Wire icing galloping amplitude prediction method, system and device and medium
By building a test platform for conductor icing galloping, generating training data, and constructing a random forest prediction model, the problem of quantitative prediction of conductor icing galloping amplitude was solved, and accurate prediction of transmission line icing galloping amplitude was achieved, thus improving power grid safety.
Patent Information
- Application Number
- CN202511846836.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-09
- Publication Date
- 2026-03-03
AI Technical Summary
There is a lack of research on the quantitative prediction of conductor ice galloping amplitude in existing technologies, making it difficult to achieve accurate early warning and prediction, which affects the safe operation of the power grid.
By building a double-split conductor ice-covered galloping test platform, training data was generated, multidimensional statistical features were extracted and normalized, redundant features were removed, a random forest prediction model was constructed, and hyperparameters were optimized using random search to establish an ice-covered galloping amplitude prediction model.
It has enabled accurate prediction of the amplitude of ice-induced galloping on transmission lines, improved the early warning capability of the power grid under severe weather conditions, and enhanced the prediction accuracy and generalization ability of the model.
Smart Images
Figure CN121598019A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of conductor icing and galloping amplitude prediction technology, and more specifically, to a method, system, device and medium for predicting conductor icing and galloping amplitude. Background Technology
[0002] With the rapid development of ultra-high voltage power transmission projects in my country and the frequent occurrence of extreme weather events, the problem of conductor icing and galloping has become increasingly prominent, becoming one of the key factors restricting the safe operation of the power grid. Overhead transmission line icing and galloping is a low-frequency, large-amplitude self-excited vibration phenomenon generated by wind excitation of conductors. Its amplitude can reach hundreds of times the conductor diameter, which can easily cause serious accidents such as phase-to-phase flashover, hardware damage, conductor strand breakage, and even collapse, resulting in significant economic losses. Therefore, accurately predicting the amplitude of conductor icing and galloping is of great significance for realizing power grid disaster prevention and mitigation and ensuring power supply reliability.
[0003] Extensive research has been conducted by scholars both domestically and internationally on predicting icing and galloping of transmission lines. Early studies were primarily based on empirical formulas and physical models, such as Nigol theory and DenHartog's vertical instability mechanism. While these methods provided a theoretical foundation, they typically required precise conductor parameters and meteorological conditions, limiting their widespread applicability in practical applications. With the development of artificial intelligence technology, machine learning algorithms have gradually been applied to the field of icing and galloping prediction. For example, ANSYS software was used to simulate the dynamic response of iced four-split conductors, and an icing and galloping early warning model was constructed based on the PSO-SVM algorithm, achieving accurate prediction of whether the conductor will gallop. Another example is the use of the IPSO algorithm to optimize a BP neural network, constructing a galloping early warning model with various meteorological and line parameters as inputs, improving prediction accuracy. Furthermore, the Internet of Things (IoT) integrated fiber optic sensing and weather forecast time series data, and through adaptive weighted fusion, K-medoids clustering, and a closed-loop attention GRU network, constructed an icing transmission line galloping early warning model, achieving accurate early warning of galloping risk. However, most existing studies focus on early warning of the risk of ice-induced galloping of conductors, while there are relatively few studies on quantitative prediction of the amplitude of ice-induced galloping. Summary of the Invention
[0004] The purpose of this invention is to provide a method, system, device, and medium for predicting the amplitude of conductor icing and dancing, so as to solve the problems existing in the above-mentioned background art.
