An overhead line galloping warning system and method based on machine learning model

Abstract

The present invention proposes an overhead line galloping warning system and method based on a machine learning model, which relate to the field of electric power technology. The method comprises the following steps: setting a plurality of monitoring nodes on the overhead line; obtaining a weight coefficient of each node according to the galloping data of each node; obtaining a first risk of each node within a preset time according to the galloping data of each node and the corresponding environmental parameters; establishing a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predicting the line galloping risk within a preset time according to the line dynamic prediction model; issuing a galloping warning and providing decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk, and by introducing a weight distribution mechanism, establishing a dynamic prediction model, optimizing the preset time setting, and improving the warning mechanism and other measures, real-time monitoring, accurate prediction, and timely warning of the overhead line galloping risk are achieved.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to an overhead line galloping warning system and method based on a machine learning model. Background Art

[0002] In the current operation and maintenance of power systems, the galloping problem of overhead lines has always been an important factor affecting the safety and stability of the power grid. Traditional galloping warning measures mainly rely on manual inspections and static analysis, which have obvious limitations. First, manual inspections are not only time-consuming and labor-intensive, but also difficult to cover all possible galloping risk points, resulting in untimely warnings. Second, static analysis methods are usually based on historical data or empirical formulas, which cannot reflect changes in the line galloping state in real time, so the prediction accuracy is limited.

[0003] With the advancement of technology, especially the rapid development of sensor technology, big data analysis and artificial intelligence algorithms, new solutions have been provided for overhead line galloping warning. However, although the existing technology has been able to monitor and predict the galloping state of the line to a certain extent, there are still the following defects: the importance of the galloping data and environmental parameters of different nodes in the overall risk prediction is often different, but the existing technology often adopts a unified weight distribution method, which cannot accurately reflect the contribution of each node to the overall risk, resulting in inaccurate prediction results; most existing prediction models are based on fixed algorithms and parameters, and cannot dynamically adjust the prediction strategy according to real-time data and environmental changes, so it is difficult to adapt to the complex and changeable actual operation conditions; the setting of the preset time is crucial to the accuracy of the prediction results. The existing technology usually sets the preset time according to experience or fixed rules, ignoring the influence of historical data, current environmental parameters and line importance on the preset time, resulting in the lack of pertinence and timeliness of the prediction results. The warning mechanism of the existing technology is often based only on a single threshold judgment, lacks multi-level warning levels and corresponding decision support, and is difficult to meet the actual needs of power grid operation and maintenance. Summary of the invention

[0004] In response to the above-mentioned defects, the purpose of this application is to provide an overhead line galloping warning system and method based on a machine learning model. By introducing a scientific weight distribution mechanism, establishing a dynamic prediction model, optimizing the preset time setting, and improving the warning mechanism, real-time monitoring, accurate prediction and timely warning of overhead line galloping risks can be achieved, providing strong guarantees for the safe and stable operation of the power grid.

[0005] The purpose of this application is achieved by the following technical solutions:

[0006] The present application provides an overhead line galloping early warning method based on a machine learning model, the method comprising:

[0007] S1. Set up multiple monitoring nodes on the overhead line;

[0008] S2. Obtain the weight coefficient of each node according to the dancing data of each node; obtain the first risk of each node within the preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within the preset time according to the line dynamic prediction model; wherein, obtain the average dancing time of the line through the historical dancing data; dynamically adjust and determine the preset time according to the current environmental parameters, the line importance score and the average dancing time of the line;

[0009] S3. issuing a galloping warning and providing decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk.

[0010] Preferably, the S1 includes:

[0011] Based on environmental parameters, topography and vegetation coverage, cluster analysis is performed on the areas where overhead lines pass through, and multiple monitoring sub-areas are divided according to the cluster analysis results;

[0012] A monitoring node is set up in each monitoring sub-area.

[0013] Preferably, S2 includes:

[0014] Acquire first dance data of each node in real time through a dance sensor, wherein the first dance data includes a first dance time, a first dance duration, a first dance amplitude, and a first dance frequency;

[0015] Obtaining a weight coefficient of each node according to the first dance data of each node;

[0016] According to the first dancing data of each node and the environmental parameters of each node, the first risk of each node within a preset time is obtained;

[0017] A line dynamic prediction model is established through the weight coefficient of each node and the environmental parameters of each node;

[0018] Collect multiple sets of continuous real-time data and input them into the line dynamic prediction model to predict the risk of line dancing within a preset time.

[0019] Preferably, the weight coefficient of each node is obtained by the following formula:

[0020]

[0021] in, w i For the i The weight coefficient of each node; Sij For the i Node j Duration of the first dance of the dance; v ij No. i Node j The amplitude of the first dance of the second dance; p ij For the i Node j The first dance frequency of the second dance; f () is the functional relationship; Z ij For the i Node j The overall score of the dance; n is the number of nodes, m The number of dances in the statistical time period; H i For Node i Importance rating; G ij For the i The node in j The median value in the dance.

[0022] Preferably, the step of obtaining the first risk of each node within a preset time according to the first dancing data of each node and the environmental parameters of each node comprises:

[0023] Establishing a first risk model through the first dancing data of each node and the environmental parameters of each node;

[0024] The first dance data includes a plurality of first dance durations, a plurality of first dance amplitudes, a plurality of first dance frequencies, a change in dance duration, a change in dance amplitude, and a change in dance frequency of nodes within a first statistical time period;

[0025] The environmental parameters include the maximum value, minimum value and average value of the environmental parameter change rate of each node dance; the environmental parameter change rate includes the temperature change rate, humidity change rate, wind direction change rate, and the average angle change rate between the wind direction and the node line;

[0026] A first risk of each node is obtained through a first risk model, where the first risk includes a first risk probability and a first dancing intensity.

[0027] Preferably, the first risk model is established by using the first dancing data of each node and the environmental parameters of each node; comprising:

[0028] Perform correlation analysis on each environmental parameter and the first dancing data to obtain each first correlation coefficient;

[0029] If the absolute value of the first correlation coefficient between the environmental parameter and any dancing data is greater than a first preset threshold, the environmental parameter is used as an input of the first risk model;

[0030] Among them, the first preset threshold is:

[0031]

[0032] in, D y is the first preset threshold, D a is the mean of the absolute values ​​of the first correlation coefficients; s is the standard deviation of the means of the first correlation coefficients; a is the adjustment coefficient, 1≤ a ≤5; k Score the importance of the line, 0< k <1.

[0033] Preferably, establishing a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node includes:

[0034] According to the weight coefficient of each node, the environmental parameters of multiple nodes in the same line in the same time period and the dancing data of each node are weighted averaged to obtain the weighted average value of each environmental parameter corresponding to the line and the average dancing data of the line;

[0035] A line dynamic prediction model is established by taking the weighted average values ​​of the environmental parameters of the line, the average dancing data of the corresponding line, and the changes of the weighted average values ​​of the environmental parameters of the corresponding line for multiple consecutive times and the changes of the average dancing data.

