A Multi-Dimensional Prediction Method and System for Icing Based on Multi-Source Data

Abstract

The application discloses a kind of based on multi-source data's icing multidimensional prediction method and system, comprising: obtaining historical icing data and pre-processing, obtain fusion multi-source data, using improved vole optimization algorithm Attention-LSTM model is optimized, obtain current optimal solution, based on the data obtained above, optimal solution Attention-LSTM model is trained, obtain transmission line icing prediction model, based on historical icing data transmission line icing prediction model is trained, obtain after evaluation transmission line icing prediction model;Real-time icing data of target area is obtained, and real-time icing data is input in the model obtained above and is predicted, and output prediction result generates decision auxiliary information.The application is optimized to Attention-LSTM model by improved vole optimization algorithm, adjusts energy factor decay rate, improves global optimal discovery rate.

Description

Technical Field

[0001] This invention belongs to the field of power transmission line technology, specifically relating to a multi-dimensional prediction method and system for icing based on multi-source data. Background Technology

[0002] In recent years, climate change has become a global environmental issue, extreme weather events have occurred frequently, and icing incidents on power transmission lines have become more and more frequent. Icing on power lines seriously threatens the normal operation of power transmission lines and brings great challenges to line maintenance. Therefore, how to accurately predict the thickness of icing on power lines has become an urgent problem to be solved.

[0003] Currently, traditional icing prediction methods are mainly based on icing data, utilizing statistical theory and intelligent computing technology to analyze meteorological, environmental, and icing data to establish a fitting model with reliable accuracy. With the upgrading and application of icing monitoring technology and equipment, more and more micrometeorological data and icing monitoring data are being used in the construction of icing prediction models.

[0004] While traditional methods can predict icing thickness to some extent, they suffer from several significant drawbacks: First, traditional icing prediction methods rely heavily on historical icing data, making them ineffective in predicting climate change and extreme weather events. Second, micrometeorological data is often limited in its availability, failing to cover all power lines, and its quality cannot be guaranteed, resulting in insufficient model accuracy. Finally, the icing process of transmission lines is complex, influenced by numerous factors, and these factors involve nonlinear and complex multidimensional relationships, demanding sophisticated algorithms that are difficult to implement using traditional machine learning methods. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-dimensional prediction method and system for icing based on multi-source data, aiming to solve the problems in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional prediction method for icing based on multi-source data, comprising:

[0007] Step S1: Obtain historical icing data for the target area;

[0008] Step S2: Preprocess the acquired historical icing data to obtain fused multi-source data;

[0009] Step S3: Improve the lemming optimization algorithm by introducing an elite archiving mechanism and an energy factor. Use the improved lemming optimization algorithm to optimize the Attention-LSTM model and obtain the current optimal solution.

[0010] Step S4: Train the Attention-LSTM model using fused multi-source data and the current optimal solution to obtain the transmission line icing prediction model; divide the historical icing data into training and testing sets, train and test the transmission line icing prediction model to obtain the evaluated transmission line icing prediction model.

[0011] Step S5: Obtain real-time icing data for the target area, input the real-time icing data into the evaluated transmission line icing prediction model for prediction, and output the prediction results of line icing thickness, icing type, risk level and melting time for the target area.

[0012] Step S6: Based on the predicted results of the line icing thickness, icing type, risk level, and melting time, generate decision support information.

[0013] Furthermore, historical icing data for the target area includes micrometeorological data, weather station data, and high-resolution meteorological grid data;

[0014] The meteorological station data comes from meteorological stations near the icing towers: the observation equipment at the meteorological stations records hourly values ​​of temperature, humidity, wind speed, and rainfall during the icing period; the meteorological station data is based on the hourly observation values ​​recorded during the icing period.

[0015] Preprocessing of the observations yields corresponding high-resolution meteorological grid data;

[0016] Micrometeorological data consists of hourly temperature and humidity data collected by a micrometeorological data acquisition device during the icing period of the icing tower.

[0017] Furthermore, the specific process of fusing multi-source data is as follows:

[0018] Calculate the distances between the icing-covered tower and the corresponding grid centers of the micro-meteorological data acquisition device, meteorological station, and high-resolution meteorological grid data, denoted as d1, d2, and d3, respectively; use the reciprocals of the squares of the distances d1, d2, and d3 as spatial distance weighting coefficients. ,express:

[0019] (1);

[0020] In the formula, d represents the distance; i is an index variable; i = 1, 2, 3; This refers to the distance from the icing-covered tower to the micrometeorological data acquisition device. This refers to the distance from the icing tower to the weather station. This represents the distance from the icing tower to the corresponding grid center in the high-resolution meteorological grid data.

[0021] A sliding window-based source data anomaly detection algorithm is used to dynamically monitor the reliability scores of micro-meteorological data, weather station data, and high-resolution meteorological gridded data. Real-time output of reliability weighting coefficients ,express:

[0022] (2);

[0023] Among them, the source data anomaly detection algorithm based on a sliding window achieves reliability scoring by dynamically monitoring multiple quality dimensions. Multiple quality dimensions include data completeness, data mutation frequency, and variance stability.

[0024] Data integrity rate: The ratio of the number of micro-meteorological data, weather station data and high-resolution meteorological grid data actually received within the sliding time window to the expected total number of micro-meteorological data, weather station data and high-resolution meteorological grid data points, calculated using a source data anomaly detection algorithm with a sliding window.

[0025] The closer the ratio of the actual number of points received to the expected total number of points is to 1, the higher the integrity of the micro-meteorological data, weather station data, and high-resolution meteorological grid data.

[0026] Data mutation frequency: Using a source data anomaly detection algorithm within a sliding window, the absolute difference between two adjacent time points in micro-meteorological data, weather station data, and high-resolution meteorological grid data within the statistical window is calculated; the absolute difference is compared with a preset threshold. The higher the absolute difference exceeds the preset threshold, the lower the reliability of the micro-meteorological data, weather station data, and high-resolution meteorological grid data.

[0027] Variance stability: The variance of micro-meteorological data, weather station data and high-resolution meteorological grid data within the sliding window is calculated using the source data anomaly detection algorithm of the sliding window, and compared with the preset long-term benchmark variance. The closer the ratio of the calculated variance to the long-term benchmark variance is to 1, the more stable the fluctuation of the micro-meteorological data, weather station data and high-resolution meteorological grid data is.

