Icing prediction method and device, equipment and storage medium

Abstract

The embodiment of the invention provides an icing prediction method and device, equipment and a storage medium, and the method comprises the steps: obtaining secondary season icing related data, carrying out the convolution operation of the secondary season icing related data, determining the space structure information of a weather system, building a reference mapping model based on the space structure information, and carrying out the calculation of the reference mapping model. Taking a model parameter in the reference mapping model as an initial value, adjusting the initial value by using historical forecast data, obtaining a correction model according to an adjustment result, inputting forecast data to be predicted into the correction model to obtain air temperature forecast and rainfall forecast, and calculating the content of supercooled liquid water according to the air temperature forecast and the rainfall forecast. And inputting the content of the supercooled liquid water into a preset icing growth physical model to obtain an icing thickness growth rate, and predicting icing by using the icing thickness growth rate. The accuracy of sub-season scale icing prediction is improved, the systematic deviation of a power mode is corrected, the reliability of a power grid disaster prevention plan is enhanced, and the operation mode is optimized.

Description

Technical Field

[0001] This disclosure relates to the field of meteorological forecasting technology, and in particular to an icing forecasting method, apparatus, equipment and storage medium. Background Technology

[0002] Conductor icing is one of the major natural disasters threatening the safe and stable operation of power grids. Accurate icing forecasts, especially sub-seasonal forecasts several weeks to a month in advance, are crucial for power grids to develop disaster prevention plans and optimize operation.

[0003] Currently, subseasonal forecasting of icing meteorological elements mainly relies on two types of models: one is subseasonal dynamical models based on physical processes, whose advantage lies in their clearly defined physical mechanisms and ability to provide long-term forecasts; the other is AI models driven by historical data, whose advantage lies in their ability to uncover complex nonlinear relationships. However, dynamical models inevitably suffer from systematic biases at the subseasonal scale, leading to inaccurate forecasts of key meteorological elements such as temperature and precipitation, directly reducing the reliability of icing predictions based on their outputs. Secondly, while purely data-driven AI models are flexible, their performance is highly dependent on a large amount of high-quality paired data. In practice, such long-term, complete historical datasets of subseasonal dynamical model forecasts are often difficult to obtain, limiting the application and performance of AI models in subseasonal forecasting. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this disclosure provides an icing prediction method, apparatus, device and storage medium.

[0005] This disclosure provides an ice accretion prediction method, the method comprising: Acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The model parameters in the baseline mapping model are used as initial values, and the initial values ​​are adjusted using the historical forecast data. The corrected model is then derived based on the adjustment results. Input the data to be predicted into the correction model to obtain temperature and precipitation forecasts; The supercooled liquid water content is calculated based on the temperature and precipitation forecasts. The supercooled liquid water content is then input into a preset icing growth physical model to obtain the icing thickness growth rate. The icing thickness growth rate is then used to predict icing.

[0006] The method provided in this disclosure includes acquiring sub-seasonal icing-related data, performing a convolution operation on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and constructing a benchmark mapping model based on the spatial structure information, comprising: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

[0007] The method provided in this disclosure uses model parameters in a baseline mapping model as initial values, adjusts the initial values ​​using historical forecast data, and derives a corrected model based on the adjustment results, including: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using preset adjustment hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

[0008] The method provided in this disclosure, which inputs the data to be predicted into the correction model to obtain temperature and precipitation forecasts, includes: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

[0009] The method provided in this disclosure calculates the supercooled liquid water content based on the temperature and precipitation forecasts, inputs the supercooled liquid water content into a preset icing growth physical model, obtains the icing thickness growth rate, and uses the icing thickness growth rate to predict icing, including: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

[0010] This disclosure also provides an icing prediction device, the device comprising: The module is used to acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The adjustment module is used to take the model parameters in the baseline mapping model as initial values, adjust the initial values ​​using the historical forecast data, and derive the corrected model based on the adjustment results. The forecast module is used to input the forecast data to be predicted into the correction model to obtain temperature forecast and precipitation forecast; The calculation module is used to calculate the supercooled liquid water content based on the temperature forecast and precipitation forecast, input the supercooled liquid water content into a preset icing growth physical model, obtain the icing thickness growth rate, and use the icing thickness growth rate to predict icing.