[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution: Firstly, this application provides a method for predicting the amplitude of conductor ice-induced galloping, comprising the following specific steps: An ice-galloping simulation test was conducted using a double-split conductor ice-galloping test platform, and training data was generated. The training data included tension data and tilt angle data under different ice conditions and different amplitudes. Multidimensional statistical features are extracted from each group of samples in the training data using a feature extraction function, and the extracted multidimensional statistical features are then normalized. Redundant features of the normalized multidimensional statistical features were removed by using the Pearson correlation coefficient, and the multidimensional statistical features after removing redundant features were further screened and dimensionality reduced by the feature importance method. A random forest prediction model is constructed, and the optimal hyperparameters of the random forest prediction model are determined by a random search method. The random forest prediction model with determined hyperparameters is then trained using the multidimensional statistical features of each group after dimensionality reduction and screening. The random forest prediction model that meets the training termination condition is identified as the icing galloping amplitude prediction model, and the icing galloping amplitude of the transmission line is predicted by the icing galloping amplitude prediction model.
[0006] Based on the above technical solution, the present invention can be further improved as follows.
[0007] Furthermore, the aforementioned double-split conductor icing galloping test platform includes: Two brackets, positioned facing each other; Two right-angle hanging plates, and two bracket hanging plates located at the top of the two brackets respectively; Two ball-head hanging rings, with their fixed ends connected to two right-angle hanging plates respectively; Two first insulator strings are connected to the suspension ends of two ball-head hanging rings, respectively; The conductor, with its two ends clamped to the ends of the two first insulator strings by tension clamps; Multiple weights are evenly distributed on the wire between the two supports; The tension and tilt sensor is fixed above the middle of the two brackets and connected to the middle of the conductor through the second insulator string; An accelerometer is mounted at the suspension point of a wire.
[0008] Furthermore, the aforementioned tension and tilt sensors are used to measure the mechanical response characteristics of conductor ice galloping. The tension sensor has a measurement range of 0-50kN and a sensitivity of ±7.5N; the tilt sensor has a measurement range of -45° to +45° and a sensitivity of ±0.01°.
[0009] Furthermore, the aforementioned accelerometer is used to monitor amplitude changes during the conductor icing galloping simulation test, with a measurement range of ±40g, a frequency response of 20-2000Hz, and a sensitivity of 62.5mV / g.
[0010] Furthermore, the required counterweight mass in the above-mentioned icing dance simulation test is calculated using the following formula: ; In the formula, Let be the required counterweight mass, and let represent the mass of ice accumulation during the simulation process. The density of ice is taken as 0.9 g / cm³; d The value represents the icing thickness, and D is the diameter of the conductor, so as to obtain the counterweight mass per unit length corresponding to different icing thicknesses.
[0011] Furthermore, the above normalization process is specifically as follows: ; In the formula, These are the multidimensional statistical features after normalization. These are unnormalized multidimensional statistical features. The mean of the data. denoted as the standard deviation of the data.
[0012] Furthermore, the multidimensional statistical features after redundant feature removal are screened and dimensionality reduced using the feature importance method, specifically as follows: Traverse all decision tree nodes and obtain the impurity of each node; Calculate the reduction in Gini caused by this split; The reduction is weighted according to the node weights to obtain the node's contribution to the feature; The importance of each feature is calculated based on the contribution of each node to each feature.
[0013] Secondly, this application provides a conductor icing galloping amplitude prediction system, applied to any of the conductor icing galloping amplitude prediction methods in the first aspect, comprising: The training data acquisition module is used to conduct ice-galloping simulation tests and generate training data through the constructed double-split conductor ice-galloping test platform. The training data includes tension data and tilt angle data under different ice conditions and different amplitudes. The data feature processing module is used to extract multidimensional statistical features from each group of samples in the training data through the feature extraction function, and to normalize the multidimensional statistical features extracted from each group of samples. The multidimensional feature dimensionality reduction module is used to remove redundant features of the normalized multidimensional statistical features by using the Pearson correlation coefficient, and to filter and reduce the multidimensional statistical features after removing redundant features by using the feature importance method. The prediction model building module is used to build a random forest prediction model. It determines the optimal hyperparameters of the random forest prediction model through a random search method, and trains the random forest prediction model with determined hyperparameters using the multidimensional statistical features of each group after dimensionality reduction and screening. The icing amplitude prediction module is used to identify the random forest prediction model that meets the training termination condition as the icing icing amplitude prediction model, and to predict the icing icing amplitude of the transmission line through the icing icing amplitude prediction model.