[0028] Taking into account data completeness, frequency of data mutations, and variance stability, a weighted geometric mean or weighted arithmetic mean is used to calculate the overall reliability score of micrometeorological data, meteorological station data, and high-resolution meteorological gridded data within a time window. ;

[0029] Comprehensive spatial distance weighting coefficient With reliability weighting coefficient The weights ω for obtaining micro-meteorological data, weather station data, and high-resolution meteorological grid data are:

[0030] (3);

[0031] In the formula, for Adjustable hyperparameters; for Adjustable hyperparameters;

[0032] The weight ω is calculated as a weighted average of micro-meteorological data, meteorological station data, and high-resolution meteorological grid data to obtain fused multi-source data.

[0033] Furthermore, the specific process of the improved lemming optimization algorithm is as follows:

[0034] The lemming optimization algorithm consists of six stages: initialization stage, long-distance migration stage, burrowing stage, foraging stage, and predator avoidance stage.

[0035] The initialization phase of the lemming optimization algorithm generates a random population in the search space as the initial population.

[0036] The initial positions of the initial population are initialized using Latin hypercube sampling, which means:

[0037] (4);

[0038] In the formula, Let i be the initial position of the i-th lemming in the j-th dimension; This is the lower bound of the j-th dimension; The random number generated for the i-th lemming in dimension j using Latin hypercube sampling ensures a uniform distribution of the initial population. This is the upper bound of the j-th dimension;

[0039] The long-distance migration stage indicates:

[0040] (5);

[0041] In the formula, Let be the position of the i-th lemming after a long migration in the (t+1)-th iteration; Let be the position of the optimal lemming in the initial population at the t-th iteration; A marker to change the search direction; is a random number vector representing the probability density function of Brownian motion; The random numbers are uniformly generated within the interval [0,1] to balance the position of the optimal lemming in the initial population; Let i be the current position of the i-th lemming at the t-th iteration; Let be the position of the a-th lemming randomly selected from the initial population during the t-th iteration;

[0042] The indicator that changes the search direction means:

[0043] (6);

[0044] In the formula, The random number is uniformly generated within the interval [0,1]. By introducing the random number, the probability of switching the search direction is determined.

[0045] The random number vector of the probability density function of Brownian motion is represented as:

[0046] (7);

[0047] In the formula, It is a natural exponential function; It is a vector of random numbers; For the exponential term of the Gaussian distribution;

[0048] The digging stage indicates:

[0049] (8);

[0050] In the formula, It is a dynamically adjustable parameter that controls the step size or exploration range of the lemming's digging behavior; Let b be the position of the b-th lemming randomly selected from the initial population during the t-th iteration;

[0051] Dynamically adjusted parameters, meaning:

[0052] (9);

[0053] In the formula, It is a sine function; For the current time step Half of;

[0054] During the foraging stage, it indicates:

[0055] (10);

[0056] In the formula, The parameters for the spiral winding mechanism are used to simulate the spiral movement path of lemmings within their foraging area.

[0057] The parameters of the helical winding mechanism are represented as follows:

[0058] (11);

[0059] (12);

[0060] In the formula, The radius of the spiral; It is a cosine function; For the dimensions of the search space; Let be the coordinate component of the optimal lemming in the initial population in the j-th dimension at the t-th iteration, where yes Coordinate components in the j-dimensional dimension; Let be the coordinate component of the i-th lemming in the j-th dimension at the t-th iteration;

[0061] During the phase of avoiding predators, it means:

[0062] (13);

[0063] (14);

[0064] In the formula, These are dynamically adjustable parameters that control the intensity or step length of lemming avoidance behavior. For the Lévy flight random number vector; This represents the maximum number of iterations.

[0065] Levy's flight random number vector, representing:

[0066] (15);

[0067] In the formula, The random number is a normally distributed number located at the numerator position; Let Lévy's distribution be the standard deviation. The random number is a normally distributed number and is located in the denominator. Γ is the Lévy exponent; Γ is the gamma function;

[0068] Introducing an elite save system:

[0069] An elite archive is established to store the locations of the best lemmings discovered so far. After each iteration, the fitness values ​​of the lemmings in the initial population are calculated according to a predefined fitness function, and the lemmings in the initial population are sorted according to their fitness values. Then, the locations of the top k best lemmings after sorting are selected. The locations of the best lemmings are compared with the locations of the best lemmings stored in the elite archive, and the location of the best lemming with the highest similarity value is selected as the elite lemming. ;

[0070] Will Joining the long-distance migration phase means:

[0071] (16);

[0072] In the formula, As a social learning factor; It generates random numbers within the range [0,1], and uses these random numbers to adjust the elite lemmings. The extent of the impact on the current lemming location update;

[0073] Will Adding the digging stage means:

[0074] (17);

[0075] Will Joining the foraging phase indicates:

[0076] (18);

[0077] Will Adding a phase of avoiding predators means:

[0078] (19);

[0079] Let energy factor E be the energy factor. When energy factor E changes, lemmings will enter different stages depending on the magnitude of energy factor E.

[0080] Energy Factor The calculation, expressed as:

[0081] (20);

[0082] In the formula, The energy factor at the t-th iteration; This is the minimum value of the energy factor; This represents the maximum value of the energy factor. represents the initial population diversity index at the t-th iteration;

[0083] The initial population diversity index at the t-th iteration is represented as:

[0084] (twenty one);

[0085] In the formula, This represents the initial population size, i.e., the number of lemmings in the initial population. For the dimensions of the search space; Let be the average position coordinate of the initial population in the j-th dimension at the t-th iteration; Let this be the initial position of the i-th lemming in the j-th dimension during the t-th iteration;

[0086] The energy factor at iteration number t is calculated according to equation (20). Determine the energy factor at iteration number t. If the set energy threshold is reached, the exploration phase begins, which includes a long-distance migration phase and a burrowing phase. If the set energy threshold is not reached, the foraging phase begins. Simultaneously, different methods are used in different phases. Determine the formula for updating lemming positions;

[0087] During the exploration phase, when If the value is less than 0.3, the lemming is considered to be in the long-distance migration stage, and the lemming's position is updated using equation (16); when If the value is greater than 0.3, the lemming is considered to be in the burrowing stage, and the position of the lemming is updated using equation (17);

[0088] when If the value is less than 0.5, the lemming is considered to be in the foraging stage, and its position is updated using equation (18); when If the value is greater than 0.5, the lemming is considered to be in the stage of avoiding predators, and the position of the lemming is updated using formula (19);

[0089] After completing the exploration and foraging phases and updating the lemming positions, the fitness values ​​of the new lemming positions updated by Equations (16), (17), (18) and (19) are evaluated. The fitness values ​​of the lemmings updated by Equations (16), (17), (18) and (19) are compared, and the optimal lemming is identified as the current optimal solution.