[0011] The apparatus provided in this disclosure, wherein the construction module is specifically used for: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

[0012] The device provided in this disclosure, wherein the prediction module is specifically used for: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using feature hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

[0013] The device provided in this disclosure, wherein the prediction module is specifically used for: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

[0014] The apparatus provided in this disclosure, wherein the computing module is specifically used for: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

[0015] This disclosure also provides an electronic device, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the icing prediction method provided in this disclosure.

[0016] This disclosure also provides a computer-readable storage medium storing a computer program for performing the icing prediction method as provided in this disclosure.

[0017] The technical solution provided in this disclosure has the following advantages compared with the prior art: The icing prediction method provided in this disclosure acquires sub-seasonal icing data, extracts the spatial structure of the weather system through convolution operations, constructs a baseline mapping model, adjusts the model parameters using historical forecast data to obtain a corrected model, inputs historical forecast data to output temperature and precipitation, and calculates the icing thickness growth rate using a physical model of icing growth, combined with the supercooled liquid water content, to achieve icing prediction. This improves the accuracy of sub-seasonal icing prediction, corrects systematic biases in dynamic models, reduces reliance on large amounts of paired data, enhances the reliability of power grid disaster prevention plans, and optimizes operational modes. Attached Figure Description

[0018] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0019] Figure 1 A schematic flowchart illustrating the icing prediction method provided in this embodiment of the disclosure; Figure 2 This is a schematic diagram of the structure of the icing prediction device provided in the embodiments of this disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0020] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0021] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0022] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0023] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0024] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0025] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0026] To address the aforementioned problems, this disclosure provides an icing prediction method, which will be described below with reference to specific embodiments.

[0027] Figure 1This is a flowchart illustrating an icing prediction method provided in an embodiment of the present disclosure. The method can be executed by an icing prediction device, which can be implemented in software and / or hardware and is generally integrated into an electronic device.

[0028] Example 1: This embodiment of the disclosure provides an icing prediction method, the method comprising: Acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The model parameters in the baseline mapping model are used as initial values, and the initial values ​​are adjusted using the historical forecast data. The corrected model is then derived based on the adjustment results. Input the data to be predicted into the correction model to obtain temperature and precipitation forecasts; The supercooled liquid water content is calculated based on the temperature and precipitation forecasts. The supercooled liquid water content is then input into a preset icing growth physical model to obtain the icing thickness growth rate. The icing thickness growth rate is then used to predict icing.

[0029] In this embodiment, historical reanalysis data is collected. The variables in the historical reanalysis data include gridded data such as sea level pressure, 500hPa / 700hPa / 850hPa geopotential height, temperature, humidity, and wind field. These serve as input features for the baseline mapping model during the pre-training phase, characterizing the atmospheric circulation state.

[0030] In this embodiment, the meteorological station observation data is the daily temperature and precipitation observation data of national meteorological stations collected and reanalyzed at the same time, which serves as the prediction target of the benchmark mapping model in the pre-training stage.

[0031] In this embodiment, the historical forecast data for the sub-seasonal to seasonal dynamic model is the historical forecast data collected from the sub-seasonal to seasonal (S2S) dynamic model. The historical forecast data corresponds to the variables in the historical reanalysis data and serves as input data for the adjustment phase and the final forecast.

[0032] In this embodiment, spatial structure information is an abstract feature extracted after the convolution operation. For example, in the feature map output from the deep layers of the convolutional network, highly activated regions may correspond to key spatial configurations of weather systems such as blocking high pressure or deep low-pressure troughs.

[0033] In this embodiment, the baseline mapping model is a deep convolutional neural network trained on historical reanalysis data, used to establish the statistical relationship between the atmospheric circulation field and local meteorological elements. The input is historical reanalysis data, and the output is meteorological station observation data for the corresponding time period.

