Deicing early warning method and device for power transmission tower

Abstract

The invention relates to a de-icing early warning method and device for a power transmission tower. The method comprises the following steps: constructing a first training data set, wherein the first training data set comprises meteorological data, radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data; a weather prediction model is constructed, the weather prediction model comprises a long-short-term memory network and a Transform model, weather data prediction values are output through the trained weather prediction model, and the weather data prediction values after the current moment are judged to determine a current deicing mode; finite element models of different power transmission towers are constructed to construct a second training data set; training a structure response prediction model by using the second training data set to obtain a trained structure response prediction model; inputting the trained structure response prediction model to obtain a response parameter prediction value; and substituting the response parameter prediction value into various failure functions to calculate a failure value, and comparing the failure value with a corresponding early warning threshold to output an early warning result. By adopting the method, the early warning accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of energy storage system prediction, and in particular to a method, device, computer equipment, computer-readable storage medium, and computer program product for warning deicing of transmission towers. Background Art

[0002] With the development of the power industry, the transmission grid system continues to improve, and the coverage of the transmission grid system in harsh environments such as high altitude areas is also increasing. However, the transmission towers of the transmission grid system in harsh environments are prone to icing. When the ice is too thick, it seriously affects the safe operation of the power grid.

[0003] In order to ensure the safe and stable operation of the power grid, the following methods are currently available for rapid assessment and early warning of transmission tower deicing: (1) Establish a refined finite element model of a static transmission tower line, then consider various deicing conditions, conduct dynamic time history analysis, extract stresses and node displacements at key locations, and verify the safety of the transmission tower structure. (2) Through experimental methods, establish a scaled or full-scale model of the transmission tower, use artificial icing, falling weights, etc. to trigger or simulate the deicing process of the transmission tower, and monitor the dynamic response of the actual tower line. Based on the results, analyze the impact of deicing on the structure. (3) Combine historical meteorological data and icing observations to predict the probability of deicing risk. Combined with the deicing load estimation formulas provided by relevant standards such as IEC and ASCE, empirical and semi-empirical methods are used to evaluate deicing conditions.

[0004] However, current rapid assessment and early warning methods for transmission tower deicing still have the following problems: (1) Related research methods are often based on a static model and cannot be used to quickly assess deicing for transmission towers of different types or with different design parameters. The experimental method is too expensive. (2) Deicing assessments are always performed using finite element analysis based on certain specific wind speed or temperature conditions (ice thickness), and the impact of complex meteorological conditions on deicing is rarely considered, resulting in errors in deicing assessments. (3) Empirical and semi-empirical methods have large errors in deicing assessments and are generally not practical.

[0005] In summary, there is an urgent need for a more efficient and accurate transmission tower de-icing warning method. Summary of the Invention

[0006] Based on this, it is necessary to provide a de-icing warning method, device, computer equipment, computer-readable storage medium and computer program product for transmission towers that can improve the accuracy of warning in response to the above technical problems.

[0007] In a first aspect, the present application provides a method for early warning of ice shedding of a transmission tower, comprising:

[0008] Constructing a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0009] Constructing a weather forecast model, the weather forecast model including a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively; the input of the long short-term memory network is weather data for a set time period; the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period; and the output of the cross-attention mechanism module is a forecast value of the weather data;

[0010] The meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted are obtained, a predicted value of meteorological data after the current moment is output using the trained meteorological prediction model, and the predicted value of meteorological data after the current moment is discriminated to determine a current deicing mode;

[0011] Constructing finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; inputting different deicing conditions into the finite element model to obtain corresponding response parameters to construct a second training data set;

[0012] Constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition and the output is a response parameter, and training the structural response prediction model using a second training data set to obtain a trained structural response prediction model;

[0013] Obtaining the current ice thickness of the transmission tower to be predicted, and inputting the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values;

[0014] Substituting the predicted value of the response parameter into a plurality of failure functions to calculate the failure value, comparing the failure value with the corresponding warning threshold value to output the warning result.

[0015] In one embodiment, the meteorological data include temperature, humidity, wind speed, wind direction, pH value, the angle between the main wind direction and the line direction, and solar radiation flux; the radar remote sensing data include reflectivity factor and liquid water content; the satellite remote sensing data include cloud top temperature and cloud phase; the analytical data include boundary layer height and vertical temperature gradient; the meteorological derived data include atmospheric heating rate, inversion layer strength, ice-water conversion index, supercooled water content, wind shear and turbulent kinetic energy.

[0016] In one embodiment, the construction of the first training data set includes: gridding the area where the transmission tower is located, and uniformly interpolating the meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data to corresponding grid points in the grid; using the meteorological data as the reference data, setting the weight of the meteorological data to a first weight; setting the horizontal weight of the radar remote sensing data to a second weight; setting the cloud-free area weight of the satellite remote sensing data to a second weight, and setting the weights of the analysis data and the meteorological derivative data to a first weight, wherein the first weight is greater than the second weight and greater than the third weight; and the meteorological data, the radar remote sensing data, the satellite remote sensing data, the analysis data, and the meteorological derivative data after the weight setting are completed constitute the first training data set.

[0017] In one embodiment, the determination of the current de-icing mode by judging the predicted value of meteorological data after the current moment includes: obtaining the temperature, wind speed and solar radiation flux from the predicted value of meteorological data after the current moment, determining the de-icing probability based on the wind speed, and calculating the inversion layer thickness and ice-water mixing ratio based on the predicted value of meteorological data; if the temperature, wind speed, solar radiation flux, de-icing probability, inversion layer thickness and ice-water mixing ratio in the predicted value of meteorological data meet the corresponding judgment conditions, then outputting the de-icing mode corresponding to the condition and using it as the current de-icing mode.

[0018] In one embodiment, the predicted value of the response parameter is substituted into a plurality of failure functions to calculate the failure value, including: obtaining a strength failure function, a stability failure function and a fatigue failure function; and substituting the predicted value of the response parameter into the strength failure function, the stability failure function and the fatigue failure function respectively to calculate the corresponding failure value.

[0019] In one embodiment, before inputting the current ice thickness, wind speed, wind direction angle, temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode into the trained structural response prediction model, the method further includes: establishing a joint probability distribution according to the current ice thickness, wind speed, wind direction angle, temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode to obtain a joint distribution function, and extracting the ice thickness, wind speed, wind direction angle, temperature change rate, and de-icing mode from the joint distribution function as input to the trained structural response prediction model.