[0014] Thirdly, this application provides an electronic device, including: at least one processor, at least one memory, and a data bus; In this system, the processor and the memory communicate with each other via a data bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute a method for predicting the amplitude of conductor icing shedding, as described in any of the first aspects.
[0015] Fourthly, this application provides a non-transitory computer-readable storage medium that stores computer instructions that cause a computer to execute a method for predicting the amplitude of conductor icing shedding as described in any of the first aspects.
[0016] Compared with the prior art, the present invention has at least the following beneficial effects: In this application, a test platform for ice galloping of double-split conductors was built to conduct experiments simulating the mechanical response characteristics of conductors under ice thickness of 0-10 mm and galloping amplitude of 0-1 m. The obtained tension and tilt angle sensing data were used to extract multidimensional important feature parameters and perform dimensionality reduction using the feature importance method. The dimensionality-reduced parameters were used as input and the ice galloping amplitude was used as output to establish a prediction model for the amplitude of ice galloping of transmission lines based on random forest (RF). The hyperparameters of the model were optimized using the random search (RS) method. Finally, the accuracy of the model prediction was verified through a test set, demonstrating that it can effectively predict the amplitude of ice galloping of transmission lines. Attached Figure Description
[0017] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the prediction method in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the double-split conductor icing galloping test platform in an embodiment of the present invention; Figure 3 This is a Pearson correlation analysis result diagram of various features related to the tension tilt angle in the embodiments of the present invention; Figure 4This is a flowchart illustrating the calculation of feature importance in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the importance of features in an embodiment of the present invention; Figure 6 This is an accuracy graph of the model under different training sample sizes in the embodiments of the present invention.
[0018] In the diagram: 1. Bracket; 2. Right-angle hanging plate; 3. Ball head hanging ring; 4. First insulator string; 5. Tension clamp; 6. Conductor; 7. Weight; 8. Tension and tilt sensor; 9. Accelerometer; 10. Second insulator string. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0021] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0022] In the description of the embodiments of the present invention, "multiple" means at least two.
[0023] Example 1: To address the issue that most existing research focuses on early warning of conductor icing galloping risk, with relatively little research on quantitative prediction of icing galloping amplitude, this example provides a method for predicting conductor icing galloping amplitude, such as... Figure 1 As shown, the specific steps include the following: S1, through the construction of a double-split conductor icing galloping test platform, conduct icing galloping simulation tests and generate training data, including tension data and tilt angle data under different icing conditions and different amplitudes.
[0024] The structural schematic diagram of the aforementioned double-split conductor icing galloping test platform can be found in [reference needed]. Figure 2 The platform includes: Two brackets 1 are arranged facing each other; Two right-angle hanging plates 2 and two brackets 1 are respectively located at the top of the two brackets 1; Two ball-head hanging rings 3, the fixed ends of the two ball-head hanging rings 3 are respectively connected to two right-angle hanging plates 2; Two first insulator strings 4 are respectively connected to the suspension ends of two ball-head hanging rings 3; The conductor 6 has its two ends clamped to the ends of the two first insulator strings 4 by tension clamps 5 respectively. Multiple weights 7 are evenly distributed on the wires 6 between the two supports 1; The tension and tilt sensor 8 is fixed above the middle of the two brackets 1 and connected to the middle of the wire 6 through the second insulator string 10; Accelerometer 9 is mounted at the suspension point of wire 6.
[0025] Among them, there are two stents 1, such as Figure 2 As shown, a conductor 6 is set between two supports 1. A right-angle hanging plate 2 is set on the top of the support 1. A ball head hanging ring 3 is set on the right-angle hanging plate 2. A first insulator string 4 is connected to one side of the ball head hanging ring 3. Tension clamps 5 are connected to both ends of the conductor 6. One end of the tension clamp 5 is connected to the first insulator string 4. A weight 7 is set on the conductor 6. A tension and tilt sensor 8 is set in the middle of the conductor 6. An acceleration sensor 9 is set on one side of the conductor 6 located on the weight 7.
