A power transmission line icing disaster early warning method and system and a storage medium

Abstract

The application belongs to the technical field of icing disaster early warning, and discloses a power transmission line icing disaster early warning method, a system and a storage medium. A weather prediction model is constructed based on a convolutional neural network, current meteorological data is input into the trained weather prediction model to output meteorological prediction data, extreme weather features are extracted from historical meteorological data, and an extreme weather comprehensive influence factor is extracted and constructed. The icing thickness of the power transmission line is calculated through a physical model, and the icing type and local icing density are obtained through icing shape recognition. The results are input into a preset icing prediction model to output icing prediction data. The icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive influence factor is established, the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive influence factor is obtained, the icing prediction correction value under the sub-interval of the corresponding comprehensive influence factor is obtained, and the icing prediction data is corrected for disaster assessment.

Description

Technical Field

[0001] This invention belongs to the field of power system maintenance technology, specifically relating to a method, system, and storage medium for early warning of icing disasters on transmission lines. Background Technology

[0002] In today's power transmission networks, transmission lines, as key infrastructure for power transmission, play a vital role in the normal operation of society due to their safe and stable operation. However, icing on transmission lines poses a serious threat, especially in northern regions such as Shanxi Province, where the problem is particularly prominent due to its specific geographical and climatic conditions. With the impact of global climate change, the frequency and intensity of extreme weather events have increased. In winter or cold weather conditions, phenomena such as rime, hoarfrost, and wet snow occur frequently. When these precipitation forms encounter low temperatures, they are very easy to freeze on the surface of transmission lines, forming icing. The formation of icing not only increases the weight of transmission lines but also changes their mechanical and electrical properties, leading to a series of serious problems.

[0003] However, existing monitoring methods often only measure basic parameters such as the thickness of ice accretion, failing to fully consider local weather conditions, ice formation mechanisms, and other factors. This results in insufficient accuracy and timeliness of early warnings, making it difficult to effectively improve the power grid's ability to defend against ice accretion disasters and the grid's operation and maintenance capabilities. Furthermore, existing early warning systems often do not fully integrate local geographical environment, meteorological characteristics, and specific structural parameters of transmission lines, making early warning information inaccurate and unable to meet the actual needs of power grid maintenance personnel for early prevention and effective response to ice accretion disasters. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, and storage medium for early warning of icing disasters on power transmission lines. The method calculates the icing trend coefficient based on icing prediction data and current real-time status data, and establishes the icing prediction accuracy density function under each sub-interval of the comprehensive impact factor of extreme weather based on the icing trend coefficient, thereby obtaining the corresponding icing prediction accuracy under the sub-interval of the comprehensive impact factor of extreme weather, thus improving the prediction accuracy.

[0005] The technical solution to achieve the purpose of this invention is as follows:

[0006] A method for predicting icing on power transmission lines includes the following steps:

[0007] An icing prediction model is constructed, and icing prediction data is obtained based on the constructed icing prediction model. The icing prediction data includes the icing location and the corresponding icing thickness prediction value.

[0008] The icing trend coefficient is calculated based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, the icing prediction accuracy density function under each sub-interval of the comprehensive impact factor of extreme weather is established. The corresponding icing prediction accuracy under each sub-interval of the comprehensive impact factor of extreme weather is obtained, and the icing prediction correction value under each sub-interval of the comprehensive impact factor is obtained.

[0009] The icing prediction data is corrected based on the icing prediction correction value to obtain the final icing prediction data.

[0010] Furthermore, an icing prediction model is constructed, and icing prediction data obtained based on the constructed icing prediction model includes:

[0011] Obtain the structural data of the transmission line and the ice thickness, ice type and local ice density under historical icing conditions, and calculate the total load of the distribution network line under icing conditions based on the structural data;

[0012] Based on the overlapping relationship between the total load and the bearing capacity of the distribution network line, a fault prediction model is established, and the AHP-entropy weight method is used to determine the parameter weights in the fault prediction model. The fault probability of the distribution network line under icing is calculated. Based on the icing thickness, icing type and local icing density, and based on the icing physical mechanism, an icing impact model is constructed. The corresponding fault probability is imported into the icing impact model to obtain multiple sets of icing disaster impact maps.

[0013] An icing prediction model is constructed based on LSTM time series. The icing disaster impact map is used as the training sample. The LSTM layer in the LSTM time series is a multi-layer bidirectional LSTM, with each layer including multiple units to capture long-term dependencies. Its fully connected layer is used to output icing prediction data within the prediction period.

[0014] Furthermore, methods for obtaining icing thickness, icing type, and local icing density include:

[0015] Real-time status data on the power transmission line is acquired through multi-source sensors, including meteorological data and image data. Based on the temperature, humidity and wind speed parameters in the meteorological data, the icing thickness of the power transmission line is calculated through a physical model. The icing morphology is identified by deep learning algorithm on the image data to obtain the icing type and local icing density.