[0036] Preferably, the method for setting the preset time is as follows:

[0037] The average dance duration of the line is obtained through historical dance data;

[0038] Determine the preset time based on the current environmental parameters, the importance score of the line and the average dancing time of the line;

[0039] The preset time is:

[0040]

[0041]

[0042] in, T y The current preset time length for the line, T a is the average dancing time of the line, Ec is the rate of change of the current environmental parameters of the line; E a is the historical average environmental parameter change rate of the line; C h The current prediction model h The maximum rate of change of an input environmental parameter; X h For the prediction model h The weight of the input environmental parameters; v The number of environmental parameters input to the prediction model; C ha For the prediction model h The average value of the maximum change rate of the input environmental parameters within the statistical time period; k Score the importance of the line, 0< k <1; β is the adjustment coefficient, 0< β <1.

[0043] Preferably, issuing a warning of galloping and providing decision support according to the prediction results include:

[0044] If the first risk probability of any node is greater than its preset threshold, or the line galloping risk probability is greater than its preset threshold, a first warning is issued, and the first warning is first graded according to the predicted galloping intensity; the warning level will be further classified according to the predicted galloping intensity;

[0045] If the first risk probabilities of multiple nodes on a line are all greater than their preset thresholds, a second warning will be issued; the level of the warning will be further classified based on the predicted intensity of the oscillation.

[0046] The present application provides a system for an overhead line galloping early warning method based on a machine learning model described in the present application, the system comprising:

[0047] A node setting module, used to set multiple monitoring nodes on the overhead line;

[0048] The prediction module is used to obtain the weight coefficient of each node according to the dancing data of each node; obtain the first risk of each node within a preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within a preset time according to the line dynamic prediction model; wherein, the average dancing time of the line is obtained through the historical dancing data; and dynamically adjust and determine the preset time according to the current environmental parameters, the line importance score and the average dancing time of the line;

[0049] The early warning module is used to issue a galloping warning and provide decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk.

[0050] The beneficial effects of the present invention include: by introducing a scientific weight distribution mechanism, the method can fully consider the importance of different nodes in the overall risk prediction, making the prediction results more accurate. At the same time, the establishment of a dynamic prediction model enables the prediction strategy to be dynamically adjusted according to real-time data and environmental changes, further improving the accuracy of the prediction; by real-time monitoring of the dancing data and environmental parameters of each node, combined with the dynamic prediction model, it is possible to achieve real-time tracking and prediction of line dancing risks; by comprehensively considering historical data, current environmental parameters, and line importance scores and other factors, it is possible to set the preset time scientifically and reasonably, so that the prediction results are more in line with the actual situation, and improve the pertinence and timeliness of the warning. Not only is the warning threshold set, but also multi-level warning levels and corresponding decision support are provided according to the prediction results; it helps power grid operation and maintenance personnel to take corresponding countermeasures according to the warning level and reduce the impact of dancing risks on power grid security. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of an overhead line dancing warning method based on a machine learning model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new embodiment.

[0053] See also Figure 1 The embodiment of the present application provides an overhead line galloping warning method based on a machine learning model, and the method comprises:

[0054] S1. Set up multiple monitoring nodes on the overhead line;

[0055] S2. Obtain the weight coefficient of each node according to the dancing data of each node; obtain the first risk of each node within the preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within the preset time according to the line dynamic prediction model; wherein, obtain the average dancing time of the line through the historical dancing data; dynamically adjust and determine the preset time according to the current environmental parameters, the line importance score and the average dancing time of the line;

[0056] S3. issuing a galloping warning and providing decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk.

[0057] In some embodiments, the S1 includes:

[0058] Based on environmental parameters, topography and vegetation coverage, cluster analysis is performed on the areas where overhead lines pass through, and multiple monitoring sub-areas are divided according to the cluster analysis results;

[0059] A monitoring node is set up in each monitoring sub-area.

[0060] Environmental parameters include, but are not limited to, wind speed and direction, temperature, humidity, snow and ice load, rainfall, and topographic features.

[0061] The dancing data include the date and time when the dancing occurs, the intensity and duration of the dancing, the frequency and amplitude of the dancing;

[0062] Galloping data also includes: the impacts caused by the galloping, such as the extent of damage to the line itself or other infrastructure; as well as the mitigation measures taken and their effectiveness.

[0063] The working principle and effect of the above technical solution are as follows: first, cluster analysis is performed on the areas through which the overhead lines pass according to environmental parameters (such as wind speed, wind direction, temperature, humidity, etc.), topography (such as mountainous areas, plains, river valleys, etc.) and vegetation coverage; areas with similar characteristics are divided into different monitoring sub-areas to more accurately set monitoring nodes; in each monitoring sub-area, multiple monitoring nodes are set according to the results of cluster analysis and actual needs; these monitoring nodes are responsible for real-time collection of conductor dancing data (such as dancing amplitude, frequency, etc.) and environmental parameters (such as wind speed, wind direction, temperature, etc.) in the area.

[0064] According to the dancing data of each node, the weight coefficient of each node is calculated through a certain algorithm. These weight coefficients reflect the importance of each node in the overall risk prediction. Combined with the dancing data of each node and the corresponding environmental parameters, the first risk of each node within the preset time is calculated using statistical or machine learning methods; the first risk refers to the possibility of wire dancing at a single node in the future and the degree of its impact. A line dynamic prediction model is established through the weight coefficient of each node and the environmental parameters of each node (such as the changing trend of wind speed and wind direction, etc.); this model can comprehensively consider the influence of data and environmental factors of multiple nodes to predict the overall dancing risk of the line. The established dynamic prediction model is used to predict the dancing risk of overhead lines within the preset time. The prediction results include the probability of line dancing, the possible scope of impact, and the severity.

[0065] According to the prediction results, when the risk of line galloping is predicted to exceed the preset threshold, the system immediately issues a galloping warning. The warning information can be sent to relevant personnel in a variety of ways (such as SMS, email, APP push, etc.) so that they can take timely measures to respond. In addition to the warning information, the system also provides decision-making support functions. According to the prediction results and actual conditions, targeted suggestions and measures are provided to operation and maintenance personnel, such as strengthening inspections, adjusting operation modes, and installing anti-galloping devices, so as to reduce the impact of galloping risks on power grid security.

[0066] To sum up, the overhead line galloping warning method based on machine learning model provided in the embodiment of the present application effectively improves the ability to prevent and control the risk of overhead line galloping through real-time monitoring, dynamic prediction and early warning mechanism, and provides a strong guarantee for the safe and stable operation of the power grid.