[0026] Specifically, the icing galloping simulation test conducted using the double-split conductor 6 icing galloping test platform includes the following steps: Step 1: The two ends of the conductor 6 are fixed with tension clamps 5 and connected to the attachment points of the high-strength brackets 1 at both ends of the platform through the first insulator string 4 and the right-angle hanging plate 2; Step 2: The tension on conductor 6 can be adjusted by the ball-head hanging ring 3 fixed by tension clamp 5. The sag rate corresponding to the simulation experiment is 5.5%. Step 3: The platform is equipped with tension and tilt sensors 8 and acceleration sensors 9. The tension and tilt sensors 8 are installed above the second insulator string 10 in the middle of the conductor 6 to record the local angle deflection and tension change of the conductor 6 during the ice-covered dancing process. Step 4: The circuit layout, dimensional parameters, and mechanical properties of the simulation platform are kept consistent with the actual circuit. Step 5: Use uniformly distributed weights 7 to simulate the additional mass generated by ice on the conductor 6. The amplitude simulation is adjusted by the limiter to precisely control the vertical lifting of the conductor 6. Step 6: Collect the tension and tilt data obtained by the tension and tilt sensor 8 each time, as well as the weight and amplitude data of the weight 7, to obtain multiple sets of data to form training data.
[0027] Specifically, the aforementioned tension and tilt sensors are used to measure the mechanical response characteristics of conductor icing and galloping. The tension sensor has a measurement range of 0-50kN and a sensitivity of ±7.5N; the tilt sensor has a measurement range of -45° to +45° and a sensitivity of ±0.01°; and the accelerometer is used to monitor the amplitude change during the conductor icing and galloping simulation test, with a measurement range of ±40g, a frequency response of 20-2000Hz, and a sensitivity of 62.5mV / g.
[0028] Optionally, a uniformly distributed counterweight is used to simulate the additional mass generated by icing on the conductor; to simplify the calculation, the shape of the iced conductor is approximated as a hollow cylinder; the counterweight mass required for the icing galloping simulation test is calculated using the following formula: ; In the formula, Let be the required counterweight mass, and let represent the mass of ice accumulation during the simulation process. The density of ice is taken as 0.9 g / cm³; d The denoting element represents the icing thickness, and D is the conductor diameter, to obtain the counterweight mass per unit length corresponding to different icing thicknesses; the conductor counterweight relationship under the condition of icing thickness of 0-10mm is shown in Table 1: Table 1 Ice thickness t (mm) Mass of counterweight per unit length (expressed in scientific notation, g / m) 1 7.01×10 2 <![CDATA[1.46×10 2 ]]> 3 <![CDATA[2.27×10 2 ]]> 4 <![CDATA[3.14×10 2 ]]> 5 <![CDATA[4.07×10 2 ]]> 6 <![CDATA[5.06×10 2 ]]> 7 <![CDATA[6.10×10 2 ]]> 8 <![CDATA[7.19×10 2 ]]> 9 <![CDATA[8.35×10 2 ]]> 10 <![CDATA[9.56×10 2 ]]>
[0029] The amplitude simulation can be adjusted using limiters to precisely control the vertical lifting of the conductor. The conductor is lifted in segments within a range of 0-1m, with each adjustment gradient being 0.1m.
[0030] S2 extracts multidimensional statistical features from each group of samples in the training data using a feature extraction function, and normalizes the multidimensional statistical features extracted from each group of samples.
[0031] Specifically, for the raw data obtained by the tension and tilt sensors, the extract_comprehensive_features function in Python can be used to extract data features, converting the time series data of each sensor channel into 50-dimensional statistical feature data. The features can include: time-domain features such as mean, standard deviation, peak value, skewness, and kurtosis; frequency-domain features such as dominant frequency, spectral energy, and spectral centroid; and nonlinear features such as Lyapunov exponent and Hurst exponent.