[0020] In a second aspect, the present application further provides an ice shedding warning device for a transmission tower, comprising:

[0021] A first training data set construction module is used to construct a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0022] A first modeling module is used to construct a weather forecast model, wherein the weather forecast model includes a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively. The input of the long short-term memory network is weather data for a set time period, the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period, and the output of the cross-attention mechanism module is a forecast value of the weather data;

[0023] a training module configured to train the meteorological prediction model using the first training data set to obtain a trained meteorological prediction model; obtain meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted, output a predicted value of meteorological data after the current moment using the trained meteorological prediction model, and determine a current deicing mode based on the predicted value of meteorological data after the current moment;

[0024] a second training data set construction module, configured to construct finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct the second training data set;

[0025] a second modeling module for constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition, the output is a response parameter, and the structural response prediction model is trained using a second training data set to obtain a trained structural response prediction model;

[0026] A real-time data acquisition module is used to obtain the current ice thickness of the transmission tower to be predicted, and input the current ice thickness, wind speed, wind direction angle, temperature change rate and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values;

[0027] The early warning module is used to substitute the predicted value of the response parameter into multiple failure functions to calculate the failure value, compare the failure value with the corresponding early warning threshold, and output the early warning result.

[0028] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0029] Constructing a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0030] Constructing a weather forecast model, the weather forecast model including a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively; the input of the long short-term memory network is weather data for a set time period; the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period; and the output of the cross-attention mechanism module is a forecast value of the weather data;

[0031] The meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted are obtained, a predicted value of meteorological data after the current moment is output using the trained meteorological prediction model, and the predicted value of meteorological data after the current moment is discriminated to determine a current deicing mode;

[0032] Constructing finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; inputting different deicing conditions into the finite element model to obtain corresponding response parameters to construct a second training data set;

[0033] Constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition and the output is a response parameter, and training the structural response prediction model using a second training data set to obtain a trained structural response prediction model;

[0034] Obtaining the current ice thickness of the transmission tower to be predicted, and inputting the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values;

[0035] Substituting the predicted value of the response parameter into a plurality of failure functions to calculate the failure value, comparing the failure value with the corresponding warning threshold value to output the warning result.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0037] Constructing a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0038] Constructing a weather forecast model, the weather forecast model including a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively; the input of the long short-term memory network is weather data for a set time period; the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period; and the output of the cross-attention mechanism module is a forecast value of the weather data;

[0039] The meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted are obtained, a predicted value of meteorological data after the current moment is output using the trained meteorological prediction model, and the predicted value of meteorological data after the current moment is discriminated to determine a current deicing mode;

[0040] Constructing finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; inputting different deicing conditions into the finite element model to obtain corresponding response parameters to construct a second training data set;

[0041] Constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition and the output is a response parameter, and training the structural response prediction model using a second training data set to obtain a trained structural response prediction model;

[0042] Obtaining the current ice thickness of the transmission tower to be predicted, and inputting the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values;

[0043] Substituting the predicted value of the response parameter into a plurality of failure functions to calculate the failure value, comparing the failure value with the corresponding warning threshold value to output the warning result.

[0044] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0045] Constructing a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0046] Constructing a weather forecast model, the weather forecast model including a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively; the input of the long short-term memory network is weather data for a set time period; the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period; and the output of the cross-attention mechanism module is a forecast value of the weather data;

[0047] The meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted are obtained, a predicted value of meteorological data after the current moment is output using the trained meteorological prediction model, and the predicted value of meteorological data after the current moment is discriminated to determine a current deicing mode;

[0048] Constructing finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; inputting different deicing conditions into the finite element model to obtain corresponding response parameters to construct a second training data set;

[0049] Constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition and the output is a response parameter, and training the structural response prediction model using a second training data set to obtain a trained structural response prediction model;

[0050] Obtaining the current ice thickness of the transmission tower to be predicted, and inputting the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values;

[0051] Substituting the predicted value of the response parameter into a plurality of failure functions to calculate the failure value, comparing the failure value with the corresponding warning threshold value to output the warning result.

[0052] The above-mentioned deicing warning method, device, computer equipment, computer-readable storage medium and computer program product for transmission towers construct a first training data set, which includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data; construct a meteorological prediction model, which includes a long short-term memory network, a Transformer model and a cross-attention mechanism module, and the input end of the cross-attention mechanism module is respectively connected to the output end of the long short-term memory network and the output end of the Transformer model, the input of the long short-term memory network is the meteorological data of a set time period, the input of the Transformer model is the radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data of a set time period, and the output of the cross-attention mechanism module is the predicted value of the meteorological data; thereby, the coupling of meteorological data with other types of data can be achieved, the accuracy of meteorological prediction can be improved, and thus the accuracy of subsequent deicing mode determination can be improved. Finite element models of different transmission towers are constructed, where the input of the finite element model is the deicing condition, and the output is the response parameter. The deicing condition includes ice thickness, wind speed, wind direction angle, temperature change rate and deicing mode. Different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct a second training data set. A structural response prediction model is constructed, where the input of the structural response prediction model is the deicing condition, and the output is the response parameter. The structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model. In this way, a large number of data sets consisting of deicing conditions and response parameters of each transmission tower are obtained using the finite element model, which is conducive to improving the prediction accuracy of the structural response prediction model, and thus conducive to improving the early warning accuracy. A meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model. Meteorological data, radar remote sensing data, satellite remote sensing data, analytical data, and meteorological derivative data for a set period of time before the current moment of the transmission tower to be predicted are obtained. The trained meteorological prediction model outputs a predicted value of the meteorological data after the current moment, and the predicted value of the meteorological data after the current moment is discriminated to determine the current deicing mode. The current ice thickness of the transmission tower to be predicted is obtained, and the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the predicted value of the meteorological data after the current moment are input into the trained structural response prediction model to obtain a predicted value of the response parameter. The predicted value of the response parameter is substituted into multiple failure functions to calculate the failure value, and the failure value is compared with the corresponding warning threshold to output a warning result. Integrating multiple failure values ​​for warning is beneficial to improving the accuracy of warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 1 is a flow chart of a method for early warning of ice shedding of a transmission tower according to an embodiment;

[0055] Figure 2 1 is a flow chart of a method for early warning of ice shedding of a transmission tower according to another embodiment;

[0056] Figure 3 1 is a structural block diagram of an ice shedding warning device for a transmission tower according to one embodiment;

[0057] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0059] In an exemplary embodiment, Figure 1 As shown, a method for early warning of ice loss on a transmission tower is provided. This embodiment uses the method applied to a terminal as an example. It is understood that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0060] Step 102: construct a first training data set, which includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data.

[0061] Meteorological data refers to atmospheric parameters collected by various observation devices. Radar remote sensing data is generated by actively transmitting microwaves and receiving echoes. Satellite remote sensing data is acquired through non-contact Earth detection using optical / microwave sensors onboard satellites. Analytical data refers to secondary analysis data derived from meteorological, radar, and satellite remote sensing data using existing analytical methods. Meteorological derivative data refers to meteorological-related parameters derived from the analysis of meteorological, radar, and satellite remote sensing data.