[0034] In this embodiment, the model parameters are all the learnable weights and biases in the neural network. The initial values ​​refer to the numerical values ​​of these parameters loaded from the baseline mapping model. For example, a convolutional layer's kernel weight matrix [[0.1,0.2],[0.3,0.4]] and a bias term [0.05]. At the start of tuning, these values ​​are directly used from the values ​​trained on the baseline model, rather than being randomly initialized.

[0035] In this embodiment, the corrected model is the final product of the adjustment process, i.e., the optimal adjusted model. The input is new, unpredictable S2S dynamical model forecast data, and the output is the corrected meteorological elements (such as temperature forecast and precipitation forecast). This output has been calibrated and is closer to the actual situation.

[0036] In this embodiment, the baseline mapping model parameters are loaded as initial values, and the input is switched to historical forecast data. The loss is calculated through forward propagation using the adjusted training set, and the high-level parameters of the network are updated through backpropagation to obtain the adjusted values. The optimal model is saved as the correction model by using the adjusted validation set to trigger an early stopping mechanism.

[0037] In this embodiment, the temperature forecast and precipitation forecast are the final, high-quality forecast products that can be used for operations, generated by combining the initial forecast with the correction.

[0038] In this embodiment, the supercooled liquid water content is the mass of water droplets per unit volume of air that are below the freezing point but still in a liquid state.

[0039] In this embodiment, the preset icing growth physical model is a combination of mathematical equations constructed based on the thermodynamic equilibrium principle of wire icing. It is used to simulate the physical process of water droplet collision-capture-freezing. The inputs are supercooled liquid water content, freezing efficiency coefficient, collision efficiency, effective capture area, icing thickness growth rate, and meteorological conditions during the target forecast period. The outputs are time step growth, continuous prediction sequence of icing thickness, and dynamic cumulative thickness.

[0040] The working principle and beneficial effects of this embodiment are as follows: Sub-seasonal icing data is acquired; the spatial structure of the weather system is extracted through convolution operations; a baseline mapping model is constructed; historical forecast data is used to adjust the model parameters to obtain a corrected model; historical forecast data is input to output temperature and precipitation; and the icing thickness growth rate is calculated using a physical model of icing growth, combined with the supercooled liquid water content, thereby achieving icing prediction. This improves the accuracy of sub-seasonal icing prediction, corrects systematic biases in dynamic models, reduces reliance on large amounts of paired data, enhances the reliability of power grid disaster prevention plans, and optimizes operational modes.

[0041] Example 2: The method provided in this embodiment of the present disclosure acquires sub-seasonal icing-related data, performs convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and constructs a benchmark mapping model based on the spatial structure information, including: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

[0042] In this embodiment, the formula corresponding to the convolution operation is: , where F represents the input feature map; K represents the preset convolution kernel; (i,j) represents the position of the input feature map; (m,n) represents the index of the preset convolution kernel K for cyclic accumulation, and the values ​​of m and n are determined by the size of the preset convolution kernel.

[0043] In this embodiment, the preset latitude and longitude grid is a predefined regular spatial grid that covers the target area. For example, a latitude and longitude grid covering a certain region will have all meteorological field data uniformly interpolated onto this grid.

[0044] In this embodiment, the processing result is a standardized, high-quality dataset obtained after the above interpolation, alignment, and elimination. For example, a data table containing daily ERA5 circulation fields, S2S historical forecast fields, and precisely matched station icing observations for a certain year.

[0045] In this embodiment, the atmospheric circulation field is a meteorological field describing the large-scale atmospheric motion. For example, the 500 hPa geopotential height field, with its fluctuating high-pressure ridges and low-pressure troughs, is the engine driving weather changes; icing events are actual icing phenomena observed at a station. For example, on a certain day, month, and year, a meteorological station recorded an icing thickness of 15 mm; the correlation is the statistical association between the two. For example, when a west-high, east-low circulation pattern appears in the 500 hPa geopotential height field, the probability of an icing event occurring at a certain station in the next 10-30 days increases significantly.