[0032] Specifically, the above normalization process is as follows: ; In the formula, These are the multidimensional statistical features after normalization. These are unnormalized multidimensional statistical features. The mean of the data. The standard deviation of the data is used for normalization. Normalization can prevent data with large or small feature values from causing an imbalance in the training of machine learning models, ensuring that the algorithm can fully learn the importance of each feature and speed up the convergence.
[0033] S3 uses the Pearson correlation coefficient to remove redundant features from the normalized multidimensional statistical features, and then uses the feature importance method to filter and reduce the dimensionality of the multidimensional statistical features after removing redundant features.
[0034] To further improve model computational efficiency and identify correlations between feature data, the Pearson correlation coefficient can be used to perform correlation analysis on the normalized 50-dimensional statistical feature data to remove redundant features. The Pearson correlation coefficient is calculated as follows: ; In the formula, The Pearson correlation coefficient is used. Let X and Y be the covariances. , are the standard deviations of features X and Y, respectively; where, Figure 3 The results of Pearson correlation analysis among the eight features are presented. The correlation coefficients among most features range from -0.36 to 0.95, indicating that the overall linear correlation among the variables is weak and there is no serious multicollinearity problem. Some variables show a high positive correlation (r≈0.95), suggesting that they may reflect similar statistical characteristics, such as "wavelet coefficient level 1 energy" and "wavelet coefficient level 1 standard deviation". Meanwhile, some variables show a slight negative correlation (r≈-0.36), such as "wavelet coefficient level 0 mean" and "Y-axis tilt peak-to-peak value".
[0035] Optionally, after removing strongly correlated feature data, the feature importance (FI) method in the RF algorithm can be used to further filter and reduce the dimensionality of the feature data, such as... Figure 4 As shown, the above-mentioned dimensionality reduction processing of multidimensional statistical features after redundant feature removal using the feature importance method can be achieved through the following steps: S31, Start, traverse all decision tree nodes and obtain the node impurity Gini(t).
[0036] S32, calculate the reduction in Gini caused by this split, ΔGini(j,t).
[0037] S33, weight the reduction according to the node weights to obtain the node's contribution to the feature.
[0038] S34, calculate the importance MDIj of the j-th feature, and end.
[0039] The larger the MDI value of a feature, the more significant the feature importance; the MDI calculation results are as follows: Figure 5 As shown, the peak importance of tension and Y-axis tilt angle are 0.575 and 0.2329, respectively, contributing the most to the model's predictions; other features, such as the standard deviation of wavelet coefficients at level 0 and level 1, are 0.0103 and 0.0105, respectively, contributing less. RF is an ensemble learning method that constructs multiple decision trees and combines their predictions to achieve a regression task. Each decision tree is trained on a different subset of the data, which is randomly sampled from the original training data using Bootstrap sampling. During the construction of the decision trees, not only are random training samples used, but also features are randomly selected to split nodes, thereby reducing the correlation between the trees. Finally, RF makes a decision by averaging the predictions of multiple decision trees. The advantage of this model is that by integrating multiple weak learners, it can significantly improve accuracy and reduce the risk of overfitting. Compared to a single decision tree, RF has stronger robustness and generalization ability. Although RF has strong predictive power and can reduce overfitting, overfitting can still occur when there are too many trees or too deep trees.
[0040] S4. Construct a random forest prediction model, determine the optimal hyperparameters of the random forest prediction model through a random search method, and train the random forest prediction model with determined hyperparameters using the multidimensional statistical features of each group after dimensionality reduction and screening.
[0041] To address the bottleneck issues encountered by RF in the prediction process, the Random Search (RS) algorithm can be used to optimize its parameters. RS can optimize key parameters such as the number, depth, and feature selection strategy of decision trees, thereby maintaining the model's predictive ability while preserving its computational efficiency. The basic principle of RS is to randomly select solutions within a predefined search space and evaluate their performance. It does not depend on the specific structure of the problem, thus having strong versatility. By continuously generating random solutions, RS can effectively avoid local optima and perform a global search.