[0067] In some embodiments, the S2 includes:

[0068] Acquire first dance data of each node in real time through a dance sensor, wherein the first dance data includes a first dance time, a first dance duration, a first dance amplitude, and a first dance frequency;

[0069] Obtaining a weight coefficient of each node according to the first dance data of each node;

[0070] According to the first dancing data of each node and the environmental parameters of each node, the first risk of each node within a preset time is obtained;

[0071] A line dynamic prediction model is established through the weight coefficient of each node and the environmental parameters of each node;

[0072] Collect multiple sets of continuous real-time data and input them into the line dynamic prediction model to predict the risk of line dancing within a preset time.

[0073] The dancing sensor in this embodiment does not refer to a specific type of sensor, but a series of sensors. The dancing sensor is usually based on inertial sensor technology, such as accelerometers, gyroscopes, or a combination of the two nine-axis sensors (including three-axis accelerometers, three-axis magnetometers, and three-axis gyroscopes). These sensors can detect the acceleration, angular velocity, and magnetic field changes of objects in space, thereby calculating the motion state of the object. The dancing sensor can monitor the dancing frequency, amplitude, and other parameters of the wire in real time, and accurately determine the dancing of the wire. The sensor uploads the monitoring data to the background center through wireless communication, so that the operation and maintenance personnel can take quick measures to avoid accidents.

[0074] The working principle of the above technical solution is:

[0075] The first dance data of each node is obtained in real time through the dancing sensors installed at each node of the overhead line. These data include the first dance time (i.e. the time when the dance starts), the first dance duration (the duration of the dance), the first dance amplitude (the maximum deviation of the conductor dance) and the first dance frequency (the number of dances per unit time). These data provide the basis for subsequent weight calculation, risk assessment and model prediction.

[0076] Based on the first dancing data of each node, the weight coefficient of each node is calculated using statistical methods or machine learning algorithms. The calculation of the weight coefficient may take into account multiple factors, such as the historical dancing frequency, amplitude, dancing duration of the node, and the importance of the node in the power grid. The weight coefficient reflects the relative importance of each node in the overall risk prediction; combined with the first dancing data of each node and the corresponding environmental parameters (such as wind speed, wind direction, temperature, humidity, etc.), the first risk of each node within a preset time is calculated through a risk assessment model. This risk assessment model can be constructed based on methods such as statistics, machine learning or expert systems, and can comprehensively consider the impact of dancing data and environmental parameters on dancing risks. A line dynamic prediction model is constructed through the weight coefficients of each node and the environmental parameters of each node. This model can capture the mutual influence between different nodes and the dynamic impact of environmental parameters on line dancing risks. The model can be constructed and optimized using methods such as time series analysis, neural networks, and support vector machines.

[0077] Continuously receive and process multiple sets of real-time data, including the galloping data and environmental parameters of each node. These real-time data are input into the established line dynamic prediction model. Based on multiple sets of continuous real-time data, the line dynamic prediction model predicts the line galloping risk within a preset time. The prediction results may include key parameters such as the probability of line galloping, the expected galloping duration, amplitude and frequency. These prediction results will be used for subsequent galloping warning and decision support.

[0078] Through the above steps, the real-time collection of overhead line galloping data at each node, the scientific calculation of weight coefficients, the accurate assessment of the first risk and the dynamic prediction of line galloping risk are realized. This process provides timely and accurate risk information for power grid operation and maintenance personnel, which helps to take effective measures to reduce the impact of galloping risks on power grid security.

[0079] In some embodiments, the weight coefficient of each node is obtained by the following formula:

[0080]

[0081] in, w i For the i The weight coefficient of each node; Sij For the i Node j Duration of the first dance of the dance; v ij No. i Node j The amplitude of the first dance of the second dance; p ij For the i Node j The first dance frequency of the second dance; f () is the functional relationship; Z ij For the i Node j The overall score of the dance; n is the number of nodes, m The number of dances in the statistical time period; H i For Node i Importance rating; G ij For the i The node in j The median value in the dance.

[0082] The working principle of the above technical solution is: first, for each node i Every dance j , according to the duration of the first dance of this dance S ij , the first dance amplitude v ij and the first dance frequency p ij , through the function f Calculate a comprehensive score Z ij ; This function f () can be a weighted sum, product, or a more complex mathematical expression, which is used to combine these three dance characteristics into a single score value to reflect the overall impact of the dance.

[0083] Each node i Each has an importance rating H i , this score reflects the relative importance of the node in the power grid. The importance score may be based on a variety of factors, such as the node's geographical location, the number of connected lines, the size of the load carried, etc. The higher the score, the more important the node.

[0084] Using the comprehensive score Z ij and node importance score H i ,H i The larger the node, the more important it is. Calculate an intermediate value G ij ; This calculation process takes into account the comprehensive score of each node dance and the importance of the node, by multiplying the number of nodes n (used for normalization) and node importance scores H i , and then divided by the sum of the comprehensive scores of all dances of all nodes in the statistical time period, to get G ij ; The purpose of this step is to combine the node's dancing characteristics and its importance to form an intermediate value that reflects the node's contribution to the overall risk.

[0085] Finally, the G ij The values ​​are added and divided by the total number of nodes in all statistical time periods. G ij The sum of the values ​​​​gets the final weight coefficient of each node w i ; This weight coefficient reflects the relative importance of the node in the overall risk prediction, that is, the degree of influence of node dancing on the overall line dancing risk.

[0086] In summary, the calculation process of this weight coefficient comprehensively considers the dancing characteristics of the node and its importance in the power grid. Through a series of mathematical operations and normalization processing, a weight coefficient that can accurately reflect the contribution of the node to the overall risk is obtained. These weight coefficients will be used in the subsequent line dynamic prediction model to improve the accuracy and reliability of the prediction results.

[0087] The effect of the above technical solution is: this method not only takes into account the dancing characteristics of the node (dancing duration S ij , dance amplitude v ij , dance frequency p ij ), and also combines the importance score of the node in the power grid H i This comprehensive consideration approach can more comprehensively reflect the contribution of nodes to the overall dancing risk, making the evaluation results more accurate and reliable.

[0088] Through the function f () The calculated comprehensive score Z ij , can quantify the comprehensive impact of each dance, avoiding the one-sidedness that may be caused by a single indicator; at the same time, Z ijand node importance score H i The combination of the two further refines the calculation of the weight coefficient and improves the accuracy of the evaluation. Since the weight coefficient is calculated based on real-time data and statistical results, it has good dynamic adaptability. As the power grid operation status changes and the dancing characteristics are updated, the weight coefficient will be adjusted accordingly to maintain the timeliness and accuracy of the evaluation results.