[0046] In this embodiment, the input feature map is the processed grid data treated as an image. For example, the sea level pressure field data of a certain area is used as a single-channel grayscale image input to the model.

[0047] In this embodiment, the preset convolution is a small weight matrix that slides across the feature map to detect specific patterns. For example, a 3x3 convolution kernel with a weight distribution might make it particularly sensitive to circular closed isobars, i.e., the center of a pressure system.

[0048] In this embodiment, the core feature representation is a digital and abstract expression of spatial structure information. For example, a feature vector of length 512, where each value encodes the degree to which a certain circulation pattern contributes to icing. The target variable is the objective that the model needs to predict. For example, icing thickness or the presence or absence of icing in meteorological station observation data.

[0049] The working principle and beneficial effects of this embodiment are as follows: Historical reanalysis data, historical forecast data, and station observation data are processed to establish a connection between the atmospheric circulation field and icing events. Based on the input feature map, spatial structure information of the weather system is extracted using convolutional kernels, and a benchmark mapping model is constructed by combining station observation target variables. This achieves a high-precision correlation mapping between meteorological elements and icing phenomena, improves the physical interpretability of sub-seasonal forecasts, provides a reliable benchmark for subsequent dynamic model bias correction, and enhances the power grid's icing risk early warning capability.

[0050] Example 3: The method provided in this embodiment of the present disclosure uses the model parameters in the baseline mapping model as initial values, adjusts the initial values ​​using the historical forecast data, and derives a corrected model based on the adjustment results, including: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using preset adjustment hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

[0051] In this embodiment, the preset adjustment hyperparameter is used to control the adjustment of the training process, rather than the configuration parameters of the model itself.

[0052] In this embodiment, the adjusted model is a model loaded with initial values ​​and prepared for training with historical forecast data. The input is historical forecast data, and the output is the predicted meteorological elements.

[0053] In this embodiment, adjusting the partition is the process of dividing historical forecast data into two parts in chronological order. The training set is adjusted to use a subset of data to update model parameters. For example, historical forecast data from year A to year B. The validation set is adjusted to use a subset of data to monitor training progress and prevent overfitting. For example, historical forecast data from year B+1 to year C.

[0054] In this embodiment, the loss function between the predicted output and the actual observation is a mathematical function used to quantify the difference between the model's predicted output and the actual observed value. For example, mean squared error. If the model predicts a temperature of 5°C, while the actual observation is 3°C, then the loss for this prediction is (5-3). 2 =4. The goal of training is to minimize the average loss of all predictions.

[0055] In this embodiment, the higher-level parameters of the network refer to the parameters of the layers closer to the output end in the model, such as the last fully connected layer.

[0056] In this embodiment, the adjustment value is the gradient of the loss function calculated using the backpropagation algorithm, which is used to update these high-level parameters. For example, it is calculated that the weights need to be increased by 0.001 or decreased by 0.002.

[0057] In this embodiment, the validation loss is a loss function value calculated using an adjusted validation set, used to determine the model's performance on unseen data. The main signal triggering early stopping is when the validation loss stops decreasing.

[0058] The working principle and beneficial effects of this embodiment are as follows: The baseline mapping model parameters are loaded as initial values, and the input is switched to historical forecast data. The loss is calculated through forward propagation using an adjusted training set, and the adjusted values ​​are obtained by updating the high-level parameters of the network through backpropagation. An early stopping mechanism is triggered using an adjusted validation set to save the optimal model as a correction model. This achieves accurate correction of the dynamic model system bias, improves the accuracy of next-season meteorological element forecasts, reduces reliance on reanalysis data, and enhances the reliability and operational application capabilities of icing forecasts.

[0059] Example 4: The method provided in this embodiment of the present disclosure inputs the forecast data to be predicted into the correction model to obtain temperature forecast and precipitation forecast, including: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

[0060] In this embodiment, the initial temperature and initial precipitation are raw, unprocessed forecast values ​​directly calculated and output by the S2S dynamical model. For example, initial temperature: the model forecasts an average temperature of -2.5°C for a certain region in the north during the fourth week of the future. Initial precipitation: the model forecasts 45 mm of precipitation for a certain region in the south during the third week of the future.