[0042] Specifically, RS can be used to randomly sample 150 sets of candidate parameter combinations from the defined parameter distribution, and 6-fold cross-validation can be used to evaluate the performance of each set of parameters. Finally, the best hyperparameter combination is automatically selected from 400 model training iterations. For example, the optimized RF model parameters are: 100 trees, 10 maximum tree depth, 4 minimum split samples per node, 1 minimum sample per leaf node, 0.8 maximum feature ratio, 1 bootstrap sampling ratio, and 0.0 impurity reduction threshold, using the bootstrap sampling method.
[0043] S5. The random forest prediction model that meets the training termination condition is identified as the icing galloping amplitude prediction model, and the icing galloping amplitude of the transmission line is predicted by the icing galloping amplitude prediction model.
[0044] The process involves establishing RF and RS-RF prediction models for the icing galloping amplitude of transmission lines within a Python 3.12 compilation environment. This process can be divided into three stages: Stage 1: Creating a database containing real-time data collected by tension and tilt sensors under different icing conditions and amplitudes, with a randomized allocation of training and test sets; Stage 2: Creating the prediction model. The prediction model first reads the training set from the database and uses the FI method to obtain an input matrix consisting of the first eight principal components. An RF prediction model is then established, and the optimal hyperparameters are determined using the RS method to establish the RS-RF prediction model; Stage 3: Based on the obtained optimal prediction model, i.e., the icing galloping amplitude prediction model, it is used to predict the icing galloping amplitude of the transmission lines.
[0045] In the above, the RF and RS-RF prediction models exhibit different adaptability and training performance with different training sample sizes; the changes in prediction accuracy of the two models are as follows: Figure 6 As shown, the prediction accuracy of both models increases with the increase of the number of training samples. When the number of training samples is small, the prediction accuracy of both models is very low and does not meet the prediction requirements. When the number of training samples reaches 225 groups, the accuracy of the RF and RS-RF models are 83.45% and 91.33%, respectively. After optimizing the RF model with RS, the prediction accuracy of the model is significantly improved.
[0046] Blind sample validation was performed on the amplitude prediction capabilities of the two models. 400 data sets were divided into training and test sets in a 70:30 ratio. The training set was used for model training to predict the amplitude of icing dance; the test set was used to validate the model accuracy. Twenty data sets were randomly selected from the test set, and the predicted values and prediction differences are shown in Table 2.
[0047] Table 2 Serial Number Test value / (m) RF prediction value / (m) RS-RF predicted value / (m) 1 0.5 0.6307 0.3770 2 0.7 0.7105 0.7352 3 0.4 0.3095 0.3162 4 0.2 0.2410 0.1443 5 0.9 0.7446 0.8846 6 0.5 0.9101 0.6240 7 0.2 0.6840 0.1451 8 0.3 0.7685 0.2281 9 0.4 0.1745 0.3614 10 0.6 0.9490 0.6656 11 0.2 0.3050 0.1894 12 0.5 0.6095 0.4598 13 0.8 0.4715 0.8294 14 0.1 0.4975 0.1015 15 0.8 0.9150 0.7965 16 0.8 0.9365 0.7364 17 0.2 0.3050 0.2691 18 0.5 0.2200 0.5242 19 0.3 0.3720 0.2991 20 0.2 0.3050 0.2037
[0048] As shown in Table 2, the predicted values of the conductor icing galloping amplitude prediction model based on RS-RF are closer to the actual values; Table 3 compares the R values of the models. 2 Evaluation metrics such as coefficient of determination (RMSE) and mean square error (MAE) show that the RS-RF prediction model outperforms the RF model in all three metrics. Therefore, the RS-RF-based transmission line ice-galling amplitude prediction model has better prediction performance.