[0090] Increment the iteration number t by 1, that is ; This represents the current iteration number;

[0091] judge Has the maximum number of iterations been reached? ,when Maximum number of iterations not reached Return to formulas (20) and (21) to recalculate the energy factor. ;when Reaching the maximum number of iterations The optimization process is complete.

[0092] Furthermore, the specific process for obtaining the evaluated transmission line icing prediction model is as follows:

[0093] Historical icing data was divided into training and testing sets. The training set was used to train the transmission line icing prediction model, and the testing set was used to verify the accuracy of the trained transmission line icing prediction model.

[0094] The coefficient of determination R² and the minimum mean square error MSE were selected as performance indicators to evaluate the accuracy performance of the trained transmission line icing prediction model, and the evaluated transmission line icing prediction model was obtained.

[0095] Furthermore, the specific process for generating decision support information is as follows:

[0096] Acquire real-time icing data of the target area, and preprocess the real-time icing data in step S2 to obtain real-time fused multi-source data;

[0097] The real-time fusion of multi-source data from the target area is input into the evaluated transmission line icing prediction model to predict the icing situation of the lines in the target area, and output the prediction results of the line icing thickness, icing type, risk level and melting time in the target area.

[0098] A decision rule base is constructed, and the predicted results of line icing thickness, icing type, risk level, and melting time in the target area are input into the decision rule base to generate decision support information.

[0099] A multi-dimensional icing prediction system based on multi-source data, used to implement a multi-dimensional icing prediction method based on multi-source data, includes:

[0100] The data acquisition module is used to acquire historical icing data for the target area;

[0101] The data processing module is used to preprocess the acquired historical icing data to obtain fused multi-source data;

[0102] The model optimization module is used to improve the lemming optimization algorithm by introducing an elite archiving mechanism and an energy factor. The improved lemming optimization algorithm is then used to optimize the Attention-LSTM model to obtain the current optimal solution.

[0103] The training module is used to train the Attention-LSTM model using fused multi-source data and the current optimal solution to obtain the transmission line icing prediction model; the historical icing data is divided into training set and test set to train and test the transmission line icing prediction model to obtain the evaluated transmission line icing prediction model.

[0104] The prediction module is used to acquire real-time icing data of the target area, input the real-time icing data into the evaluated transmission line icing prediction model for prediction, and output the prediction results of line icing thickness, icing type, risk level and melting time in the target area.

[0105] The generation module is used to generate decision support information based on the predicted results of the line icing thickness, icing type, risk level, and melting time.

[0106] A computer device includes: one or more processors; the processors are configured to store one or more programs; when the one or more programs are executed by the one or more processors, they are configured to implement a multi-dimensional icing prediction method based on multi-source data.

[0107] A computer-readable storage medium includes a computer program stored thereon, which, when executed, is used to implement a multi-dimensional icing prediction method based on multi-source data.

[0108] Compared with existing technologies, the present invention has the following advantages:

[0109] (1) By integrating micro-meteorological data, meteorological station data and high-resolution meteorological grid data, and by combining spatial distance weight and reliability weight, this invention solves the problems of insufficient single data sources and inconsistent data quality in traditional methods, reduces the dependence on historical icing data, and thus significantly improves the accuracy of icing thickness prediction.

[0110] (2) The improved lemming optimization algorithm proposed in this invention effectively solves the problems of traditional lemming optimization algorithms being prone to getting trapped in local optima and having slow convergence speed by introducing innovative methods such as Latin hypercube sampling to initialize the population, an elite archiving mechanism to save the position of the best lemming, and dynamic adjustment of the energy factor according to the diversity of the initial population. It enhances the global optimization ability, accelerates the convergence speed, provides better initial parameters for the Attention-LSTM model, and significantly improves the accuracy and stability of the evaluated transmission line icing prediction model.

[0111] (3) This invention breaks through the limitation of traditional icing prediction being limited to thickness prediction, and realizes multi-dimensional prediction of icing thickness, icing type, risk level and melting time. It also generates decision support information by combining expert knowledge rule base, providing comprehensive icing situation awareness and accurate decision support for the operation and maintenance management of transmission lines, and directly reducing line maintenance costs and risks. Attached Figure Description

[0112] Figure 1 This is a flowchart of the method of the present invention.

[0113] Figure 2 This is a comparison chart of the initial population diversity indices of this invention.

[0114] Figure 3 This is a comparison chart of the fitness values ​​of the present invention. Detailed Implementation

[0115] like Figure 1 As shown, the present invention provides a technical solution: a multi-dimensional prediction method for icing based on multi-source data, comprising:

[0116] Step S1: Obtain historical icing data for the target area;

[0117] Step S2: Preprocess the acquired historical icing data to obtain fused multi-source data;

[0118] Step S3: Improve the lemming optimization algorithm by introducing an elite archiving mechanism and an energy factor. Use the improved lemming optimization algorithm to optimize the Attention-LSTM model and obtain the current optimal solution.

[0119] Step S4: Train the Attention-LSTM model using fused multi-source data and the current optimal solution to obtain the transmission line icing prediction model; divide the historical icing data into training and testing sets, train and test the transmission line icing prediction model to obtain the evaluated transmission line icing prediction model.

[0120] Step S5: Obtain real-time icing data for the target area, input the real-time icing data into the evaluated transmission line icing prediction model for prediction, and output the prediction results of line icing thickness, icing type, risk level and melting time for the target area.

[0121] Step S6: Based on the predicted results of the line icing thickness, icing type, risk level, and melting time, generate decision support information.

[0122] Historical icing data for the target area includes micrometeorological data, meteorological station data, and high-resolution meteorological grid data;

[0123] The meteorological station data comes from meteorological stations near the icing towers: the observation equipment of the meteorological stations (including thermometers, hygrometers, anemometers, rain gauges, etc.) records the hourly temperature, humidity, wind speed and rainfall during the icing period; the meteorological station data is based on the hourly observations recorded during the icing period.

[0124] Preprocessing the observations yields high-resolution meteorological grid data for the corresponding 1km×1km grid points of the ice-covered towers;

[0125] Micrometeorological data consists of hourly temperature and humidity data collected by a micrometeorological data acquisition device during the icing period of the icing tower.