[0089] The accurate calculation results of the weight coefficient provide powerful decision-making support for power grid operation and maintenance personnel; by comparing the weight coefficients of different nodes, operation and maintenance personnel can give priority to high-risk nodes, take targeted prevention and response measures, and reduce the impact of dancing risks on power grid security. The calculation of weight coefficients also helps to optimize the allocation of power grid resources. When resources are limited, operation and maintenance personnel can allocate resources according to the weight coefficients to ensure that high-risk nodes receive sufficient attention and resource support, thereby improving the efficiency of resource use.

[0090] In some embodiments, obtaining the first risk of each node within a preset time according to the first dancing data of each node and the environmental parameters of each node includes:

[0091] A first risk model is established through the first dancing data of each node and the environmental parameters of each node; and iterative optimization is performed according to the predicted results and actual results;

[0092] The first dance data includes a plurality of first dance durations, a plurality of first dance amplitudes, a plurality of first dance frequencies, a change in dance duration, a change in dance amplitude, and a change in dance frequency of nodes within a first statistical time period;

[0093] The environmental parameters include the maximum value, minimum value and average value of the environmental parameter change rate of each node dance; the environmental parameter change rate includes the temperature change rate, humidity change rate, wind direction change rate, and the average angle change rate between the wind direction and the node line;

[0094] A first risk of each node is obtained through a first risk model, where the first risk includes a first risk probability and a first dancing intensity.

[0095] The working principle and effect of the above technical solution are as follows: first, collect detailed dancing data of each node in the first statistical time period, including multiple first dancing durations, multiple first dancing amplitudes, multiple first dancing frequencies, and the changes of these parameters over time (such as the changes of dancing durations, the changes of dancing amplitudes, and the changes of dancing frequencies). These data reflect the dancing characteristics and dynamic changes of the nodes over a period of time.

[0096] At the same time, the environmental parameters of the node are collected each time it dances, and the change rates of these environmental parameters are calculated, including the temperature change rate, humidity change rate, wind direction change rate, and the average angle change rate between the wind direction and the node line. The maximum, minimum, and average values ​​of the change rates of these environmental parameters are also recorded as important factors affecting the risk of dancing; by comprehensively considering multiple dancing data in the first statistical time period (including dancing duration, amplitude, frequency and its change amount), as well as the change rate of environmental parameters (including temperature, humidity, wind direction and the change rate of its angle with the line), it is possible to capture more comprehensively the various factors affecting the risk of dancing. This comprehensive data input makes the assessment results of the first risk model more accurate and can more truly reflect the actual risk level of the node.

[0097] Based on the collected dancing data and environmental parameters, a first risk model is established. This model can be a complex mathematical model or a prediction model based on machine learning. There is no specific limitation here. The core of the model is to capture the correlation between the dancing data and environmental parameters, and how these parameters jointly affect the dancing risk of the node.

[0098] In the process of building the model, a large amount of historical data can be analyzed and studied through statistical methods, data mining techniques or machine learning algorithms to find out the key factors and rules that affect dancing risks.

[0099] The collected data is divided into a training set and a test set. The training set is used to train the model, and the test set is used to evaluate the performance of the model. The model is trained using the training set data, and the prediction error is minimized by adjusting the parameters and structure of the model. The trained model is predicted using the test set data, and the predicted results are compared with the actual results to evaluate the prediction accuracy and generalization ability of the model. Based on the evaluation results, the model is iteratively optimized. If the prediction results are not ideal, the model can be improved by reselecting features, adjusting model parameters, or trying different algorithms. The iterative process may include multiple training, prediction, and evaluation until the model performance meets the requirements.

[0100] In a specific embodiment of the present invention, the first risk model is as follows:

[0101]

[0102] in,[ M ] represents the structural mass matrix at the node, [ C ] represents the structural damping matrix at the node, [ K ] represents the structural stiffness matrix at the node, m& represents the acceleration matrix at the node, & represents the node velocity array, mrepresents the node displacement matrix, { F ( t )} represents the load on the transmission tower-line; according to the output load on the transmission tower-line, the galloping risk and galloping intensity under the corresponding load are obtained by using the table lookup method;

[0103] In this embodiment, the node velocity array is a matrix containing the velocity components of all nodes in the finite element model in a specific direction. It is an important component in structural dynamics analysis and is used to describe the velocity changes of each node when the structure is subjected to external forces. In finite element analysis, the node velocity array, node displacement array, node acceleration array, mass matrix, damping matrix, stiffness matrix, etc., together constitute the basic equation group describing the dynamic response of the structure.

[0104] In practical applications, node velocity arrays can be obtained in a variety of ways. One common method is to obtain it through experimental measurement, such as using a laser vibrometer or accelerometer to directly measure the vibration velocity of the structure. Another method is to obtain it through numerical simulation calculations, such as in finite element analysis software, by inputting information such as the geometry, material properties, boundary conditions, and loads of the structure, and performing dynamic analysis to obtain the node velocity array.

[0105] It should be noted that the specific form and content of the node velocity array depends on factors such as the type of structure being studied, the purpose of analysis, and the analysis method used. Therefore, in practical applications, it is necessary to select appropriate methods and tools to obtain and process the node velocity array according to the specific situation.

[0106] By inputting the real-time or predicted first dancing data and environmental parameters of each node through the established first risk model, the model will automatically calculate the first risk of each node within the preset time. This risk includes two main aspects: the first risk probability and the first dancing intensity; the first risk probability indicates the possibility of a dancing event occurring at the node within the preset time. It comprehensively considers multiple factors such as the dancing history of the node, the current state and change trend of the environmental parameters. The first dancing intensity is used to quantify the severity or scope of the dancing event. It may be calculated based on the predicted values ​​of the dancing duration, amplitude and frequency, or it may take into account the impact of environmental parameters on the dancing intensity.

[0107] The calculated first risk (including risk probability and galloping intensity) will be output to the power grid operation and maintenance personnel or related systems. The operation and maintenance personnel can evaluate the galloping risk level of the power grid based on this information and formulate corresponding prevention and control measures.

[0108] At the same time, this risk information can also be used as input data to further optimize the grid's operating strategy, adjust equipment maintenance plans, or enhance the grid's emergency response capabilities.

[0109] In summary, this embodiment achieves accurate calculation and evaluation of the node dancing risk within a preset time by establishing a first risk model and comprehensively considering the changes in dancing data and environmental parameters. This provides strong technical support and decision-making basis for the safe operation of the power grid. In some embodiments, the first risk model is established through the first dancing data of each node and the environmental parameters of each node; including:

[0110] Perform correlation analysis on each environmental parameter and the first dancing data to obtain each first correlation coefficient;

[0111] If the absolute value of the first correlation coefficient between the environmental parameter and any dancing data is greater than a first preset threshold, the environmental parameter is used as an input of the first risk model;

[0112] Among them, the first preset threshold is:

[0113]

[0114] in, D y is the first preset threshold, D a is the mean of the absolute values ​​of the first correlation coefficients; s is the standard deviation of the means of the first correlation coefficients; a is the adjustment coefficient, 1≤ a ≤5; k Score the importance of the line, 0< k <1.