[0049] Table 3 result RF prediction model RS-RF prediction model <![CDATA[R 2 ]]> 0.8388 0.9228 RMSE 0.1136 0.0788 MAE 0.0855 0.0596
[0050] This embodiment studies the amplitude response of conductors to icing galloping based on a transmission line icing galloping simulation experimental platform. By analyzing the relationship between the 8-dimensional features of 400 sets of amplitude galloping tension-tilt angle variation data (0-10mm (gradient 1mm) icing and 0-1m (gradient 0.1m)) and the conductor icing galloping amplitude, an RS-RF-based conductor icing galloping amplitude prediction model was established and compared with an RF-based algorithm. The RS-RF model outperforms the RF model, with a prediction accuracy of 92.28%, RMSE of 0.0788, and MAE of 0.0596. Therefore, monitoring icing galloping status based on tension-tilt angle sensors combined with machine learning algorithms can also meet the requirements; thus providing a new approach for monitoring transmission line icing galloping, achieving accurate prediction of icing galloping while reducing sensor weight, and improving the early warning capability of transmission lines under severe weather conditions.
[0051] Example 2: This application provides a conductor icing galloping amplitude prediction system, applied to the conductor icing galloping amplitude prediction method of Example 1, including: The training data acquisition module is used to conduct ice-galloping simulation tests and generate training data through the constructed double-split conductor ice-galloping test platform. The training data includes tension data and tilt angle data under different ice conditions and different amplitudes. The data feature processing module is used to extract multidimensional statistical features from each group of samples in the training data through the feature extraction function, and to normalize the multidimensional statistical features extracted from each group of samples. The multidimensional feature dimensionality reduction module is used to remove redundant features of the normalized multidimensional statistical features by using the Pearson correlation coefficient, and to filter and reduce the multidimensional statistical features after removing redundant features by using the feature importance method. The prediction model building module is used to build a random forest prediction model. It determines the optimal hyperparameters of the random forest prediction model through a random search method, and trains the random forest prediction model with determined hyperparameters using the multidimensional statistical features of each group after dimensionality reduction and screening. The icing amplitude prediction module is used to identify the random forest prediction model that meets the training termination condition as the icing icing amplitude prediction model, and to predict the icing icing amplitude of the transmission line through the icing icing amplitude prediction model.
[0052] Example 3: This application provides an electronic device, including: at least one processor, at least one memory, and a data bus; The processor and memory communicate with each other via a data bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute a method for predicting the amplitude of conductor icing dance, as in Example 1.
[0053] Example 4: This application provides a non-transitory computer-readable storage medium that stores computer instructions, which cause the computer to execute a method for predicting the amplitude of conductor icing and dancing according to Example 1.
[0054] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0055] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will 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 program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0056] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0057] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0058] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.
[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for predicting the amplitude of conductor icing galloping, characterized in that, The specific steps include the following: An ice-galloping simulation test was conducted using a double-split conductor ice-galloping test platform, and training data was generated. The training data included tension data and tilt angle data under different ice conditions and different amplitudes. Multidimensional statistical features are extracted from each group of samples in the training data using a feature extraction function, and the extracted multidimensional statistical features are then normalized. Redundant features of the normalized multidimensional statistical features were removed by using the Pearson correlation coefficient, and the multidimensional statistical features after removing redundant features were further screened and dimensionality reduced by the feature importance method. A random forest prediction model is constructed, and the optimal hyperparameters of the random forest prediction model are determined through rescaled range analysis. The random forest prediction model with determined hyperparameters is then trained using the multidimensional statistical features of each group after dimensionality reduction and screening. The random forest prediction model that meets the training termination condition is identified as the icing galloping amplitude prediction model, and the icing galloping amplitude of the transmission line is predicted by the icing galloping amplitude prediction model.