[0126] The specific process for obtaining fused multi-source data is as follows:

[0127] Calculate the distances between the icing-covered tower and the corresponding grid centers of the micro-meteorological data acquisition device, meteorological station, and high-resolution meteorological grid data, denoted as d1, d2, and d3, respectively; use the reciprocals of the squares of the distances d1, d2, and d3 as spatial distance weighting coefficients. ,express:

[0128] (1);

[0129] In the formula, d represents the distance; i is an index variable; i = 1, 2, 3; This refers to the distance from the icing-covered tower to the micrometeorological data acquisition device. This refers to the distance from the icing tower to the weather station. This represents the distance from the icing tower to the corresponding grid center in the high-resolution meteorological grid data.

[0130] A sliding window-based source data anomaly detection algorithm is used to dynamically monitor the reliability scores of micro-meteorological data, weather station data, and high-resolution meteorological gridded data. Real-time output of reliability weighting coefficients ,express:

[0131] (2);

[0132] Among them, the source data anomaly detection algorithm based on a sliding window achieves reliability scoring by dynamically monitoring multiple quality dimensions. Multiple quality dimensions include data completeness, data mutation frequency, and variance stability.

[0133] Data integrity rate: The ratio of the number of micro-meteorological data, weather station data and high-resolution meteorological grid data actually received within the sliding time window to the expected total number of micro-meteorological data, weather station data and high-resolution meteorological grid data points, calculated using a source data anomaly detection algorithm with a sliding window.

[0134] The closer the ratio of the actual number of points received to the expected total number of points is to 1, the higher the integrity of the micro-meteorological data, weather station data, and high-resolution meteorological grid data.

[0135] Data mutation frequency: Using a source data anomaly detection algorithm within a sliding window, the absolute difference between two adjacent time points in micro-meteorological data, weather station data, and high-resolution meteorological grid data within the statistical window is calculated; the absolute difference is compared with a preset threshold. The higher the absolute difference exceeds the preset threshold, the lower the reliability of the micro-meteorological data, weather station data, and high-resolution meteorological grid data.

[0136] Variance stability: The variance of micro-meteorological data, weather station data and high-resolution meteorological grid data within the sliding window is calculated using the source data anomaly detection algorithm of the sliding window, and compared with the preset long-term benchmark variance (such as the average variance of the past 30 days). The closer the ratio of the calculated variance to the long-term benchmark variance is to 1, the more stable the fluctuation of the micro-meteorological data, weather station data and high-resolution meteorological grid data is.

[0137] Taking into account data completeness, frequency of data mutations, and variance stability, a weighted geometric mean or weighted arithmetic mean is used to calculate the overall reliability score of micrometeorological data, meteorological station data, and high-resolution meteorological gridded data within a time window. ;

[0138] Comprehensive spatial distance weighting coefficient With reliability weighting coefficient The weights ω for obtaining micro-meteorological data, weather station data, and high-resolution meteorological grid data are:

[0139] (3);

[0140] In the formula, for Adjustable hyperparameters; for Adjustable hyperparameters;

[0141] The weighted average of micro-meteorological data, meteorological station data, and high-resolution meteorological grid data is calculated based on the obtained weight ω to obtain the fused multi-source data.

[0142] The specific process of the improved lemming optimization algorithm is as follows:

[0143] The traditional lemming optimization algorithm mainly includes six stages: initialization stage, long-distance migration (exploration) stage, burrowing (exploration) stage, foraging (development) stage, avoiding predators (development) stage, and transition from exploration to development.

[0144] The traditional lemming optimization algorithm generates a random population in the search space as the initial population during the initialization phase, thus starting the optimization process.

[0145] Random initialization during the initialization phase may lead to uneven distribution of the initial population, affecting convergence efficiency. This invention uses Latin hypercube sampling to initialize the initial position of the initial population, which can significantly improve the performance of the lemming optimization algorithm. (Example:)

[0146] (4);

[0147] In the formula, Let i be the initial position of the i-th lemming in the j-th dimension; This is the lower bound of the j-th dimension; The random number generated for the i-th lemming in dimension j using Latin hypercube sampling ensures a uniform distribution of the initial population. This is the upper bound of the j-th dimension;

[0148] The long-distance migration stage refers to the random long-distance migration of lemmings when their numbers become too large and food becomes scarce. Lemmings will explore the search space based on their current location and the locations of random lemmings in the initial population, looking for habitats rich in food resources, thereby obtaining better living conditions and resources. It is also worth noting that the direction and distance of lemming migration are not fixed and are affected by a variety of factors such as the ecological environment.

[0149] The long-distance migration stage indicates:

[0150] (5);

[0151] In the formula, Let be the position of the i-th lemming after a long migration in the (t+1)-th iteration; Let be the position of the optimal lemming in the initial population at the t-th iteration; A marker to change the search direction; is a random number vector representing the probability density function of Brownian motion; The random numbers are uniformly generated within the interval [0,1] to balance the position of the optimal lemming in the initial population; Let i be the current position of the i-th lemming at the t-th iteration; Let be the position of the a-th lemming randomly selected from the initial population during the t-th iteration;

[0152] The indicator that changes the search direction means:

[0153] (6);

[0154] In the formula, The random number is uniformly generated within the interval [0,1]. By introducing the random number, the probability of switching the search direction is determined.

[0155] The random number vector of the probability density function of Brownian motion is represented as:

[0156] (7);

[0157] In the formula, It is a natural exponential function; It is a vector of random numbers; For the exponential term of the Gaussian distribution;

[0158] The burrowing stage refers to the process by which lemmings dig burrows in their habitat, forming complex tunnels that provide them with safe refuges and places to store food. Lemmings will randomly dig new burrows based on the current location of the existing burrow and the locations of random lemmings in the initial population. This design helps lemmings quickly escape the threat of predators and find food more efficiently.