[0062] Exemplarily, meteorological data, radar remote sensing data, and satellite remote sensing data are collected by equipment, and the collected meteorological data, radar remote sensing data, and satellite remote sensing data are analyzed to obtain analytical data and meteorological derivative data.

[0063] Step 104, construct a meteorological forecast model, which includes a long short-term memory network, a Transformer model and a cross-attention mechanism module. The input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the Transformer model respectively. The input of the long short-term memory network is the meteorological data of the set time period, and the input of the Transformer model is the radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data of the set time period. The output of the cross-attention mechanism module is the predicted value of the meteorological data.

[0064] Among them, the long short-term memory network is a recurrent neural network. The Transformer model is a deep learning architecture based entirely on the self-attention mechanism. The cross-attention mechanism module is a special attention mechanism used to process the correlation between two different input sequences, achieving information fusion by calculating the attention of one sequence element to another sequence element.

[0065] Exemplarily, a long short-term memory network, a transformer model, and a cross-attention mechanism module are obtained, and the input of the cross-attention mechanism module is connected to the output of the long short-term memory network and the output of the transformer model, respectively, to construct a weather forecast model. The input of the long short-term memory network in the weather forecast model is weather data for a set time period, the input of the transformer model in the weather forecast model is radar remote sensing data, satellite remote sensing data, analytical data, and weather-derived data for the set time period, and the output of the cross-attention mechanism module (i.e., the output of the weather forecast model) is the predicted value of the weather data.

[0066] Step 106: Train a meteorological prediction model using the first training data set to obtain a trained meteorological prediction model; obtain meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before the current moment of the transmission tower to be predicted, use the trained meteorological prediction model to output a predicted value of the meteorological data after the current moment, and determine the current de-icing mode based on the predicted value of the meteorological data after the current moment.

[0067] Exemplarily, a meteorological prediction model is trained using a first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data of a set period before the current moment of the transmission tower to be predicted are obtained, the obtained meteorological data, radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data are input into the trained meteorological prediction model to output a predicted value of the meteorological data after the current moment, and the predicted value of the meteorological data after the current moment is judged to determine the current de-icing mode.

[0068] Step 108: Construct finite element models of different transmission towers. The input of the finite element model is the deicing condition, and the output is the response parameter. The deicing condition includes ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode. Different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct a second training data set.

[0069] The finite element model is a model established using the finite element analysis method, and the response parameter is data related to changes in the transmission tower structure.

[0070] Exemplarily, finite element models of different transmission towers are constructed, and different de-icing conditions are input into the finite element models to obtain corresponding response parameters to construct a second training data set.

[0071] Step 110 , constructing a structural response prediction model, wherein the input of the structural response prediction model is the de-icing condition, and the output is the response parameter. The structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model.

[0072] The de-icing operating condition is operating condition data related to de-icing.

[0073] Exemplarily, a structural response prediction model is constructed, and the structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model.

[0074] Step 112: Obtain the current ice thickness of the transmission tower to be predicted, and input the current ice thickness, wind speed, wind direction angle, temperature change rate, and current de-icing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values.

[0075] The current ice thickness refers to the thickness of ice currently covering the object, and the current de-icing mode refers to the current de-icing status.

[0076] Exemplarily, the current ice thickness of the transmission tower to be predicted is obtained, and the current ice thickness, wind speed, wind direction angle, temperature change rate and current de-icing mode in the meteorological data prediction value after the current moment are used as inputs of the trained structural response prediction model, and the response parameter prediction value is obtained using the trained structural response prediction model.

[0077] Step 114 , substitute the predicted values ​​of the response parameters into various failure functions to calculate failure values, compare the failure values ​​with corresponding warning thresholds, and output warning results.

[0078] The failure function is a function used to describe the failure probability of an object.

[0079] Exemplarily, the predicted value of the response parameter is substituted into a variety of failure functions to calculate the corresponding failure value, and each failure value is compared with the corresponding warning threshold. Based on the interval in which the failure value is located, the corresponding warning result is obtained.

[0080] In the above-mentioned de-icing warning method for transmission towers, a first training data set is constructed, and the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data; a meteorological prediction model is constructed, and the meteorological prediction model includes a long short-term memory network, a Transformer model, and a cross-attention mechanism module. The input end of the cross-attention mechanism module is respectively connected to the output end of the long short-term memory network and the output end of the Transformer model. The input of the long short-term memory network is the meteorological data of a set time period, and the input of the Transformer model is the radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data of a set time period. The output of the cross-attention mechanism module is the predicted value of the meteorological data; thereby, the coupling of meteorological data with other types of data can be achieved, the accuracy of meteorological prediction can be improved, and thus it is beneficial to improve the accuracy of subsequent de-icing mode determination. Finite element models of different transmission towers are constructed, where the input of the finite element model is the deicing condition, and the output is the response parameter. The deicing condition includes ice thickness, wind speed, wind direction angle, temperature change rate and deicing mode. Different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct a second training data set. A structural response prediction model is constructed, where the input of the structural response prediction model is the deicing condition, and the output is the response parameter. The structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model. In this way, a large number of data sets consisting of deicing conditions and response parameters of each transmission tower are obtained using the finite element model, which is conducive to improving the prediction accuracy of the structural response prediction model, and thus conducive to improving the early warning accuracy. A meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model. Meteorological data, radar remote sensing data, satellite remote sensing data, analytical data, and meteorological derivative data for a set period of time before the current moment of the transmission tower to be predicted are obtained. The trained meteorological prediction model outputs a predicted value of the meteorological data after the current moment, and the predicted value of the meteorological data after the current moment is discriminated to determine the current deicing mode. The current ice thickness of the transmission tower to be predicted is obtained, and the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the predicted value of the meteorological data after the current moment are input into the trained structural response prediction model to obtain a predicted value of the response parameter. The predicted value of the response parameter is substituted into multiple failure functions to calculate the failure value, and the failure value is compared with the corresponding warning threshold to output a warning result. Integrating multiple failure values ​​for warning is beneficial to improving the accuracy of warning.

[0081] In an exemplary embodiment, meteorological data include temperature, humidity, wind speed, wind direction, pH value, the angle between the main wind direction and the line direction, and solar radiation flux; radar remote sensing data include reflectivity factor and liquid water content; satellite remote sensing data include cloud top temperature and cloud phase; analytical data include boundary layer height and vertical temperature gradient; meteorological derived data include atmospheric heating rate, inversion layer strength, ice-water conversion index, supercooled water content, wind shear and turbulent kinetic energy.