2. The method for predicting the amplitude of conductor icing galloping according to claim 1, characterized in that, The double-split conductor icing galloping test platform includes: Two supports, the two supports being arranged facing each other; Two right-angled hanging plates, the two bracket hanging plates are respectively located at the top of the two brackets; Two ball-head hanging rings, the fixed ends of the two ball-head hanging rings are respectively connected to the two right-angle hanging plates; Two first insulator strings are respectively connected to the suspension ends of the two ball-head hanging rings; The conductor, the two ends of which are respectively clamped to the ends of the two first insulator strings by tension clamps; Multiple weights are evenly distributed on a wire between two supports; The tension and tilt sensor is fixed above the middle of the two brackets and connected to the middle of the conductor through a second insulator string; An accelerometer is mounted at the suspension point of the wire.
3. The method for predicting the amplitude of conductor icing galloping according to claim 2, characterized in that, The tension and tilt sensors are used to measure the mechanical response characteristics of the conductor's ice-covered galloping. The tension sensor has a measurement range of 0-50kN and a sensitivity of ±7.5N; the tilt sensor has a measurement range of -45° to +45° and a sensitivity of ±0.01°.
4. The method for predicting the amplitude of conductor icing galloping according to claim 2, characterized in that, The accelerometer is used to monitor the amplitude change during the conductor icing dance simulation test. The measurement range is ±40g, the frequency response is 20-2000Hz, and the sensitivity is 62.5mV / g.
5. The method for predicting the amplitude of conductor icing galloping according to claim 1, characterized in that, The required counterweight mass in the ice-covered dancing simulation test is calculated using the following formula: ； In the formula, Let represent the required counterweight mass, and let represent the mass of ice accumulation during the simulation process. The density of ice is taken as 0.9 g / cm³; d represents the ice thickness, and D is the diameter of the wire, so as to obtain the counterweight mass per unit length corresponding to different ice thicknesses.
6. The method for predicting the amplitude of conductor icing galloping according to claim 1, characterized in that, The normalization process specifically involves: ； In the formula, These are the multidimensional statistical features after normalization. These are unnormalized multidimensional statistical features. The mean of the data. denoted as the standard deviation of the data.
7. The method for predicting the amplitude of conductor icing galloping according to claim 1, characterized in that, The process of filtering and reducing the dimensionality of multidimensional statistical features after removing redundant features using the feature importance method is as follows: Traverse all decision tree nodes and obtain the impurity of each node; Calculate the reduction in Gini caused by this split; The reduction is weighted according to the node weights to obtain the node's contribution to the feature; The importance of each feature is calculated based on the contribution of each node to each feature.
8. A system for predicting the amplitude of conductor icing and galloping, characterized in that, include: The training data acquisition module is used to conduct ice-galloping simulation tests and generate training data through the constructed double-split conductor ice-galloping test platform. The training data includes tension data and tilt angle data under different ice conditions and different amplitudes. The data feature processing module is used to extract multidimensional statistical features from each group of samples in the training data through the feature extraction function, and to normalize the multidimensional statistical features extracted from each group of samples. The multidimensional feature dimensionality reduction module is used to remove redundant features of the normalized multidimensional statistical features by using the Pearson correlation coefficient, and to filter and reduce the multidimensional statistical features after removing redundant features by using the feature importance method. The prediction model building module is used to build a random forest prediction model. It determines the optimal hyperparameters of the random forest prediction model through rescaled range analysis and trains the random forest prediction model with determined hyperparameters using the multidimensional statistical features of each group after dimensionality reduction and screening. The icing amplitude prediction module is used to identify the random forest prediction model that meets the training termination condition as the icing icing amplitude prediction model, and to predict the icing icing amplitude of the transmission line through the icing icing amplitude prediction model.
9. An electronic device, characterized in that, include: At least one processor, at least one memory, and a data bus; The processor and the memory communicate with each other via the data bus. The memory stores program instructions that can be executed by the processor, which calls the program instructions to execute a method for predicting the amplitude of conductor icing shedding as described in any one of claims 1-7.
10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the method for predicting the amplitude of conductor icing shedding as described in any one of claims 1-7.