[0115] The working principle of the above technical solution is: to perform correlation analysis on each environmental parameter and each first dance data, and usually use statistical methods (such as correlation coefficient calculation) to calculate the degree of correlation between them; what is calculated here is the correlation coefficient, which can be the Pearson correlation coefficient, Spearman rank correlation coefficient, etc., depending on the nature of the data and the needs of the analysis.

[0116] Since the correlation coefficient can be positive or negative, indicating positive and negative correlation respectively, but here we are concerned with the strength of the correlation rather than the direction, so the absolute value of each correlation coefficient is used as an indicator to evaluate the strength of the correlation. Calculate the mean of the absolute values ​​of all first correlation coefficients D a , and the standard deviation σ of these absolute values. These two statistics will be used to calculate the first preset threshold later.

[0117] Using the given formula Calculate the first preset threshold D y .in, D ais the mean of the absolute values ​​of the first correlation coefficients, s is the standard deviation of these absolute values, a is the adjustment coefficient (used to adjust the tightness of the threshold, with a value range of 1≤ a ≤5), k is the line importance score (used to consider the importance of different lines in the power grid, with a value range of 0< k <1). The absolute value of the first correlation coefficient between each environmental parameter and any first dancing data is compared with the first preset threshold D y If the absolute value of the first correlation coefficient between a certain environmental parameter and at least one first dancing data is greater than D y , then it is considered that this environmental parameter has a significant impact on the dancing risk and should be selected as the input variable of the first risk model.

[0118] The effect of the above technical solution is: by selecting environmental parameters that are significantly correlated with the dancing data as model input through correlation analysis, the negative impact of irrelevant or weakly correlated variables on the model prediction performance can be reduced, thereby improving the accuracy and reliability of the model. The screened environmental parameters usually have a direct physical or chemical connection with the dancing phenomenon, which makes the results of the model easier to understand and explain. Reducing the number of input variables can reduce the complexity of the model, reduce the amount of calculation, and improve the operating efficiency of the model. This is particularly important for real-time risk assessment systems, which can ensure that accurate prediction results are given within a limited time. Introducing line importance scoring k As a regulating factor, the threshold can be adjusted according to the importance of different lines in the power grid, making the model more flexible and adaptable. For important lines, a lower threshold can be set to capture more potential risks; for minor lines, a higher threshold can be set to reduce the false alarm rate. Formula Combined with the mean of the absolute values ​​of the first correlation coefficient D a and standard deviation s , which takes into account both the average level of the overall correlation strength and the fluctuation range of the correlation strength. This combination helps to more comprehensively evaluate the correlation strength and reduce the impact of extreme values ​​on the threshold. a The introduction of allows the threshold to be adjusted according to actual conditions. a The value of can adjust the sensitivity and specificity of the model to meet different risk assessment needs while maintaining a certain accuracy. k Incorporating this into the formula enables the model to fully consider the importance differences of different lines in the power grid. This personalized threshold setting helps to improve the pertinence and effectiveness of risk assessment.

[0119] In summary, by screening input variables through association analysis and using reasonable formulas to calculate thresholds, the accuracy, interpretability, adaptability and operating efficiency of the first risk model can be significantly improved, providing strong support for the safe and stable operation of the power grid.

[0120] In some embodiments, establishing a line dynamic prediction model by using the weight coefficient of each node and the environmental parameters of each node includes:

[0121] According to the weight coefficient of each node, the environmental parameters of multiple nodes in the same line in the same time period and the dancing data of each node are weighted averaged to obtain the weighted average value of each environmental parameter corresponding to the line and the average dancing data of the line;

[0122] A line dynamic prediction model is established by taking the weighted average values ​​of the environmental parameters of the line, the average dancing data of the corresponding line, and the changes of the weighted average values ​​of the environmental parameters of the corresponding line for multiple consecutive times and the changes of the average dancing data.

[0123] In another specific embodiment of the present invention, the algorithm formula of the line dynamic prediction model is similar to the first risk model, except that the input of the line dynamic prediction model is the structural quality matrix obtained by weighted averaging the environmental parameters of each node and the dancing data of each node. M ], structural damping matrix[ C ], structural stiffness matrix [ K ], acceleration matrix m& , velocity array & and displacement matrix m ;

[0124] And then decompose and apply the corresponding load:

[0125] ;

[0126] in, m s As the tower shape factor, it is generally taken as 2.5; A s As the projection area of ​​each segment of the tower, it has been calculated; V is the corresponding wind speed time series, where the coefficient 0.5 is the load at two points on each tower end, only half of it is taken;

[0127] The wind load time history of each node of the transmission line is divided into units with a spacing of 1m. F l Calculate as follows:

[0128] ;

[0129] in, m scis the transmission line shape factor, which is 1.2 for the conductor and 1.1 for the ground wire respectively; d is the diameter of the transmission line, the conductor diameter is the single conductor diameter;

[0130] When wind load is applied to the insulator string, refer to the following formula:

[0131] ;

[0132] Where: A i is the calculated value of the wind pressure area of ​​the insulator string;

[0133] According to the load size finally output by the model, the galloping risk and galloping intensity under the corresponding load are obtained by using the table lookup method;

[0134] The working principle of the above technical solution is that the weight coefficient of each node in the power grid or network reflects its importance or influence. These weight coefficients may be calculated based on multiple factors such as the physical location of the node, connection strength, historical data, etc. Collect environmental parameters of multiple nodes on the same line in the same time period, such as temperature, humidity, wind speed, wind direction, etc. Among them, if the time interval between two data sampling intervals is within a certain time range, it is considered to be the same time period, and the time range can be the average of two adjacent sampling intervals of multiple data; these parameters have a significant impact on the dancing behavior of the line. At the same time, collect the dancing data of each node on the same line, such as dancing duration, dancing amplitude, dancing frequency, etc. These data are an important basis for assessing the risk of line dancing. According to the weight coefficient of each node, the environmental parameters of multiple nodes on the same line in the same time period are weighted averaged. In this way, the weighted average values ​​of the environmental parameters corresponding to the line can be obtained, and these values ​​reflect the overall environmental status of the line. Similarly, the dancing data of each node are weighted averaged according to the weight coefficient to obtain the average dancing data of the line. These data reflect the overall dancing behavior characteristics of the line. In addition to the weighted average of environmental parameters at the current time point, it is also necessary to consider the changing trends of the weighted averages of environmental parameters of the corresponding lines for multiple consecutive times (i.e. historical data). These changing trends can reveal the long-term impact of environmental parameters on the dancing behavior of the line. Similarly, analyze the changing trends of the average dancing data of the corresponding lines for multiple consecutive times. These changing trends reflect the stability and predictability of the dancing behavior of the line. Based on the above weighted averages and their changing trends, combined with statistics, machine learning or other prediction algorithms, a line dynamic prediction model is established. The model can comprehensively consider the changes in environmental parameters and dancing data to predict the future dancing behavior of the line. The environmental parameters collected in real time are input into the model to predict the future dancing behavior of the line. According to the comparison between the prediction results and the actual observation data, the model is fed back and optimized to improve the accuracy and reliability of the prediction.