[0061] In this embodiment, the corrected temperature and corrected precipitation functions are the correction amounts and functions calculated by the correction model based on the learned bias patterns, adjusting the initial values ​​accordingly. They represent the direction and magnitude of the adjustment the AI ​​model believes the original forecast needs. For example, a corrected temperature of +1.2°C indicates that the model believes the original forecast may be too cold and requires warming. A corrected precipitation function of 0.8 indicates that the model believes the original forecast may be too wet and requires adjustment.

[0062] In this embodiment, the temperature forecast and precipitation forecast are final, operationally usable, high-quality forecast products generated by combining the initial forecast with corrections. Temperature forecast = initial temperature + corrected temperature. Precipitation forecast = corrected precipitation function (initial precipitation).

[0063] In this embodiment, the formulas corresponding to temperature forecast and precipitation forecast are as follows: ,in, This indicates the temperature forecast value; Indicates the initial temperature value; Indicates the corrected temperature reading; Indicates the corrected precipitation function; This indicates the predicted precipitation value; This represents the initial precipitation value.

[0064] The working principle and beneficial effects of this embodiment are as follows: Initial temperature and initial precipitation are obtained based on a sub-seasonal dynamic model. The forecast data to be predicted is input into a correction model to obtain corrected temperature and corrected precipitation functions. Combined with the initial values, the final temperature and precipitation forecasts are derived. This effectively corrects the systematic bias of the dynamic model at the sub-seasonal scale, improves the accuracy and reliability of temperature and precipitation forecasts, and provides more accurate meteorological input for icing prediction.

[0065] Example 5: The method provided in this embodiment calculates the supercooled liquid water content based on the temperature and precipitation forecasts, inputs the supercooled liquid water content into a preset icing growth physical model, obtains the icing thickness growth rate, and uses the icing thickness growth rate to predict icing, including: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

[0066] In this embodiment, relative humidity data represents the percentage of water vapor saturation in the air. It is a key factor in calculating the liquid water content in clouds. Higher humidity increases the likelihood of supercooled water droplets forming, and the content is typically higher. For example, a relative humidity of 85% obtained from a weather forecast.

[0067] In this embodiment, the temperature condition is a threshold range of air temperature used to determine whether supercooled liquid water can exist. Supercooled water is only stable within a specific low-temperature range. Outside this range, water droplets will freeze or evaporate rapidly. For example, if the forecast temperature is -5°C, this temperature falls within the critical range for the existence of supercooled water.

[0068] In this embodiment, the target forecast period refers to the specific future time range targeted by the sub-seasonal scale icing prediction.

[0069] In this embodiment, the initial icing growth rate is the rate of increase in icing thickness or mass calculated based on current meteorological conditions and physical parameters at a certain instant.

[0070] In this embodiment, the continuous prediction sequence is a complete prediction curve of the ice thickness evolution over time, rather than a single value. For example, a data sequence shows no ice in week 1, ice starting to appear in the middle of week 2, gradually increasing in thickness to 8.5 mm by the end of week 3, and then melting starting in week 4 due to rising temperatures.

[0071] In this embodiment, the formula corresponding to the ice thickness growth rate is: ,in, Indicates the thickness of the ice layer; Indicates collision efficiency; Indicates freezing efficiency; The value represents the supercooled liquid water content; U represents the wind speed; A represents the cross-sectional area of ​​the icing surface. This indicates the density of ice.

[0072] The working principle and beneficial effects of this embodiment are as follows: Based on temperature and precipitation forecasts and relative humidity, the supercooled liquid water content is calculated. An initial growth rate is derived through an icing growth model, combining freezing efficiency, collision efficiency, and capture area. The thickness growth rate is obtained by integrating meteorological conditions over the target period, and an icing thickness prediction sequence is generated through time integration. This achieves quantitative prediction of icing thickness at the sub-seasonal scale, integrates meteorological conditions and physical processes, improves the physical consistency and accuracy of the prediction results, and provides a forward-looking decision-making basis for power grid disaster prevention.