[0016] Furthermore, the methods for calculating icing thickness, icing type, and local icing density include:

[0017] Obtain the icing time t, and calculate the radius R of the transmission line after icing: Where C is a constant preset according to the material of the transmission line, and T is the absolute temperature;

[0018] Calculation of icing mass per unit length based on dry growth model Where R0 is the radius of the transmission line, v is the impact velocity of the supercooling water droplets, and Wt is the liquid water content in the air;

[0019] Calculate ice thickness , where ρ is the density of ice;

[0020] A deep learning algorithm is used to construct an ice classification model. Features are input into the trained ice classification model, and the probability distribution of ice types is output. The category with the highest probability is selected as the final result to obtain the classification result.

[0021] The number of pixels in the icing area is counted in the segmentation mask based on the pixel statistics method, and the icing cross-sectional area is calculated by combining the wire diameter. Then, it is converted into a density level by empirical formula. After integration, an icing density heat map is obtained, and the local icing density is extracted based on the icing density heat map.

[0022] Furthermore, methods for constructing comprehensive impact factors of extreme weather include:

[0023] Based on the current forecast cycle, a historical judgment cycle is set, historical meteorological data within the historical judgment cycle is obtained, extreme meteorological characteristics are extracted based on the historical meteorological data, and the meteorological forecast data is judged to be in an extreme weather state based on the preset meteorological judgment criteria. The comprehensive impact factor of extreme weather is extracted and constructed.

[0024] Furthermore, the comprehensive impact factors of extreme weather were extracted and constructed, including:

[0025] Get the current forecast period T1, set the historical judgment period T2 with years as the retrospective unit, and get the historical meteorological data within the historical judgment period;

[0026] Extreme weather features are extracted from historical meteorological data and used to characterize extreme weather.

[0027] Obtain meteorological forecast data, extract extreme weather characteristics under the current state based on the meteorological forecast data and current meteorological data, and determine that the meteorological forecast data is under extreme weather conditions if the extreme weather characteristics under the current state exceed the preset meteorological judgment criteria.

[0028] The comprehensive impact factor of extreme weather can be constructed based on the following formula:

[0029] , where i = 1, 2, 3, ..., n, and n is the specific number of days in the current prediction period;

[0030] Among them, Px i Let ΔPx be the comprehensive impact factor of extreme weather on day i under extreme weather characteristic x. i,max Let ΔPx be the maximum fluctuation value among all characteristic quantities on day i under extreme weather condition x.i,avg Let Px be the average fluctuation value of each characteristic quantity on the i-th day under the x-th extreme weather characteristic. i，avg It represents the average value of each characteristic quantity on the i-th day under the x-th extreme weather characteristic.

[0031] Furthermore, the calculation of the icing trend coefficient includes:

[0032] Analysis sites were selected at each icing location, and the radius R of the transmission line after icing at each analysis site was obtained. The icing trend coefficient was then calculated using the following formula. Ui :

[0033]

[0034] Where H is the predicted icing thickness, α is the preset proportional coefficient, and R0 is the radius of the transmission line.

[0035] Furthermore, the icing prediction correction values ​​obtained for the corresponding comprehensive impact factors in sub-intervals include:

[0036] The meteorological data of the comprehensive impact factors of extreme weather are divided into n intervals, as shown below:

[0037]

[0038] in, To construct the maximum value of meteorological data for the comprehensive impact factor of extreme weather, To construct the minimum value of meteorological data for the comprehensive impact factors of extreme weather, The unit for dividing intervals;

[0039] The prediction accuracy loss is calculated based on real-time status data and predicted ice thickness, and the prediction accuracy loss is divided into corresponding intervals.

[0040] Based on meteorological data from various intervals, the accuracy of icing prediction is calculated using nonparametric estimation methods. The probability density function of the difference between the actual icing data and the predicted icing data is obtained. :

[0041]

[0042] in, To predict the accuracy loss, N is The number of parameters, h is a preset smoothing parameter, and G is the kernel function. Ui This represents the icing trend coefficient.

[0043] S504. Fit the icing prediction accuracy density function for each sub-interval of the comprehensive impact factors of different extreme weather, and obtain its cumulative probability distribution function by integration to obtain the icing prediction correction value for each sub-interval of the comprehensive impact factors of extreme weather.

[0044] Furthermore, after obtaining the final icing prediction data, the following is also included:

[0045] Obtain the preset icing disaster assessment range (Hmin, Hmax), where Hmin is the minimum value of icing disaster assessment and Hmax is the maximum value of icing disaster assessment. If the icing hazard coefficient is less than Hmax and greater than Hmin, a level one warning signal is generated.