[0082] Where temperature is the ambient temperature, humidity is the proportion of water vapor in the air, wind speed is the rate of horizontal movement of air relative to the ground, wind direction is the direction of atmospheric movement, pH is the negative logarithm of the hydrogen ion concentration of a solution, the angle between the prevailing wind direction and the power line orientation is the minimum angle between the prevailing wind direction and the axis of the power line, and solar radiation flux is the amount of solar radiation that passes vertically through a unit area per unit time.

[0083] In this embodiment, by comprehensively considering multiple data, the deep coupling of meteorological data and other data can be better achieved, which is beneficial to the accuracy of subsequent predictions.

[0084] In an exemplary embodiment, constructing a first training data set includes: gridding the area where the transmission tower is located, and uniformly interpolating meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data to corresponding grid points in the grid; using meteorological data as baseline data, setting the weight of the meteorological data to a first weight; setting the horizontal weight of the radar remote sensing data to a second weight; setting the cloud-free area weight of the satellite remote sensing data to a second weight, and setting the weights of the analysis data and the meteorological derivative data to a first weight, wherein the first weight is greater than the second weight and greater than the third weight; and the meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data after the weight setting are completed constitute the first training data set.

[0085] In an embodiment, by gridding and assigning weights to the data at each grid point, the difference in data sources can be better eliminated, which is beneficial to the accuracy of subsequent model predictions.

[0086] In an exemplary embodiment, the meteorological data prediction value after the current moment is judged to determine the current de-icing mode, including: obtaining the temperature, wind speed and solar radiation flux from the meteorological data prediction value after the current moment, determining the de-icing probability based on the wind speed, and calculating the inversion layer thickness and ice-water mixing ratio based on the meteorological data prediction value; if the temperature, wind speed, solar radiation flux, de-icing probability, inversion layer thickness and ice-water mixing ratio in the meteorological data prediction value meet the corresponding judgment conditions, then outputting the de-icing mode corresponding to the condition and using it as the current de-icing mode.

[0087] The current deicing mode refers to different deicing scenarios, including unilateral deicing, bilateral deicing, random deicing, and segmented deicing. Temperature is the ambient temperature. Wind speed is the rate of horizontal air movement relative to the ground. Solar radiation flux is the amount of solar radiation energy passing vertically through a unit area per unit time. Deicing probability is the likelihood of natural deicing of an ice-covered object under specific meteorological conditions. Inversion layer thickness is the vertical thickness of the atmospheric layer where the ground temperature increases with altitude. Ice-water mixing ratio is the mass concentration ratio of solid ice crystals to liquid water droplets in a cloud.

[0088] In this embodiment, by comprehensively judging whether the temperature, wind speed, solar radiation flux, de-icing probability, inversion layer thickness, and ice-water mixing ratio in the meteorological data forecast values ​​meet the corresponding judgment conditions, it is possible to more accurately determine whether de-icing occurs, and then judge the de-icing mode when de-icing occurs, thereby improving the efficiency of de-icing mode judgment.

[0089] In an exemplary embodiment, the predicted values ​​of the response parameters are substituted into a plurality of failure functions to calculate failure values, including: obtaining a strength failure function, a stability failure function, and a fatigue failure function; and substituting the predicted values ​​of the response parameters into the strength failure function, the stability failure function, and the fatigue failure function respectively to calculate the corresponding failure values.

[0090] The strength failure function is used to determine whether stress or deformation exceeds the material limit. The stability failure function is used to calculate the critical buckling load of a compressive component. The fatigue failure function is used to calculate the life prediction or crack growth rate under cyclic loading.

[0091] In this embodiment, by integrating the strength failure function, stability failure function and fatigue failure function, the corresponding strength failure value, stability failure value and fatigue failure value can be calculated, which can more comprehensively determine the failure value and is conducive to improving the accuracy of subsequent warnings.

[0092] In an exemplary embodiment, before inputting the current ice thickness, the wind speed, wind direction angle, temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode into the trained structural response prediction model, the method further includes: establishing a joint probability distribution according to the current ice thickness, the wind speed, wind direction angle, temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode to obtain a joint distribution function, and extracting the ice thickness, wind speed, wind direction angle, temperature change rate, and de-icing mode from the joint distribution function as input to the trained structural response prediction model.

[0093] Among them, the joint distribution function is a function that describes the relationship between the values ​​of multiple random variables.

[0094] In this embodiment, a joint probability distribution is established based on the current ice thickness, the wind speed, wind direction angle, temperature change rate in the meteorological data predicted values ​​after the current moment, and the current de-icing mode to obtain a joint distribution function. The ice thickness, wind speed, wind direction angle, temperature change rate and de-icing mode are extracted from the joint distribution function as inputs to the trained structural response prediction model, which can improve the accuracy of the model prediction.

[0095] Exemplarily, a method for early warning of deicing of a transmission tower specifically includes:

[0096] Step 202: construct a first training data set, which includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data.

[0097] Among them, the meteorological data obtained include but are not limited to temperature, humidity, wind speed, wind direction, pH value, the angle between the main wind direction and the line direction, and solar radiation flux; the radar remote sensing data obtained include but are not limited to reflectivity factor and liquid water content; the satellite remote sensing data obtained include but are not limited to cloud top temperature and cloud phase; the analytical data obtained include but are not limited to boundary layer height and vertical temperature gradient; it is also necessary to obtain key meteorological derivative data from the four types of data obtained above, and the meteorological derivative data include but are not limited to atmospheric heating rate, inversion layer intensity, ice-water conversion index, supercooled water content, wind shear edge, and turbulent kinetic energy.

[0098] The acquired meteorological data can be connected to the local meteorological database through the network so as to obtain and update the meteorological data in real time.

[0099] The process of constructing the first training dataset specifically involves establishing a regular grid, for example, 1 km × 1 km, and uniformly interpolating the various data points acquired above onto the grid points. A three-dimensional variational assimilation method is used to resolve conflicts between different data sources. Specifically, ground station data (i.e., meteorological data) is used as the absolute benchmark, with a first weighting (e.g., 0.7). Radar remote sensing data is optimally interpolated in the vertical direction, with a second weighting (e.g., 0.5). Satellite data is weighted down to a third weighting (e.g., 0.3) in cloud-free areas. Satellite data is weighted to the first weighting (e.g., 0.7) in clouded areas. The weights of both analytical and meteorologically derived data are set to the first weighting, generating a fused multidimensional feature field. This includes near-surface (e.g., 10-meter) temperature, wind speed vector, ice-water mixing ratio (e.g., the integral of the vertical radar reflectivity profile over the -5°C to 0°C layer), and a thermal instability index (e.g., the potential temperature difference between 925 hPa and the ground).