[0159] The digging stage indicates:

[0160] (8);

[0161] In the formula, It is a dynamically adjustable parameter used to control the step size or exploration range of lemming digging behavior; Let b be the position of the b-th lemming randomly selected from the initial population during the t-th iteration;

[0162] Dynamically adjusted parameters, meaning:

[0163] (9);

[0164] In the formula, It is a sine function; For the current time step Half of;

[0165] The foraging phase refers to the lemming's extensive and random movement within its burrows, relying on its keen sense of smell and hearing to locate food sources. Lemmings typically establish a relatively small foraging area within their habitat, depending on the abundance and availability of food. To ingest as much food as possible, lemmings will wander freely within this foraging area. To simulate the lemming's random foraging behavior, a spiral entanglement mechanism is used to represent the foraging phase, as shown below:

[0166] (10);

[0167] In the formula, The parameters are for the spiral winding mechanism, used to simulate the spiral movement path of lemmings in the foraging area;

[0168] The parameters of the helical winding mechanism are represented as follows:

[0169] (11);

[0170] (12);

[0171] In the formula, The radius of the spiral; It is a cosine function; For the dimensions of the search space; Let be the coordinate component of the optimal lemming in the initial population in the j-th dimension at the t-th iteration, where yes Coordinate components in the j-dimensional dimension; Let be the coordinate component of the i-th lemming in the j-th dimension at the t-th iteration;

[0172] The predator avoidance stage refers to the lemming's avoidance and protective behavior when encountering danger; burrows in its habitat serve as refuges for lemmings; upon spotting an enemy, lemmings will use their special running abilities to escape back to their burrows; simultaneously, lemmings will also make deceptive movements to evade predators; the predator avoidance stage indicates:

[0173] (13);

[0174] (14);

[0175] In the formula, These are dynamically adjustable parameters used to control the intensity or stride length of lemming avoidance behavior; For the Lévy flight random number vector; This represents the maximum number of iterations.

[0176] Levy's flight random number vector, representing:

[0177] (15);

[0178] In the formula, The random number is a normally distributed number located at the numerator position; Let Lévy's distribution be the standard deviation. The random number is a normally distributed number and is located in the denominator. Γ is the Lévy exponent; Γ is the gamma function;

[0179] Introducing an elite save system:

[0180] An elite archive is established to store the locations of the best lemmings discovered so far. After each iteration, the fitness values ​​of the lemmings in the initial population are calculated according to a predefined fitness function, and the lemmings in the initial population are sorted according to their fitness values. Then, the locations of the top k best lemmings after sorting are selected. The locations of the best lemmings are compared with the locations of the best lemmings stored in the elite archive, and the location of the best lemming with the highest similarity value is selected as the elite lemming. ;

[0181] Will Joining the long-distance migration phase means:

[0182] (16);

[0183] As a social learning factor; It generates random numbers within the range [0,1], and uses these random numbers to adjust the elite lemmings. The extent of the impact on the current lemming position update; Adding the digging stage means:

[0184] (17);

[0185] Will Joining the foraging phase indicates:

[0186] (18);

[0187] Will Adding a phase of avoiding predators means:

[0188] (19);

[0189] To ensure that lemmings iterate between different stages, an energy factor E is designed. When the energy factor changes, the lemmings will enter different stages according to the magnitude of the energy factor E.

[0190] When lemmings have sufficient energy factor E, they selectively migrate or burrow; otherwise, they forage for food and avoid predators. Traditional lemming optimization algorithms use a linearly decreasing energy factor E, which cannot adapt to nonlinear optimization requirements and does not consider the actual search state of the initial population, easily leading to premature convergence or overexploration. Therefore, this invention proposes an adaptive energy factor E calculation method, dynamically adjusting the decay rate of energy factor E based on the real-time state of the initial population (initial population diversity). When the initial population diversity is high, the decay of energy factor E is slowed down, extending the exploration time; when the initial population converges too quickly, the energy factor E is increased, restarting exploration; when the initial population stagnates, the decay of energy factor E is accelerated, transitioning to the foraging phase. The adaptive energy factor E calculation method proposed in this invention can significantly improve the convergence speed and the global optimum discovery rate, as shown below:

[0191] (20);

[0192] In the formula, The energy factor at the t-th iteration; This is the minimum value of the energy factor; This represents the maximum value of the energy factor. represents the initial population diversity index at the t-th iteration;

[0193] The initial population diversity index at the t-th iteration is represented as:

[0194] (twenty one);

[0195] In the formula, This represents the initial population size, i.e., the number of lemmings in the initial population. For the dimensions of the search space; Let be the average position coordinate of the initial population in the j-th dimension at the t-th iteration; Let this be the initial position of the i-th lemming in the j-th dimension during the t-th iteration;

[0196] The energy factor at iteration number t is calculated according to equation (20). Determine the energy factor at iteration number t. If the set energy threshold is reached, the exploration phase begins, which includes a long-distance migration phase and a burrowing phase. If the set energy threshold is not reached, the foraging phase begins. Simultaneously, different methods are used in different phases. Determine the formula for updating lemming positions;

[0197] During the exploration phase, when If the value is less than 0.3, the lemming is considered to be in the long-distance migration stage, and the lemming's position is updated using equation (16); when If the value is greater than 0.3, the lemming is considered to be in the burrowing stage, and the position of the lemming is updated using equation (17);

[0198] when If the value is less than 0.5, the lemming is considered to be in the foraging stage, and its position is updated using equation (18); when If the value is greater than 0.5, the lemming is considered to be in the stage of avoiding predators, and the position of the lemming is updated using formula (19);

[0199] After completing the exploration and foraging phases and updating the lemming positions, the fitness values ​​of the new lemming positions updated by Equations (16), (17), (18) and (19) are evaluated. The fitness values ​​of the lemmings updated by Equations (16), (17), (18) and (19) are compared, and the optimal lemming is identified as the current optimal solution.

[0200] Increment the iteration number t by 1, that is ; This represents the current iteration number;

[0201] judge Has the maximum number of iterations been reached? ,when Maximum number of iterations not reached Return to formulas (20) and (21) to recalculate the energy factor. ;when Reaching the maximum number of iterations The optimization process is complete.

[0202] The historical icing data was divided into a training set and a test set. The training set was used to train the transmission line icing prediction model, and the test set was used to verify the accuracy of the trained transmission line icing prediction model.

[0203] The coefficient of determination R² and the minimum mean square error MSE were selected as performance indicators to evaluate the accuracy performance of the trained transmission line icing prediction model, and the evaluated transmission line icing prediction model was obtained.

[0204] Among them, an attention mechanism is introduced on the basis of the Long Short-Term Memory (LSTM) network to construct the Attention-LSTM model;

[0205] Micro-meteorological data, weather station data, and high-resolution meteorological grid data are input into the Attention-LSTM model. The micro-meteorological data, weather station data, and high-resolution meteorological grid data are processed through a long short-term memory network to obtain a hidden state sequence. The attention calculation layer is used to calculate the attention weights of the hidden state sequence. The attention weights of the hidden states are weighted and summed to generate a fixed context vector. The context vector is then passed through a fully connected output layer to obtain the corresponding output result.