[0135] The effect of the above technical solution is that the introduction of weight coefficients enables the contribution of different nodes in the model to be reflected differently. According to the node's dance history data (such as dance duration, amplitude, frequency) and the node's importance score ( H i ) can more accurately reflect the impact of the node on the overall dancing behavior of the line; this differentiated processing method helps to improve the accuracy of the prediction model; the calculation formula of the weight coefficient explicitly considers the multi-dimensional characteristics of the node (dancing duration, amplitude, frequency) and the importance score of the node, making the model results easier to understand and explain. Due to differences in the importance, geographical location, environmental conditions, etc. of different lines in the power grid, the weight coefficients of their nodes will also be different. By introducing the node importance score ( H i ) and the comprehensive score of dancing data ( Z ij ), the model can more flexibly adapt to the characteristics of different lines and provide customized prediction services for each line. When performing weighted averaging, the environmental parameters and dancing data of each node are weighted using weight coefficients, which can reduce the interference of irrelevant or low-influence nodes on the model prediction results. This helps to optimize the quality of the model's input data and improve the stability and reliability of the model.

[0136] The calculation formula of the weight coefficient provides a clear direction for improving the model. With the accumulation of data and the advancement of technology, the functional relationship can be continuously optimized. f () and node importance score H i The calculation method of the weight coefficient is to improve the accuracy and rationality of the weight coefficient. This will further promote the continuous improvement and enhancement of the model performance. Since the weight coefficient is calculated based on the real-time galloping data and importance score of the node, the model can reflect the changes in the line galloping risk in real time. This helps operation and maintenance personnel to promptly identify potential risks and take corresponding prevention and control measures to ensure the safe and stable operation of the power grid. By establishing a dynamic prediction model and evaluating the line galloping risk in real time, operation and maintenance personnel can more accurately locate high-risk areas and nodes and reduce unnecessary inspections and maintenance work. This will significantly improve operation and maintenance efficiency and reduce operation and maintenance costs.

[0137] In summary, establishing a line dynamic prediction model by introducing a specific weight coefficient calculation formula can bring many benefits and effects, including improving prediction accuracy, enhancing model interpretability, adapting to different line characteristics, optimizing model input, promoting continuous model improvement, supporting real-time risk assessment, and improving operation and maintenance efficiency.

[0138] In some embodiments, the method for setting the preset time is as follows:

[0139] The average dance duration of the line is obtained through historical dance data;

[0140] Determine the preset time based on the current environmental parameters, the importance score of the line and the average dancing time of the line;

[0141] The preset time is:

[0142]

[0143]

[0144] in, T y The current preset time length for the line, T a is the average dancing time of the line, E c is the rate of change of the current environmental parameters of the line; E a is the historical average environmental parameter change rate of the line; C h The current prediction model h The maximum rate of change of an input environmental parameter; X h For the prediction model h The weight of the input environmental parameters; v The number of environmental parameters input to the prediction model; C ha For the prediction model h The average value of the maximum change rate of the input environmental parameters within the statistical time period; k Score the importance of the line, 0< k <1; β is the adjustment coefficient, 0< β <1.

[0145] The working principle of the above technical solution is: first, the average dancing time of the overhead line is obtained through historical records ( T a ) This is an important reference for evaluating the possible dancing time in the future; real-time monitoring and recording of changes in current environmental parameters (such as wind speed, temperature, humidity, etc.). For each input environmental parameter in the prediction model, calculate its current maximum rate of change ( C h ), and multiplied by the weight of that parameter ( X h ) to perform weighted summation to obtain the current environmental parameter change rate ( E c ). This step takes into account the different degrees of impact of different environmental parameters on dancing risk.

[0146] Count the maximum change rate of each environmental parameter within the statistical time period and calculate its mean ( C ha ). Similarly, for each environmental parameter C ha The weighted sum of the values ​​is used to obtain the historical average environmental parameter change rate ( E a This step provides a basis for assessing the degree of deviation of current environmental parameters from historical average levels. Route importance score ( k ) reflects the importance of the line in the power grid. Important lines require a shorter preset time to ensure timely response. k The larger it is, the more important the line is.

[0147] Adjustment coefficient ( β ) is used to adjust the preset time change caused by changes in environmental parameters, making the preset time setting more flexible and reasonable. E c and E a The comparison result sets the preset time ( T y ):when E c Less than E a , it indicates that the change rate of the current environmental parameters is lower than the historical average, so the preset time can be appropriately extended. Specifically, through the formula T y = (1 + β * k * (E a - E c ) / E a ) * T a Calculated.

[0148] when E c Greater than or equal to E a , it indicates that the change rate of the current environmental parameters is higher than or close to the historical average level, and the preset time needs to be shortened to deal with possible dancing risks. At this time, through the formula T y = (1 + (1 - k) * ((E a -E c ) / E a + 0.1)) * T a Calculated.

[0149] The preset time setting is not fixed, but is dynamically adjusted according to the changes in current environmental parameters and the importance of the line. This helps ensure that the prediction model can timely and accurately assess the galloping risk of the line based on actual conditions.

[0150] The effect of the above technical solution is that the setting of the preset time takes into account the change rate of the current environmental parameters ( E c ) and the historical average rate of change ( E a ). E c Greater than E a , it means that the current environmental conditions are more likely to cause a disturbance than the historical average, so the preset time will be shortened, ensuring that the warning can be issued more promptly, giving the operation and maintenance personnel more response time. k ) and the adjustment coefficient ( β ), the preset time setting can be more flexible to meet the needs of different lines and different environmental conditions. Lines with high importance scores are more likely to receive attention when environmental parameters change, and the introduction of adjustment coefficients allows fine-tuning of the preset time to further improve the accuracy of early warning.

[0151] Reasonable setting of preset time helps optimize the configuration of operation and maintenance resources. For lines with high importance and large changes in current environmental parameters, inspections and prevention and control measures can be arranged in advance to avoid waste of resources and delays. At the same time, for lines with low importance or relatively stable environmental parameters, the preset time can be appropriately extended to reduce unnecessary inspection times.

[0152] By comprehensively considering multiple factors to determine the preset time, the system can maintain a high degree of robustness when facing complex and changing environmental conditions. Even in some extreme cases, when the changes in individual environmental parameters exceed the expected range, the system can maintain the effectiveness of the warning by adjusting the preset time. The setting of the preset time not only provides a time basis for the warning, but also provides an important reference for the decision-making of operation and maintenance personnel. By understanding the calculation method of the preset time and the logic behind it, operation and maintenance personnel can more scientifically evaluate the risk of line dancing and formulate corresponding prevention and control strategies.