[0100] Step 204, construct a meteorological forecast model, which includes a long short-term memory network, a Transformer model and a cross-attention mechanism module. The input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the Transformer model respectively. The input of the long short-term memory network is the meteorological data of the set time period, and the input of the Transformer model is the radar remote sensing data, satellite remote sensing data, analysis data and meteorological derivative data of the set time period. The output of the cross-attention mechanism module is the predicted value of the meteorological data.

[0101] The weather forecast model uses a dual-channel spatiotemporal hybrid model, which includes a long short-term memory (LSTM) network, a Transformer model, and a cross-attention mechanism module. The input of the cross-attention mechanism module is connected to the output of the LSTM network and the output of the Transformer model, respectively. The LSTM branch processes meteorological data (for example, 6 hours of historical meteorological data with a temporal resolution of 10 minutes), using 128 hidden units and a dropout of 0.2. Input features can include current values ​​and first-order differences of temperature, wind speed, humidity, and air pressure.

[0102] The Transformer model branch processes radar remote sensing data (e.g., a 5 km × 5 km area with a 1-hour time step), satellite remote sensing data, analytical data, and meteorologically derived data. The Transformer model uses a 4-layer encoder and an 8-head multi-attention system. Input features include radar remote sensing data, satellite remote sensing data, analytical data, and meteorologically derived data. For example, reflectivity profiles from satellite remote sensing data, cloud top temperature from meteorological data, and wind divergence from meteorologically derived data can be selected.

[0103] Finally, the dynamic coupling of ground and remote sensing data is achieved through the Cross-Attention module. The output of the Cross-Attention module is the weather data forecast, for example, weather data for the next six hours.

[0104] Step 206: Train a meteorological prediction model using the first training data set to obtain a trained meteorological prediction model; obtain meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period before the current moment of the transmission tower to be predicted, use the trained meteorological prediction model to output a predicted value of the meteorological data after the current moment, and determine the current de-icing mode based on the predicted value of the meteorological data after the current moment.

[0105] Among them, the de-icing sufficiency prediction method uses the predicted value of meteorological data, de-icing probability, and inversion layer thickness to conduct time series trend analysis (see Figure 2) to determine the de-icing mode. The prediction method specifically includes: obtaining temperature, wind speed and solar radiation flux from the meteorological data forecast value after the current moment, determining the de-icing probability based on the wind speed (for example, when the wind speed is greater than 15m / s, the corresponding de-icing probability is greater than 0.8), and calculating the inversion layer thickness and ice-water mixing ratio based on the meteorological data forecast value through relevant calculation methods; the judgment conditions include, for example, the predicted temperature is continuously greater than 0℃ in the next 3 hours, the near-ground wind speed is between 2-12m / s, the de-icing probability output by the model is greater than 0.65, and the solar radiation flux is greater than 300W / m 2 (Daytime), inversion layer thickness < 200m, and ice-water mixing ratio decrease rate > 15% per hour. If the temperature, wind speed, solar radiation flux, de-icing probability, inversion layer thickness, and ice-water mixing ratio in the meteorological data forecast meet these criteria, the de-icing mode corresponding to these conditions is output and used as the current de-icing mode, combined with on-site line parameters (such as topography and terrain).

[0106] Step 208: Construct finite element models of different transmission towers. The input of the finite element model is the deicing condition, and the output is the response parameter. The deicing condition includes ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode. Different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct a second training data set.

[0107] Specifically, if Figure 2 As shown, a parametric modeling approach is used to partition the finite element model database of a typical transmission tower structure into four levels of parameters: geometric parameters (e.g., tower height, root span dimensions, and crossarm spacing); topological parameters (e.g., crossarm configuration, web member system, and number of ground wire attachment points); material parameters (e.g., main material specifications, diagonal member diameter, and gusset plate thickness); and load parameters (e.g., ground wire tension and wind load distribution). Based on these four levels of parameters, a geometric model of the transmission tower is constructed. Constraints defined in design specifications (e.g., those related to transmission towers) are used to define multi-parameter deicing conditions. Dynamic time history analysis is performed using a dynamic time history analysis module (based on an automated script in ANSYS APDL) to generate a parameter-driven finite element model. This dynamic time history analysis utilizes dynamic unloading simulations of ice accretion loads.

[0108] The input of the finite element model is the de-icing condition, and the output is the response parameters. Optionally, the response parameters can be derived using a hybrid method of proper orthogonal decomposition (POD) and dynamic modal decomposition (DMD). This method extracts the first 100 modes of the structure, calculates the participation factor for each mode, and retains modes with a participation factor greater than 5% for the finite element model simulation, thereby improving the speed and accuracy of the finite element calculation.

[0109] By setting different deicing conditions and performing calculations using the finite element model, corresponding response parameters are output. Each deicing condition and its corresponding response parameters form a high-quality structural response database (i.e., the second training dataset).

[0110] The typical operating condition module in step 3 should include factors such as different ice thickness, wind speed, wind direction, and temperature change rate. The specific implementation method can be carried out according to Table 1.

[0111] Table 1 Input operating conditions

[0112]

[0113] In this embodiment, the response parameters output by the finite element model (also called structural response time history data) include key node displacements (cross arm ends, tower tops), member stresses (main members, diagonal members), accelerations, base reactions, etc.

[0114] In this embodiment, after obtaining the second training data set, the second training data set can also be subjected to hierarchical management, physical rationality verification, and abnormal data processing. Specifically, hierarchical management includes: setting the second training data set as a training set (80%), covering conventional working conditions, for basic training of the model; a validation set (10%), including boundary working conditions, for parameter adjustment; and a test set (10%), retaining extreme working conditions, for evaluating the results of data training. The physical rationality verification should perform an energy conservation check to determine whether the error between the input wind energy + ice load work and the structural strain energy + damping dissipation energy is less than 5%, and perform a modal consistency test. For abnormal data processing: use the 3σ principle to filter abnormal data, eliminate samples with abnormally large displacements / stresses, and perform mechanical logic verification at the same time. For example, if the cross arm displacement under a certain working condition is greater than the tower top displacement, it will be judged as unreasonable and eliminated.

[0115] Step 210 : constructing a structural response prediction model. The input of the structural response prediction model is the de-icing condition, and the output is the response parameter. The structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model.

[0116] The trained structural response prediction model learns the mapping relationship between input de-icing conditions and output response parameters. In some cases, the model extracts time-domain features (peak-to-valley values, root mean square (RMS) values), frequency-domain features, and nonlinear features (such as Lyapunov exponents) of the response parameters to train the mapping relationship between the input de-icing conditions and the output response parameters.

[0117] The structural response prediction model (also known as the physics-guided neural network) can adopt the CNN-LSTM-Attention model (convolutional long short-term memory attention model).