[0206] Specifically, the improved lemming optimization algorithm is used to optimize the learning rate, number of hidden layer units, attention weight matrix, and context vector weights of the Attention-LSTM model to obtain the current optimal solution. The specific process is as follows:

[0207] Learning rate: The learning rate is a hyperparameter in the improved lemming optimization algorithm, used to control the step size of the Attention-LSTM model parameter updates. If the learning rate is too large, the hyperparameter update will be too large, causing oscillations; if the learning rate is too small, the hyperparameter update will be too small, and the convergence speed will become slow.

[0208] Number of hidden layer units: The number of hidden layer units is a hyperparameter of the Attention-LSTM model, representing the feature dimension of the hidden state in each layer of the Long Short-Term Memory network. It determines the capacity of the Attention-LSTM model. Too few hidden layer units will lead to underfitting, while too many units will easily lead to overfitting and high computational cost.

[0209] Attention weight matrix: The attention weight matrix is ​​a hyperparameter of the Attention-LSTM model, used to map the hidden state to attention weights. Optimizing the attention weight matrix enables the Attention-LSTM model to more accurately evaluate the importance of micro-meteorological data, weather station data, and high-resolution meteorological grid data at different historical moments.

[0210] Context vector weights: Context vector weights are also hyperparameters of the Attention-LSTM model, used to calculate the context vector. Optimizing the context vector weights enables the Attention-LSTM model to more accurately assess the importance of micro-meteorological data, weather station data, and high-resolution meteorological grid data at different historical moments.

[0211] By optimizing hyperparameters such as learning rate, number of hidden layer units, attention weight matrix, and context vector weights using the improved lemming optimization algorithm, the Attention-LSTM model's prediction performance parameters are optimized, which is the current optimal solution.

[0212] The specific process for generating decision support information is as follows:

[0213] Staff retrieved real-time icing data of the target area from different sources for the past or the next 12 hours, and preprocessed the real-time icing data in step S2 to obtain real-time fused multi-source data.

[0214] The real-time fusion of multi-source data from the target area is input into the evaluated transmission line icing prediction model to predict the icing situation of the lines in the target area. Based on the corresponding output results, further analysis is performed, and the predicted results of line icing thickness, icing type, risk level, and melting time in the target area are output.

[0215] Line icing thickness: The continuous line icing thickness values ​​are output in a regression manner, and the prediction results are calculated by the evaluated transmission line icing prediction model based on real-time fused multi-source data.

[0216] Icing type: Outputs the probability distribution of major icing types such as rime, hoarfrost, and mixed rime in a classification manner; represents the probability assessment of different types of icing by the transmission line icing prediction model after evaluation;

[0217] Risk Level: Outputs a preset risk level in a classification manner (e.g., Level I: Mild; Level II: Moderate; Level III: Severe; Level IV: Extreme).

[0218] Melting time: Outputs the predicted melting time in a regression manner; the predicted melting time is based on forecasts of at least 24 hours of gridded meteorological data (temperature, wind speed, solar radiation);

[0219] A decision rule base is constructed, which encapsulates industry knowledge and expert experience in the form of "if-then" rules. For example, if the risk level is Level III (severe) and the icing type is rime ice, and the line icing thickness is high, then the warning level is orange warning, and the decision support information suggests starting the DC de-icing device 4 hours before the predicted de-icing time.

[0220] The predicted results of line icing thickness, icing type, risk level, and de-icing time are input into the decision rule base, which automatically matches and generates decision support information. The decision support information is presented visually through a human-computer interaction interface or automatically sent to the mobile terminal of operation and maintenance personnel and the control system of the de-icing device through a communication interface.

[0221] A system for predicting icing thickness based on multi-source data, used to implement the aforementioned multi-dimensional icing prediction method based on multi-source data, includes:

[0222] The data acquisition module is used to acquire historical icing data for the target area;

[0223] The data processing module is used to preprocess the acquired historical icing data to obtain fused multi-source data;

[0224] The model optimization module is used to improve the lemming optimization algorithm by introducing an elite archiving mechanism and an energy factor. The improved lemming optimization algorithm is then used to optimize the Attention-LSTM model to obtain the current optimal solution.

[0225] The training module is used to train the Attention-LSTM model using fused multi-source data and the current optimal solution to obtain the transmission line icing prediction model; the historical icing data is divided into training set and test set to train and test the transmission line icing prediction model to obtain the evaluated transmission line icing prediction model.

[0226] The prediction module is used to acquire real-time icing data of the target area, input the real-time icing data into the evaluated transmission line icing prediction model for prediction, and output the prediction results of line icing thickness, icing type, risk level and melting time in the target area.

[0227] The generation module is used to generate decision support information based on the predicted results of the line icing thickness, icing type, risk level, and melting time.

[0228] A computer device includes: one or more processors; the processors are configured to store one or more programs; when the one or more programs are executed by the one or more processors, a multi-dimensional prediction method for icing based on multi-source data is implemented.

[0229] A computer-readable storage medium having a computer program stored thereon, which, when executed, implements a multi-dimensional prediction method for icing based on multi-source data.

[0230] Figure 2The study demonstrates the changing trend of initial population diversity before and after optimization of the lemming algorithm. The original lemming algorithm initializes the population in a random manner, which easily leads to initial population aggregation and insufficient diversity, causing the Attention-LSTM model to get stuck in a local optimum.

[0231] The improved lemming optimization algorithm uses Latin hypercube sampling in the initialization phase, which significantly improves the uniformity and diversity of the initial population.

[0232] from Figure 2 As can be seen, in the early stages, the initial population diversity index of both the original lemming algorithm and the improved lemming optimization algorithm was at a high level. However, as the number of iterations increased, the initial population diversity index of the original lemming algorithm dropped rapidly, falling to 0.3 after 10 iterations and almost to 0 after 50 iterations.

[0233] The improved lemming optimization algorithm exhibits a more gradual decline in initial population diversity, with the number of iterations remaining above 0.6 after 10 iterations and around 0.3 after 50 iterations, showing a more gradual decline. This indicates that the improved lemming optimization algorithm can better maintain initial population diversity, enhance global search capabilities, and effectively avoid local optima.

[0234] Figure 3 The comparison of fitness curves is shown; the original lemming algorithm does not introduce an elite archiving mechanism, which makes the original lemming algorithm converge faster in terms of the number of iterations. The number of iterations tends to stabilize around the 25th iteration. The fitness value of the original lemming algorithm is larger, indicating that the Attention-LSTM model is prone to getting trapped in local optima.