[0153] In summary, the method of determining the preset time by comprehensively considering multiple factors such as historical dancing data, current environmental parameter change rate, line importance score, and adjustment coefficient has significant effects in improving the timeliness and accuracy of warning, optimizing resource allocation, and enhancing system robustness.

[0154] In some embodiments, issuing a warning of galloping and providing decision support according to the prediction results include:

[0155] If the first risk probability of any node is greater than the corresponding probability threshold, or the line galloping risk probability is greater than the corresponding probability threshold, a first warning is issued, and the first warning is first graded according to the predicted galloping intensity;

[0156] If the first risk probabilities of multiple nodes on a line are all greater than the corresponding probability threshold, a second warning is issued; and the second warning is second graded according to the predicted pulsation intensity;

[0157] Establish an early warning feedback mechanism to collect and analyze the actual effects of early warning and feedback from operation and maintenance personnel. Continuously optimize and improve the judgment logic and / or early warning system based on the feedback results to improve its adaptability and accuracy.

[0158] The preset threshold is:

[0159]

[0160] in, POS yi1 is the probability threshold corresponding to the node, POS a1 is the probability reference value corresponding to the node; POS y2 is the probability threshold corresponding to the line; POS a2 is the probability reference value corresponding to the line; H i Score the node importance, k Score the importance of the line. b is a constant, 3≤ b ≤5.

[0161] The working principle of the above technical solution is: using the established mathematical model or machine learning algorithm, based on the collected data, calculate the risk probability of dancing for each node and the entire line (i.e. the first risk probability and the line dancing risk probability).

[0162] According to the importance score of the node ( H i ) and the importance score of the line ( k ), and the corresponding probability reference value ( POS a1 、POS a2 ), calculate the probability threshold of each node and line ( POS yi1 、POS y2 ). These thresholds are used to determine whether an early warning is needed.

[0163] If the first risk probability of any node is greater than its preset threshold, or the galloping risk probability of the entire line is greater than its preset threshold, the system will issue a first warning. At the same time, the first warning is graded according to the predicted galloping intensity so that the operation and maintenance personnel can take corresponding emergency measures according to the warning level.

[0164] If the first risk probabilities of multiple nodes on a line are all greater than their preset thresholds, indicating that the line faces a high risk of swaying, the system will issue a second warning. Similarly, the second warning is graded according to the predicted swaying intensity.

[0165] In the present invention, the first warning and the second warning are two different signals; the level of the warning signal is based on the corresponding dancing intensity. The dancing intensity and the warning level are mapping relationships pre-established by the staff in this field according to the actual local conditions, and can be quickly graded by table lookup.

[0166] Establish an early warning feedback mechanism to collect and analyze the actual effects of early warnings and feedback from operation and maintenance personnel. These feedbacks include the timeliness, accuracy, and rationality of early warning levels.

[0167] Based on the collected feedback, the judgment logic and / or early warning system will be continuously optimized and improved. This may include adjusting the calculation method of probability thresholds, optimizing prediction models, improving early warning classification standards, etc., to improve the adaptability and accuracy of the system.

[0168] Through four main links, namely data collection and analysis, model prediction, early warning mechanism, and early warning feedback and optimization, the prediction, early warning and continuous optimization of power line galloping risks are achieved. This mechanism helps to detect potential galloping risks in advance and provide timely decision-making support for operation and maintenance personnel, thereby reducing the impact of galloping disasters on power lines.

[0169] The effect of the above technical solution is that by real-time monitoring and predicting the risk of galloping of power lines and their nodes, the system can issue an early warning before a galloping event occurs. This early warning mechanism provides operation and maintenance personnel with valuable time to prepare countermeasures, reducing the potential impact of galloping events on power lines and power grid operations.

[0170] The early warning system not only issues warnings, but also grades them according to the predicted intensity of the galloping. This graded warning mechanism helps operation and maintenance personnel take corresponding emergency measures according to the warning level and achieve accurate response. High-level warnings can trigger more urgent response processes to ensure that quick action can be taken in the event of serious galloping risks.

[0171] The preset thresholds are set taking into account the importance scores of nodes and lines, so that each node and line has a personalized threshold. bIf the value of is too small, the importance score will have a great impact on the preset threshold, causing the preset threshold to change significantly even if the importance score of a node or line is slightly different. This will make the system too sensitive to small changes, generate too many warnings, and increase the false alarm rate. On the contrary, when b When the value of is too large, the impact of the importance score on the probability threshold will be reduced, which may cause the system to be insensitive to changes in importance and thus miss some important warnings. Choose a moderate b Value range (such as 3≤ b ≤5), a balance can be found between sensitivity and accuracy; values ​​within this range will make the system neither too sensitive nor too slow, thus being able to more accurately reflect the actual risk situation of nodes or lines.

[0172] This personalized setting improves the adaptability of the early warning system and can more accurately reflect the actual risk situation of different nodes and lines. The established early warning feedback mechanism allows the system to collect and analyze the actual effect after the early warning and the feedback from the operation and maintenance personnel. These feedbacks provide an important basis for the continuous optimization and improvement of the system, which helps to improve the accuracy and reliability of the early warning system. By continuously optimizing the judgment logic and early warning system, it can be more adaptable to the complex and changeable actual operating environment.

[0173] The embodiment of the present application provides a system for an overhead line galloping early warning method based on a machine learning model as described in the embodiment of the present application, the system comprising:

[0174] A node setting module, used to set multiple monitoring nodes on the overhead line;

[0175] The prediction module is used to obtain the weight coefficient of each node according to the dancing data of each node; obtain the first risk of each node within a preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within a preset time according to the line dynamic prediction model; wherein, the average dancing time of the line is obtained through the historical dancing data; and dynamically adjust and determine the preset time according to the current environmental parameters, the line importance score and the average dancing time of the line;

[0176] The early warning module is used to issue a galloping warning and provide decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk.

[0177] The working principle and effect of the above technical solution are the same as those in the embodiment of the method of the present application and will not be described in detail here.

[0178] This application is explained from the perspectives of purpose of use, effectiveness, progress and novelty, and has met the functional enhancement and usage requirements emphasized by the Patent Law. The above description and drawings of this application are only the preferred embodiments of this application, and are not intended to limit this application. Therefore, all structures, devices, features, etc. that are similar or identical to this application, that is, all equivalent replacements or modifications made in accordance with the scope of the patent application of this application, should fall within the scope of protection of the patent application of this application.