[0118] In some examples, the constructed structural response prediction model can employ a multi-branch hybrid network structure, embedding physical laws into the core framework of the data-driven model. A spatiotemporal convolution-attention module processes input operating parameters (such as ice thickness and wind speed) and predicted structural response time history data (such as displacement and stress). Three-dimensional convolution captures local mechanical characteristics, and a multi-head attention mechanism focuses on key time steps and sensitive measurement points. After the displacement field is output from the network's intermediate layer, the product of the stiffness matrix and the load vector is automatically calculated and compared with the measured nodal forces to generate equilibrium residuals. The Von Mises yield condition is applied to the predicted stress tensor, triggering a nonlinear penalty term when the stress intensity exceeds the material's yield limit (e.g., 345 MPa for Q345 steel).

[0119] The training loss function can be a single loss function, such as the mean squared error (MSE) to measure the deviation between the prediction and the true response. Alternatively, the training loss function can be multiple loss functions, such as setting the mean squared error (MSE) to measure the deviation between the prediction and the true response, defining the equilibrium residual norm, yield violation, and energy conservation error. The final loss function is defined as a weighted sum of these terms.

[0120] In some embodiments, transfer learning can also be used to train the structural response prediction model. Specifically, the weights of the underlying spatiotemporal feature extraction layers (e.g., convolutional kernels and LSTM units) in the backbone network are retained. The mechanical response patterns learned by these layers (e.g., vibration propagation and stress concentration patterns) are universal across tower types. The dedicated adaptation layer is fine-tuned, adding learnable parameters to map input dimensions (e.g., tower height and root span) into a normalized latent space, eliminating scale differences between tower types. A graph attention network (GAT) dynamically adjusts node feature weights to accommodate load transfer path variations with different crossarm configurations. Subsequently, a small-sample fine-tuning process is performed to complete the physics-guided network training on the backbone tower database (over 10,000 load cases). Deep features are extracted from a small amount of data (50-100 load cases) from the new tower type using the universal feature layer. The maximum mean difference (MMD) is used to measure the distribution difference from the backbone data. A gradient reversal layer (GRL) is introduced to minimize the feature distribution difference between the new and old tower types through domain adversarial training, while simultaneously optimizing the parameters of the dedicated adaptation layer. Finally, using the full-operational validation set for the new tower type, fine-tuning was performed with stronger physical constraint weights to ensure that the migrated model still adhered to mechanical principles. Finally, a dynamic update mechanism was introduced. When the new tower type detected samples with low prediction confidence (e.g., entropy > threshold), finite element simulations were automatically triggered to generate supplementary data, which was then manually reviewed and added to the training set. Elastic Weight Consolidation (EWC) technology was employed to constrain the fluctuation range of important parameters (identified by the Fisher Information Matrix) during fine-tuning to prevent the loss of existing knowledge. This enabled hierarchical parameter transfer learning for modular design, enabling generalization from the main tower type to the new tower structure.

[0121] Step 212: Obtain the current ice thickness of the transmission tower to be predicted. Input the current ice thickness, wind speed, wind direction, temperature change rate, and current de-icing mode from the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values.

[0122] In step 212, the current ice thickness of the transmission tower to be predicted is obtained in real time. The current ice thickness, the wind speed, wind direction angle, temperature change rate in the meteorological data prediction value obtained above after the current moment, and the current de-icing mode obtained by discrimination are input into the trained structural response prediction model to obtain the response parameter prediction value.

[0123] In some embodiments, as Figure 2 As shown, an uncertainty propagation model is set up. The uncertainty propagation model includes a variety of distribution processing. For example, considering random uncertainty, a joint probability distribution is established for the input operating condition parameters of the trained structural response prediction model. For example, the temperature can be taken as a truncated normal distribution T~N(μ,σ 2 )[-20,5]℃, wind speed can be assumed to be a three-parameter Weibull distribution W(k, λ, θ), and ice thickness can be assumed to be a Gamma distribution Γ(α, β). Ice thickness, wind speed, wind direction, temperature change rate, and deicing pattern are extracted from the joint distribution function as inputs to the trained structural response prediction model to obtain the corresponding response parameter prediction values.

[0124] In addition, considering epistemic uncertainty, the prediction interval of the deep learning model output can be obtained through MC Dropout, and the posterior distribution of the response parameter prediction value (such as maximum stress) output by the trained structural response prediction model can be fitted using t distribution.

[0125] Step 214 , substitute the predicted values ​​of the response parameters into various failure functions to calculate failure values, compare the failure values ​​with corresponding warning thresholds, and output warning results.

[0126] The warning results include hazard levels and structural failure modes. The hazard levels are classified into three levels: safe, warning, and high-risk. Structural failure modes include bending, torsion, and buckling.

[0127] In some embodiments, as Figure 2 As shown, a probabilistic reliability analysis method is set up. This method is based on the structural safety criterion analysis framework and establishes a multi-modal limit state function system, including: strength failure function, stability failure function and fatigue failure function.

[0128] The trained structural response prediction model is used as input to obtain predicted response parameter values. These predicted response parameter values ​​are then substituted into the strength failure function, stability failure function, and fatigue failure function to calculate the corresponding failure values. The failure values ​​are then compared with the corresponding warning thresholds to output a warning result, thereby implementing a warning criterion (i.e., early warning determination) for the possibility of ice shedding on the transmission tower. This consideration includes a safety margin for uncertainty, and the warning threshold can be a dynamic warning threshold.

[0129] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0130] This method considers regional differences in meteorological parameters and the uncertainty of their rate of change. It uses a long short-term memory (LSTM) or Transformer time series model to predict trends in temperature, wind speed, and other parameters over the next several hours. A deicing sufficiency prediction method based on time series trend modeling is proposed. A finite element model database covering various typical transmission tower structures is constructed, and a parametric modeling mechanism is introduced to enable automatic model reconstruction and expansion. This adapts to the differences in forces acting on different tower types during deicing. Using the finite element model, extensive dynamic time-history analysis is conducted to analyze typical operating conditions, such as ice thickness, wind speed, wind direction, and temperature change rate, to determine the mechanical responses of key tower components under varying deicing conditions. A large-scale dataset is established, capturing the relationship between input operating conditions and output response parameters. Based on the finite element calculation results, a physics-guided neural network is constructed to rapidly fine-tune the original model using a small amount of new tower data, achieving generalization from mainstream tower types to new tower structures. A probabilistic reliability analysis method is introduced, combining the uncertainty distribution of the physics-guided neural network output to estimate the probability of a tower reaching a failure state. Taking into account the fluctuation range of input operating conditions, a more robust reliability early warning criterion for transmission tower deicing is constructed using Monte Carlo sampling + uncertainty propagation model (Bayesian Inference).