[0073] To achieve the above embodiments, this disclosure also proposes an icing prediction device.

[0074] Figure 2 This is a schematic diagram of the structure of an icing prediction device provided in an embodiment of this disclosure. The device can be implemented by software and / or hardware, and is generally integrated into an electronic device. Figure 2 As shown, the device includes: a construction module, an adjustment module, a forecasting module, and a calculation module, wherein, The module is used to acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The adjustment module is used to take the model parameters in the baseline mapping model as initial values, adjust the initial values ​​using the historical forecast data, and derive the corrected model based on the adjustment results. The forecast module is used to input the forecast data to be predicted into the correction model to obtain temperature forecast and precipitation forecast; The calculation module is used to calculate the supercooled liquid water content based on the temperature forecast and precipitation forecast, input the supercooled liquid water content into a preset icing growth physical model, obtain the icing thickness growth rate, and use the icing thickness growth rate to predict icing.

[0075] The apparatus provided in this disclosure, wherein the construction module is specifically used for: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

[0076] The device provided in this disclosure, wherein the prediction module is specifically used for: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using feature hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

[0077] The device provided in this disclosure, wherein the prediction module is specifically used for: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

[0078] The apparatus provided in this disclosure, wherein the computing module is specifically used for: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

[0079] The icing prediction device provided in this disclosure can execute the icing prediction method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the method execution.

[0080] To implement the above embodiments, this disclosure also proposes a computer program product, including a computer program / instructions, which, when executed by a processor, implements the icing prediction method in the above embodiments.

[0081] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure.

[0082] The following is a detailed reference. Figure 3 The diagram illustrates a structural schematic suitable for implementing the electronic device 300 in the embodiments of this disclosure. The electronic device 300 in the embodiments of this disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0083] like Figure 3 As shown, the electronic device 300 may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a memory 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processor 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0084] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0085] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from memory 308, or installed from ROM 302. When the computer program is executed by processor 301, it performs the functions defined in the icing prediction method of embodiments of this disclosure.

[0086] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0087] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0088] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0089] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the aforementioned icing prediction method.

[0090] Electronic devices can be programmed with computer program code in one or more programming languages ​​or combinations thereof to perform the operations of this disclosure. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0091] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0092] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0093] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0094] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0095] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0096] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0097] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for predicting icing, characterized in that, include: Acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The model parameters in the baseline mapping model are used as initial values, and the initial values ​​are adjusted using the historical forecast data. The corrected model is then derived based on the adjustment results. Input the data to be predicted into the correction model to obtain temperature and precipitation forecasts; The supercooled liquid water content is calculated based on the temperature and precipitation forecasts. The supercooled liquid water content is then input into a preset icing growth physical model to obtain the icing thickness growth rate. The icing thickness growth rate is then used to predict icing.

2. The method according to claim 1, characterized in that, Acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a baseline mapping model based on the spatial structure information, including: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

3. The method according to claim 2, characterized in that, Using the model parameters in the baseline mapping model as initial values, the initial values ​​are adjusted using the historical forecast data. Based on the adjustment results, a revised model is derived, including: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using preset adjustment hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

4. The method according to claim 1, characterized in that, The data to be predicted is input into the correction model to obtain temperature and precipitation forecasts, including: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

5. The method according to claim 1, characterized in that, The supercooled liquid water content is calculated based on the temperature and precipitation forecasts. This supercooled liquid water content is then input into a preset icing growth physical model to obtain the icing thickness growth rate. The icing thickness growth rate is then used to predict icing, including: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

6. An icing prediction device, the device comprising: The module is used to acquire sub-seasonal icing-related data, perform convolution operations on the sub-seasonal icing-related data to determine the spatial structure information of the weather system, and construct a benchmark mapping model based on the spatial structure information. The sub-seasonal icing-related data includes historical reanalysis data, meteorological station observation data, and historical forecast data from the sub-seasonal to seasonal dynamic models. The adjustment module is used to take the model parameters in the baseline mapping model as initial values, adjust the initial values ​​using the historical forecast data, and derive the corrected model based on the adjustment results. The forecast module is used to input the forecast data to be predicted into the correction model to obtain temperature forecast and precipitation forecast; The calculation module is used to calculate the supercooled liquid water content based on the temperature forecast and precipitation forecast, input the supercooled liquid water content into a preset icing growth physical model, obtain the icing thickness growth rate, and use the icing thickness growth rate to predict icing.