Claims

1. An overhead line galloping warning method based on a machine learning model, characterized in that: The method comprises: S1. Set up multiple monitoring nodes on the overhead line; S2. Obtain a weight coefficient of each node according to the dancing data of each node; obtain a first risk of each node within a preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within a preset time according to the line dynamic prediction model; Among them, the average dancing time of the line is obtained through historical dancing data; according to the current environmental parameters, the importance score of the line and the average dancing time of the line, the preset time is dynamically adjusted and determined; S3. issuing a galloping warning and providing decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk; The line dynamic prediction model is established by using the weight coefficient of each node and the environmental parameters of each node, including: According to the weight coefficient of each node, the environmental parameters of multiple nodes in the same line in the same time period and the dancing data of each node are weighted averaged to obtain the weighted average value of each environmental parameter corresponding to the line and the average dancing data of the line; A line dynamic prediction model is established by taking the weighted average values ​​of the environmental parameters of the line, the average dancing data of the corresponding line, and the changes of the weighted average values ​​of the environmental parameters of the corresponding line for multiple consecutive times and the changes of the average dancing data; The preset time is obtained by the following formula: ,in, T y The current preset time length for the line, T a is the average dancing time of the line, E c is the rate of change of the current environmental parameters of the line; E a is the historical average environmental parameter change rate of the line; C h The current prediction model h The maximum rate of change of an input environmental parameter; X h For the prediction model h The weight of the input environmental parameters; v The number of environmental parameters input to the prediction model; C ha For the prediction model h The average value of the maximum change rate of the input environmental parameters within the statistical time period; k Score the importance of the line, 0< k <1; β is the adjustment coefficient, 0< β <1.

2. The method according to claim 1, characterized in that The S1 includes: Based on environmental parameters, topography and vegetation coverage, cluster analysis is performed on the areas where overhead lines pass through, and multiple monitoring sub-areas are divided according to the cluster analysis results; A monitoring node is set up in each monitoring sub-area.

3. The method according to claim 1, characterized in that The S2 includes: Acquire first dance data of each node in real time through a dance sensor, wherein the first dance data includes a first dance time, a first dance duration, a first dance amplitude, and a first dance frequency; Obtaining a weight coefficient of each node according to the first dance data of each node; According to the first dancing data of each node and the environmental parameters of each node, the first risk of each node within the preset time is obtained; through the weight coefficient of each node and the environmental parameters of each node, a line dynamic prediction model is established; Collect multiple sets of continuous real-time data and input them into the line dynamic prediction model to predict the risk of line dancing within a preset time.

4. The method according to claim 3, characterized in that The weight coefficient of each node is obtained by the following formula: ,in, w i For the i The weight coefficient of each node; S ij For the i Node j Duration of the first dance of the dance; v ij No. i Node j The amplitude of the first dance of the second dance; p ij For the i Node j The first dance frequency of the second dance; f () is the functional relationship; Z ij For the i Node j The overall score of the dance; n is the number of nodes, m The number of dances in the statistical time period; H i For Node i Importance rating; G ij For the i The node in j The median value in the dance.

5. The method according to claim 3, characterized in that: The step of obtaining the first risk of each node within a preset time according to the first dancing data of each node and the environmental parameters of each node comprises: Establishing a first risk model through the first dancing data of each node and the environmental parameters of each node; The first dance data includes a plurality of first dance durations, a plurality of first dance amplitudes, a plurality of first dance frequencies, a change in dance duration, a change in dance amplitude, and a change in dance frequency of nodes within a first statistical time period; The environmental parameters include the maximum value, minimum value and average value of the environmental parameter change rate of each node dance; the environmental parameter change rate includes the temperature change rate, humidity change rate, wind direction change rate, and the average angle change rate between the wind direction and the node line; A first risk of each node is obtained through a first risk model, where the first risk includes a first risk probability and a first dancing intensity.

6. The method according to claim 5, characterized in that The first risk model is established by using the first dancing data of each node and the environmental parameters of each node; including: Perform correlation analysis on each environmental parameter and the first dancing data to obtain each first correlation coefficient; If the absolute value of the first correlation coefficient between the environmental parameter and any dancing data is greater than a first preset threshold, the environmental parameter is used as an input of the first risk model; Among them, the first preset threshold is: ,in, D y is the first preset threshold, D a is the mean of the absolute values ​​of the first correlation coefficients; σ is the standard deviation of the means of the first correlation coefficients; a is the adjustment coefficient, 1≤ a ≤5; k Score the importance of the line, 0< k <1.

7. The method according to claim 1, characterized in that The method of issuing a warning of galloping and providing decision support based on the prediction results includes: If the first risk probability of any node is greater than its preset threshold, or the line dancing risk probability is greater than its preset threshold, a first warning will be issued, and the level of the warning will be further classified according to the predicted dancing intensity; if the first risk probability of multiple nodes on a line is greater than its preset threshold, a second warning will be issued; the level of the warning will be further classified according to the predicted dancing intensity.

8. A system for implementing the method for early warning of overhead line galloping based on a machine learning model as claimed in claim 1, characterized in that: The system comprises: A node setting module, used to set multiple monitoring nodes on the overhead line; The prediction module is used to obtain the weight coefficient of each node according to the dancing data of each node; obtain the first risk of each node within a preset time according to the dancing data of each node and the corresponding environmental parameters; establish a line dynamic prediction model through the weight coefficient of each node and the environmental parameters of each node; predict the line dancing risk within a preset time according to the line dynamic prediction model; wherein, the average dancing time of the line is obtained through the historical dancing data; and dynamically adjust and determine the preset time according to the current environmental parameters, the line importance score and the average dancing time of the line; An early warning module, used to issue a galloping warning and provide decision support according to the prediction results, wherein the prediction results include the first risk and the line galloping risk; The line dynamic prediction model is established by using the weight coefficient of each node and the environmental parameters of each node, including: According to the weight coefficient of each node, the environmental parameters of multiple nodes in the same line in the same time period and the dancing data of each node are weighted averaged to obtain the weighted average value of each environmental parameter corresponding to the line and the average dancing data of the line; A line dynamic prediction model is established by taking the weighted average values ​​of the environmental parameters of the line, the average dancing data of the corresponding line, and the changes of the weighted average values ​​of the environmental parameters of the corresponding line for multiple consecutive times and the changes of the average dancing data; The preset time is obtained by the following formula: ,in, T y The current preset time length for the line, T a is the average dancing time of the line, E c is the rate of change of the current environmental parameters of the line; E a is the historical average environmental parameter change rate of the line; C h The current prediction model h The maximum rate of change of an input environmental parameter; X h For the prediction model h The weight of the input environmental parameters; v The number of environmental parameters input to the prediction model; C ha For the prediction model h The average value of the maximum change rate of the input environmental parameters within the statistical time period; k Score the importance of the line, 0< k <1; β is the adjustment coefficient, 0< β <1.

Images