[0131] The method of the present invention first realizes the rapid reconstruction and efficient calculation of different tower structures through parametric modeling-constructed finite element database, which significantly improves the simulation efficiency; the de-icing criterion method that integrates multi-source meteorological data and deep learning time series prediction can more accurately capture regional meteorological characteristics and dynamic change trends, and greatly improve the prediction accuracy of natural de-icing patterns; the combination of physical constraint-based neural network design and transfer learning mechanism not only ensures that the model prediction results conform to the laws of mechanics, but also enhances the generalization ability across tower types; finally, the evaluation framework of probabilistic reliability analysis and Bayesian uncertainty propagation is introduced to establish a more scientific and robust risk warning system by quantifying the combined impact of cognitive uncertainty and random uncertainty. The overall solution realizes full-chain technological innovation from meteorological monitoring, mechanical response prediction to risk assessment, while ensuring engineering accuracy while improving computing efficiency by an order of magnitude, providing intelligent and scalable technical support for power grid anti-icing and disaster reduction decision-making.

[0132] The method of this invention encompasses a parametric modeling mechanism for various typical transmission tower structures, the construction of meteorological criteria for deicing sufficiency, dynamic time-history analysis, the construction and training of a deep learning model, and a rapid reliability assessment method for transmission tower structures. First, a finite element model database of typical transmission tower structures is established. Based on meteorological data, a deicing sufficiency criterion is constructed. Subsequently, finite element dynamic time-history simulations are performed to generate training samples. Finally, a deep learning model is constructed and trained, ultimately resulting in a method for rapidly assessing the reliability of transmission tower structures under deicing conditions.

[0133] The system has the following advantages: A three-dimensional assimilation system for ground stations, radar reflectivity, and satellite cloud images has been constructed. A spatiotemporal alignment algorithm based on an attention mechanism has been developed to achieve deep coupling of multi-source meteorological features, improving the accuracy of de-icing predictions. A parameterized model library has been established, enabling automatic reconstruction of finite element models through parameterized associations of 18 design variables, such as tower height and crossarm configuration. The integration of order reduction techniques significantly reduces simulation computation speed while ensuring accuracy. A large-scale dataset has been established, showcasing the relationship between input conditions and output responses. An evaluation framework based on reliability analysis and Bayesian uncertainty propagation has been introduced, enabling automatic and flexible adjustment of warning thresholds to account for uncertainty.

[0134] Based on the same inventive concept, embodiments of the present application also provide a de-icing warning device for a transmission tower, which is used to implement the aforementioned de-icing warning method for a transmission tower. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the de-icing warning device for a transmission tower provided below can be found in the above-mentioned limitations of the de-icing warning method for a transmission tower, and will not be further elaborated here.

[0135] In an exemplary embodiment, Figure 3As shown, a de-icing warning device for a transmission tower is provided, comprising: a first training data set construction module, a first modeling module, a training module, a second training data set construction module, a second modeling module, a real-time data acquisition module, and a warning module, wherein:

[0136] A first training data set construction module is used to construct a first training data set, the first training data set including meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data;

[0137] The first modeling module is used to build a meteorological forecast model. The meteorological forecast model includes a long short-term memory network, a transformer model, and a cross-attention mechanism module. The input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model respectively. The input of the long short-term memory network is the meteorological data of a set time period. The input of the transformer model is the radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data of the set time period. The output of the cross-attention mechanism module is the predicted value of the meteorological data;

[0138] A training module is configured to train a meteorological prediction model using a first training data set to obtain a trained meteorological prediction model; obtain meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted; output a predicted value of the meteorological data after the current moment using the trained meteorological prediction model; and determine a current deicing mode based on the predicted value of the meteorological data after the current moment;

[0139] A second training dataset construction module is used to construct finite element models of different transmission towers. The input of the finite element model is the deicing operating conditions, and the output is the response parameters. The deicing operating conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode. Different deicing operating conditions are input into the finite element model to obtain corresponding response parameters to construct the second training dataset.

[0140] The second modeling module is used to build a structural response prediction model, the input of the structural response prediction model is the de-icing condition, the output is the response parameter, and the structural response prediction model is trained using the second training data set to obtain a trained structural response prediction model;

[0141] The real-time data acquisition module is used to obtain the current ice thickness of the transmission tower to be predicted. The current ice thickness, wind speed, wind direction angle, temperature change rate and current deicing mode in the meteorological data predicted values ​​after the current moment are input into the trained structural response prediction model to obtain the response parameter prediction value;

[0142] The early warning module is used to substitute the predicted value of the response parameter into multiple failure functions to calculate the failure value, compare the failure value with the corresponding early warning threshold, and output the early warning result.

[0143] In an exemplary embodiment, in the first training data set construction module, meteorological data include temperature, humidity, wind speed, wind direction, pH value, the angle between the main wind direction and the line direction, and solar radiation flux; radar remote sensing data include reflectivity factor and liquid water content; satellite remote sensing data include cloud top temperature and cloud phase; analysis data include boundary layer height and vertical temperature gradient; meteorological derivative data include atmospheric heating rate, inversion layer strength, ice-water conversion index, supercooled water content, wind shear and turbulent kinetic energy.

[0144] In an exemplary embodiment, in a first training data set construction module, the first training data set is constructed, including: gridding the area where the transmission tower is located, and uniformly interpolating meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data to corresponding grid points in the grid; using meteorological data as baseline data, setting the weight of meteorological data to a first weight; setting the horizontal weight of radar remote sensing data to a second weight; setting the cloud-free area weight of satellite remote sensing data to a second weight, and setting the weights of analysis data and meteorological derivative data to a first weight, wherein the first weight is greater than the second weight and greater than the third weight; and the meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data after the weight setting are completed constitute the first training data set.

[0145] In an exemplary embodiment, in a training module, the predicted values ​​of meteorological data after the current moment are judged to determine the current de-icing mode, including: obtaining the temperature, wind speed and solar radiation flux from the predicted values ​​of meteorological data after the current moment, determining the de-icing probability based on the wind speed, and calculating the inversion layer thickness and ice-water mixing ratio based on the predicted values ​​of meteorological data; if the temperature, wind speed, solar radiation flux, de-icing probability, inversion layer thickness and ice-water mixing ratio in the predicted values ​​of meteorological data meet the corresponding judgment conditions, then the de-icing mode corresponding to the conditions is output and used as the current de-icing mode.

[0146] In an exemplary embodiment, in the early warning module, the predicted value of the response parameter is substituted into a plurality of failure functions to calculate the failure value, including: obtaining the strength failure function, the stability failure function and the fatigue failure function; and substituting the predicted value of the response parameter into the strength failure function, the stability failure function and the fatigue failure function respectively to calculate the corresponding failure value.