[0235] After introducing an elite archiving mechanism, the improved lemming optimization algorithm, although it converges slightly slower in terms of the number of iterations and reaches stability in the 39th iteration, ultimately reduces the fitness value of the improved lemming optimization algorithm to 0.46, demonstrating a stronger global optimization capability.

[0236] In summary, the improved lemming optimization algorithm not only maintains good convergence efficiency, but also significantly improves the search performance for the optimal solution.

[0237] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional icing prediction method based on multi-source data, characterized in that, Comprise: Step S1: obtaining historical icing data of a target area; Step S2: preprocessing the obtained historical icing data to obtain fused multi-source data; Using a sliding window-based source data anomaly detection algorithm, the reliability scores of micro-meteorological data, meteorological station data and high-resolution meteorological grid point data are dynamically monitored , and the reliability weight coefficients are output in real time , indicating that （2）； The source data anomaly detection algorithm based on a sliding window is realized by dynamically monitoring multiple quality dimensions to obtain a reliability score The multiple quality dimensions include data integrity, data mutation frequency, and variance stability Data completeness rate: through the source data anomaly detection algorithm of the sliding window, the ratio of the actual received micro-meteorological data, meteorological station data and high-resolution meteorological grid data points in the sliding time window to the expected total number of micro-meteorological data, meteorological station data and high-resolution meteorological grid data points is calculated; The closer the ratio of the actual received points to the expected total number of points is to 1, the higher the integrity of the micro-meteorological data, meteorological station data and high-resolution meteorological grid data is; Data mutation frequency: through the source data anomaly detection algorithm of the sliding window, the absolute difference value of the adjacent two time point data in the statistical window is calculated; compared with the preset threshold value, the higher the value of the absolute difference value exceeding the preset threshold value, the lower the reliability of the micro-meteorological data, meteorological station data and high-resolution meteorological grid data; Variance stability: through the source data anomaly detection algorithm of the sliding window, the variance of the micro-meteorological data, meteorological station data and high-resolution meteorological grid data value in the sliding window is calculated, and compared with the preset long-term reference variance; the closer the ratio of the calculated variance to the long-term reference variance is to 1, the more stable the fluctuation of the micro-meteorological data, meteorological station data and high-resolution meteorological grid data is; The reliability score of the microclimate data, the weather station data and the high-resolution weather grid data in the time window is calculated by using a weighted geometric mean or a weighted arithmetic mean in combination with the data integrity, the data mutation frequency and the variance stability ; Synthesis spatial distance weight coefficient With reliability weight coefficient Obtain the weight ω of microclimate data, weather station data and high-resolution weather grid data: （3）； wherein is an adjustable hyperparameter; is an adjustable hyperparameter; is an adjustable hyperparameter; is an adjustable hyperparameter; According to the obtained weight ω, the weighted average value of the micro-meteorological data, meteorological station data and high-resolution meteorological grid data is calculated to obtain the fused multi-source data; Step S3: the lemming optimization algorithm is improved by introducing the elite archive mechanism and energy factor, and the improved lemming optimization algorithm is obtained; the Attention-LSTM model is optimized by using the improved lemming optimization algorithm, and the current optimal solution is obtained; Step S4: the Attention-LSTM model is trained by using the fused multi-source data and the current optimal solution, and the transmission line icing prediction model is obtained; the historical icing data is divided into training set and test set, the transmission line icing prediction model is trained and tested, and the evaluated transmission line icing prediction model is obtained; Step S5: obtaining real-time icing data of a target area, inputting the real-time icing data into the evaluated transmission line icing prediction model for prediction, and outputting the prediction results of the line icing thickness, icing type, risk level and ice melting time of the target area; Step S6: based on the prediction results of the line icing thickness, icing type, risk level and ice melting time, decision aid information is generated.

2. The multi-dimensional icing prediction method based on multi-source data according to claim 1, characterized in that: The historical icing data of the target area includes micro-meteorological data, meteorological station data and high-resolution meteorological grid data; The meteorological station data is derived from the meteorological station near the icing tower: the observation equipment of the meteorological station records the observation value of temperature, humidity, wind speed and rainfall every hour in the icing period; the meteorological station data is composed based on the observation value recorded every hour in the icing period; The observation value is preprocessed to obtain the corresponding high-resolution meteorological grid data; The microclimate data are data of temperature and humidity of the icing tower in each hour collected by the microclimate collecting device during the icing period.

3. The multi-dimensional icing prediction method based on multi-source data according to claim 2, characterized in that: The specific process of obtaining the fused multi-source data is: The distances of the iced pole tower from the micro-meteorological collection device, the weather station, and the corresponding grid center in the high-resolution meteorological grid point data are calculated and denoted as d1, d2, and d3, respectively; and the reciprocal of the square of the distances d1, d2, and d3 is taken as a spatial distance weight coefficient , which indicates that （1）； In the formula, d is distance; i is an index variable; i = 1, 2, 3; is the distance from the iced pole to the micro-meteorological acquisition device; is the distance from the iced pole to the meteorological station; is the distance from the iced pole to the center of the corresponding grid in the high-resolution meteorological grid data.