7. The apparatus according to claim 6, characterized in that, The building module is specifically used for: Acquire the sub-seasonal icing-related data, interpolate the historical reanalysis data and historical forecast data in the sub-seasonal icing-related data onto a preset latitude and longitude grid, and match and align them with the meteorological station observation data in time and space. Remove outliers and missing values ​​in the meteorological station observation data to obtain the processing results. Based on the processing results, a connection between the atmospheric circulation field and icing events is established. Based on the processing results and connectivity relationships, an input feature map is determined, and the spatial structure information of the weather system is determined by sliding calculation on the input feature map based on a preset convolutional kernel. Using historical reanalysis data as input data, spatial structure information as core feature representation, and meteorological station observation data as target variables, a benchmark mapping model is constructed.

8. The apparatus according to claim 7, characterized in that, The forecast module is specifically used for: All model parameters that have converged during the training of the baseline mapping model are used as initial values. The input data of the baseline mapping model is switched to historical forecast data. The baseline mapping model corresponding to the initial values ​​is adjusted using feature hyperparameters to obtain the adjusted model. Historical forecast data are adjusted and divided to obtain an adjusted training set and an adjusted validation set. The adjusted training set is input into the adjusted model for forward propagation. The loss function between the predicted output and the actual observation is calculated. The adjusted value is obtained by updating the high-level parameters of the baseline mapping model through the backpropagation algorithm. The adjusted validation set is input into the adjusted model for forward propagation, the validation loss is calculated, the optimal adjusted model is saved by triggering an early stopping mechanism based on the adjusted validation set, and the optimal adjusted model is selected as the correction model.

9. The apparatus according to claim 6, characterized in that, The forecast module is specifically used for: Initial temperature and initial precipitation are determined based on the sub-seasonal to seasonal dynamic model; The data to be predicted is input into the correction model to obtain the corrected temperature and the corrected precipitation function. The temperature forecast is obtained based on the initial temperature and the corrected temperature, and the precipitation forecast is obtained based on the initial precipitation and the corrected precipitation function.

10. The apparatus according to claim 6, characterized in that, The computing module is specifically used for: Relative humidity data is extracted from the sub-seasonal icing-related data. The temperature conditions for the presence of supercooled liquid water are identified based on temperature forecasts. The content of supercooled liquid water is determined based on precipitation forecasts, relative humidity data, and temperature conditions. The freezing efficiency coefficient is calculated based on the ambient temperature in the forecast data to be predicted. The freezing ratio after water droplets collide with the guide wire is determined, and the collision efficiency is determined. The proportion of water droplets that can be captured by the guide wire is determined based on the collision efficiency. The effective capture area is determined based on the proportion of water droplets and the route characteristic parameters. The supercooled liquid water content, along with the freezing efficiency, collision efficiency, and capture area, are input into a preset icing growth model to obtain the initial icing growth rate. Temperature and precipitation forecasts determine the meteorological conditions for the target forecast period corresponding to the forecast data to be predicted, and the ice thickness growth rate is obtained by combining the meteorological conditions with the initial ice growth rate. The ice thickness growth rate is input into a preset ice growth physical model to obtain the growth amount at each time step within the target forecast period. The growth amount is then added to the total thickness to generate a continuous prediction sequence of ice thickness evolution over time, and ice prediction is performed.

11. An electronic device, characterized in that, include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement the icing prediction method as described in any one of claims 1-5.

12. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, causes the processor to implement the icing prediction method provided in any one of claims 1-5.

Images