[0147] In an exemplary embodiment, in the real-time data acquisition module, before the current ice thickness, the wind speed, wind direction angle, the temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode are input into the trained structural response prediction model, the module also includes: establishing a joint probability distribution according to the current ice thickness, the wind speed, wind direction angle, the temperature change rate in the meteorological data prediction value after the current moment, and the current de-icing mode to obtain a joint distribution function, and extracting the ice thickness, wind speed, wind direction angle, the temperature change rate, and the de-icing mode from the joint distribution function as the input of the trained structural response prediction model.

[0148] Each module in the aforementioned transmission tower de-icing warning device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0149] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 4 As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless means. The wireless means can be implemented via Wi-Fi, mobile cellular networks, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements a de-icing warning method for transmission towers. The display unit of the computer device is used to produce a visual image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0150] Those skilled in the art will understand that Figure 4The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0151] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0152] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0153] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0154] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0155] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), data processing logic devices based on quantum computing, artificial intelligence (AI) processors, and the like.

[0156] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0157] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for early warning of deicing of transmission towers, characterized in that: The method comprises: Constructing a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data; Constructing a weather forecast model, the weather forecast model including a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively; the input of the long short-term memory network is weather data for a set time period; the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period; and the output of the cross-attention mechanism module is a forecast value of the weather data; The meteorological prediction model is trained using the first training data set to obtain a trained meteorological prediction model; meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted are obtained, a predicted value of meteorological data after the current moment is output using the trained meteorological prediction model, and the predicted value of meteorological data after the current moment is discriminated to determine a current deicing mode; Constructing finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; inputting different deicing conditions into the finite element model to obtain corresponding response parameters to construct a second training data set; Constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition and the output is a response parameter, and training the structural response prediction model using a second training data set to obtain a trained structural response prediction model; Obtaining the current ice thickness of the transmission tower to be predicted, and inputting the current ice thickness, wind speed, wind direction angle, temperature change rate, and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values; Substituting the predicted value of the response parameter into a plurality of failure functions to calculate the failure value, comparing the failure value with the corresponding warning threshold value to output the warning result.

2. The method according to claim 1, characterized in that The meteorological data include temperature, humidity, wind speed, wind direction, pH value, the angle between the main wind direction and the line direction, and solar radiation flux; the radar remote sensing data include reflectivity factor and liquid water content; the satellite remote sensing data include cloud top temperature and cloud phase; the analytical data include boundary layer height and vertical temperature gradient; the meteorological derived data include atmospheric heating rate, inversion layer intensity, ice-water conversion index, supercooled water content, wind shear and turbulent kinetic energy.

3. The method according to claim 1, characterized in that The constructing of the first training data set includes: The area where the transmission tower is located is gridded, and meteorological data, radar remote sensing data, satellite remote sensing data, analytical data, and meteorological derivative data are uniformly interpolated to corresponding grid points in the grid; Taking meteorological data as reference data, the weight of the meteorological data is set to a first weight; the horizontal weight of the radar remote sensing data is set to a second weight; the cloud-free area weight of the satellite remote sensing data is set to a second weight, and the weights of the analytical data and the meteorological-derived data are both set to a first weight, wherein the first weight is greater than the second weight and greater than the third weight; The meteorological data, the radar remote sensing data, the satellite remote sensing data, the analysis data, and the meteorological derivative data after weight setting are used to form a first training data set.

4. The method according to claim 1, wherein The determining of the current de-icing mode based on the meteorological data forecast value after the current moment includes: Obtaining temperature, wind speed, and solar radiation flux from predicted meteorological data after the current moment, determining a de-icing probability based on the wind speed, and calculating an inversion layer thickness and an ice-water mixing ratio based on the predicted meteorological data; If the temperature, wind speed, solar radiation flux, deicing probability, inversion layer thickness, and ice-water mixing ratio in the meteorological data forecast values ​​meet the corresponding judgment conditions, the deicing mode corresponding to the conditions is output and used as the current deicing mode.

5. The method according to claim 1, wherein Substituting the predicted response parameter value into various failure functions to calculate the failure value, including: Obtain strength failure function, stability failure function and fatigue failure function; Substitute the predicted values ​​of the response parameters into the strength failure function, the stability failure function and the fatigue failure function respectively to calculate the corresponding failure values.

6. The method according to claim 1, characterized in that Before inputting the current ice thickness, wind speed, wind direction angle, temperature change rate and current deicing mode in the meteorological data prediction value after the current moment into the trained structural response prediction model, the method further includes: A joint probability distribution is established based on the current ice thickness, the wind speed, wind direction angle, temperature change rate in the meteorological data predicted values ​​after the current moment, and the current deicing mode to obtain a joint distribution function. The ice thickness, wind speed, wind direction angle, temperature change rate and deicing mode are extracted from the joint distribution function as inputs to the trained structural response prediction model.

7. A de-icing warning device for a transmission tower, characterized in that: The device comprises: A first training data set construction module is used to construct a first training data set, wherein the first training data set includes meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data; A first modeling module is used to construct a weather forecast model, wherein the weather forecast model includes a long short-term memory network, a transformer model, and a cross-attention mechanism module, wherein the input end of the cross-attention mechanism module is connected to the output end of the long short-term memory network and the output end of the transformer model, respectively. The input of the long short-term memory network is weather data for a set time period, the input of the transformer model is radar remote sensing data, satellite remote sensing data, analysis data, and weather-derived data for the set time period, and the output of the cross-attention mechanism module is a forecast value of the weather data; a training module configured to train the meteorological prediction model using the first training data set to obtain a trained meteorological prediction model; obtain meteorological data, radar remote sensing data, satellite remote sensing data, analysis data, and meteorological derivative data for a set period of time before a current moment of the transmission tower to be predicted, output a predicted value of meteorological data after the current moment using the trained meteorological prediction model, and determine a current deicing mode based on the predicted value of meteorological data after the current moment; a second training data set construction module, configured to construct finite element models of different transmission towers, wherein the input of the finite element model is deicing conditions, and the output is response parameters, wherein the deicing conditions include ice thickness, wind speed, wind direction angle, temperature change rate, and deicing mode; different deicing conditions are input into the finite element model to obtain corresponding response parameters to construct the second training data set; a second modeling module for constructing a structural response prediction model, wherein the input of the structural response prediction model is a de-icing condition, the output is a response parameter, and the structural response prediction model is trained using a second training data set to obtain a trained structural response prediction model; A real-time data acquisition module is used to obtain the current ice thickness of the transmission tower to be predicted, and input the current ice thickness, wind speed, wind direction angle, temperature change rate and current deicing mode in the meteorological data predicted values ​​after the current moment into the trained structural response prediction model to obtain response parameter prediction values; The early warning module is used to substitute the predicted value of the response parameter into multiple failure functions to calculate the failure value, compare the failure value with the corresponding early warning threshold, and output the early warning result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images