4. The multi-dimensional icing prediction method based on multi-source data according to claim 3, characterized in that: The specific process of the improved vole optimization algorithm is: The vole optimization algorithm includes six stages: initialization stage, long-distance migration stage, hole digging stage, foraging stage, and predator avoidance stage. The initialization stage of the vole optimization algorithm generates a random population in the search space as the initial population. The initial position of the initial population is initialized using Latin hypercube sampling, which is represented by: （4）； wherein is the initial position of the i-th vole in the j-th dimension; is the lower bound of the j-th dimension; is a random number generated by Latin hypercube sampling for the i-th vole in the j-th dimension, ensuring uniform distribution of the initial population; is the upper bound of the j-th dimension; The long-distance migration stage is represented by: （5）； wherein is the position of the i-th vole after long-distance migration at the (t+1)-th iteration; is the position of the best vole in the initial population at the t-th iteration; is a flag to change the search direction; is a random number vector of the probability density function of Brownian motion; is a random number generated uniformly in the interval [0, 1] to balance the position of the best vole in the initial population; is the current position of the i-th vole at the t-th iteration; is the position of the a-th vole randomly selected from the initial population at the t-th iteration; The flag F that changes the search direction is represented by: （6）； In the formula, is a random number uniformly generated in the interval [0, 1], and by introducing the random number, the probability of switching the search direction is determined. The random number vector of the probability density function of Brownian motion is represented by: （7）； wherein is the natural exponential function; is a random number vector; is an exponential term of Gaussian distribution; The hole digging stage is represented by: （8）； In the formula, is a dynamically adjusted parameter that controls the step size or exploration range of the lemming digging behavior; is the position of the bth lemming randomly selected from the initial population at the tth iteration. The dynamically adjusted parameter is represented by: （9）； wherein is a sine function; is half of the current time step ; The foraging stage is represented by: （10）； wherein is a parameter of the spiral wrapping mechanism, simulating the spiral movement path of the lemming within the foraging area; The parameter of the spiral winding mechanism is represented by: （11）； （12）； wherein is the radius of the spiral; is the cosine function; is the dimension of the search space; is the coordinate component of the best lemming in the initial population in the jth dimension at the tth iteration, wherein is the coordinate component of the best lemming in the initial population in the jth dimension at the tth iteration, wherein is the coordinate component of the best lemming in the initial population in the jth dimension at the tth iteration, wherein is the coordinate component of the i th lemming in the jth dimension at the tth iteration; The predator avoidance stage is represented by: （13）； （14）； wherein is a parameter that is dynamically adjusted to control the strength or step size of the vole's evasion behavior; is a Levy flight random number vector; is the maximum number of iterations; The Lévy flight random number vector is represented by: （15）； wherein is a random number following a normal distribution, in the numerator; is the standard deviation of the Levy distribution; is a random number following a normal distribution, in the denominator; is the Levy index; Γ is the gamma function; The elite archive mechanism is introduced: establishing an elite archive for storing the positions of the best lemmings found so far, at the end of each iteration, calculating the fitness values of the lemmings in the initial population according to a predefined fitness function, and sorting the lemmings in the initial population according to the fitness values, then selecting the positions of the top k best lemmings in the sorting; comparing the positions of the best lemmings with the positions of the best lemmings stored in the elite archive, and selecting the position of a best lemming with the highest similarity value as an elite lemming ​ Will Joining the long-distance migration phase means: （16）； wherein is a social learning factor; is a random number in the interval [0, 1] that regulates the elite lemming by a random number the degree of influence on the current lemming position update; Will Adding the digging phase means: （17）； Will Joining the foraging phase indicates: （18）； Add Join the evading stage, indicating: （19）； The energy factor E is set, and when the energy factor E changes, the vole enters different stages according to the size of the energy factor E. Energy factor The calculation of the energy factor, E, indicates that: （20）； wherein is the energy factor at the tth iteration; is the minimum value of the energy factor; is the maximum value of the energy factor; is the initial population diversity index at the tth iteration; The initial population diversity index at the tth iteration is represented by: （21）； wherein is the initial population size, i.e. the number of voles in the initial population; is the dimension of the search space; is the average position coordinate of the initial population in the j-th dimension at the t-th iteration; is the initial position of the i-th vole in the j-th dimension at the t-th iteration; The energy factor at the iteration number t is calculated according to formula (20) The energy factor at the iteration number t is determined When the energy factor at the iteration number t reaches the set energy threshold value, the exploration stage is entered, the exploration stage includes the long-distance migration stage and the digging stage, and when the energy factor at the iteration number t does not reach the set energy threshold value, the foraging stage is entered, and The formula for determining the updated lemming position is used in different stages. In the exploration phase, when less than 0.3, the lemming is considered to be in the long migration phase, and the lemming's position is updated using equation (16); when greater than 0.3, the lemming is considered to be in the burrowing phase, and the lemming's position is updated using equation (17). When Less than 0.5, the lemming is considered to be in the foraging stage, and the position of the lemming is updated using equation (18); when Greater than 0.5, the lemming is considered to be in the predator avoidance stage, and the position of the lemming is updated using equation (19). After the vole position is updated in the exploration stage and the foraging stage, the new position of the vole after updating formula (16), formula (17), formula (18), and formula (19) is evaluated for fitness value. The iteration number t is increased by 1, i.e. ; is the current iteration number; determining whether a maximum number of iterations is reached , when the maximum number of iterations is not reached , returning to equations (20) and (21) to recalculate the energy factor ; when the maximum number of iterations is reached the optimization process is finished.

5. The multi-dimensional icing prediction method based on multi-source data according to claim 4, characterized in that: The specific process of obtaining the evaluated power line icing prediction model is: The historical icing data is divided into a training set and a test set. The training set is used to train the power line icing prediction model, and the test set is used to test the accuracy of the trained power line icing prediction model.

6. The multi-dimensional icing prediction method based on multi-source data according to claim 5, characterized in that: The coefficient of determination R2 and the minimum mean square error MSE are selected as performance indicators to evaluate the accuracy performance of the trained power line icing prediction model. The specific process of generating decision assistance information is: Real-time icing data of the target area is obtained, and the real-time fusion multi-source data is obtained by preprocessing the real-time icing data in step S2. The real-time fusion multi-source data of the target area is input into the evaluated power line icing prediction model to predict the line icing condition of the target area.

7. An icing multi-dimensional prediction system based on multi-source data, used to implement the icing multi-dimensional prediction method based on multi-source data in any one of claims 1-6. The prediction results of the line icing thickness, icing type, risk level, and ice melting time of the target area are input into the decision rule library to generate decision assistance information. The specific process of generating decision assistance information is: The data acquisition module is used to obtain historical icing data of the target area. The data processing module is used to preprocess the obtained historical icing data to obtain fused multi-source data. The model optimization module is used to introduce the elite archive mechanism and the energy factor into the vole optimization algorithm to obtain the improved vole optimization algorithm. The improved vole optimization algorithm is used to optimize the Attention-LSTM model to obtain the current optimal solution. The training module is configured to train the Attention-LSTM model by using the fused multi-source data and the current optimal solution to obtain a transmission line icing prediction model; divide historical icing data into a training set and a test set, train and test the transmission line icing prediction model, and obtain an evaluated transmission line icing prediction model; The prediction module is configured to acquire real-time icing data of a target area, input the real-time icing data into the evaluated transmission line icing prediction model for prediction, and output a prediction result of line icing thickness, icing type, risk level and ice-melting time of the target area. The generation module is configured to generate decision assistance information based on the prediction result of the line icing thickness, the icing type, the risk level and the ice-melting time.

8. A computer device, comprising: comprise: one or more processors; the processor is configured to store one or more programs; when the one or more programs are executed by the one or more processors, a method for implementing the multi-dimensional icing prediction method based on multi-source data according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, comprise: a computer program is stored thereon, and the computer program is executed to implement the multi-dimensional icing prediction method based on multi-source data according to any one of claims 1-6.

Images