[0046] If the icing hazard factor is greater than or equal to Hmax, a level II warning signal will be generated;

[0047] The system sends a warning signal containing location information to the maintenance terminal via a wireless communication module.

[0048] This invention also discloses a transmission line icing prediction system, comprising:

[0049] The icing prediction module constructs an icing prediction model and obtains icing prediction data based on the constructed icing prediction model. The icing prediction data includes the icing location and the corresponding icing thickness prediction value.

[0050] The icing prediction correction module calculates the icing trend coefficient based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, it establishes the icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive impact factor, obtains the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive impact factor, and obtains the icing prediction correction value under the corresponding sub-interval of the comprehensive impact factor.

[0051] The correction and evaluation module corrects the icing prediction data based on the icing prediction correction value to obtain the final icing prediction data.

[0052] The present invention also discloses a computer storage medium storing a computer program thereon, wherein when the computer executes the computer program, it implements the transmission line icing prediction method described in any of the above claims.

[0053] The present invention also discloses an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program stored in the memory, wherein the computer program, when executed, implements the transmission line icing prediction method described in any of the above claims.

[0054] Compared with the prior art, the significant advantages of this invention are:

[0055] This invention constructs a weather prediction model based on a convolutional neural network. It uses current meteorological data as input to train the model and outputs meteorological prediction data. It also extracts extreme weather characteristics from historical meteorological data, extracts and constructs a comprehensive impact factor for extreme weather, calculates the icing thickness of transmission lines using a physical model, and identifies icing morphology to determine icing type and local icing density. The results are input into a pre-defined icing disaster assessment model, outputting icing prediction data. An icing prediction accuracy density function is established for each sub-interval of the comprehensive impact factor for extreme weather, thereby obtaining the corresponding icing prediction accuracy for each sub-interval of the comprehensive impact factor for extreme weather. Furthermore, it obtains the icing prediction correction value for each sub-interval of the comprehensive impact factor for correcting the icing prediction data and conducting disaster assessment. Based on existing icing analysis and prediction technologies, this invention considers the impact of extreme weather on icing prediction, thus improving prediction accuracy. Attached Figure Description

[0056] Figure 1 This is a flowchart of the transmission line icing prediction method in this embodiment;

[0057] Figure 2 This is a block diagram of the transmission line icing disaster early warning system in this embodiment;

[0058] Figure 3 This is a flowchart of the transmission line icing disaster early warning method in this embodiment. Detailed Implementation

[0059] The principle of this invention is as follows: Calculate the icing trend coefficient based on icing prediction data and current real-time status data; establish the icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive impact factor based on the icing trend coefficient; thereby obtain the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive impact factor; and obtain the icing prediction correction value under the corresponding sub-interval of the comprehensive impact factor to correct the icing prediction data and conduct disaster assessment.

[0060] Example:

[0061] like Figure 1 As shown, a method for predicting icing on transmission lines includes the following steps:

[0062] An icing prediction model is constructed, and icing prediction data is obtained based on the constructed icing prediction model. The icing prediction data includes the icing location and the corresponding icing thickness prediction value.

[0063] The icing trend coefficient is calculated based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, the icing prediction accuracy density function under each sub-interval of the comprehensive impact factor of extreme weather is established. The corresponding icing prediction accuracy under each sub-interval of the comprehensive impact factor of extreme weather is obtained, and the icing prediction correction value under each sub-interval of the comprehensive impact factor is obtained.

[0064] The icing prediction data is corrected based on the icing prediction correction value to obtain the final icing prediction data.

[0065] This transmission line icing prediction method constructs a weather prediction model based on a convolutional neural network. It uses current meteorological data as input to train the model and outputs meteorological prediction data. It also extracts extreme weather characteristics from historical meteorological data, extracts and constructs a comprehensive impact factor for extreme weather, calculates the icing thickness of the transmission line using a physical model, and identifies the icing type and local icing density through icing morphology identification. The results are input into a pre-set icing prediction model, outputting icing prediction data. A density function for the icing prediction accuracy under each sub-interval of the comprehensive impact factor for extreme weather is established, thereby obtaining the corresponding icing prediction accuracy under each sub-interval of the comprehensive impact factor for extreme weather. Furthermore, the icing prediction correction value under each sub-interval of the comprehensive impact factor is obtained to correct the icing prediction data and conduct disaster assessment. Based on existing icing analysis and prediction technologies, this method considers the impact of extreme weather on icing prediction, improving prediction accuracy.

[0066] Another embodiment, such as Figure 2 As shown, a transmission line icing prediction system includes:

[0067] The icing prediction module constructs an icing prediction model and obtains icing prediction data based on the constructed icing prediction model. The icing prediction data includes the icing location and the corresponding icing thickness prediction value.

[0068] The icing prediction correction module calculates the icing trend coefficient based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, it establishes the icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive impact factor, obtains the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive impact factor, and obtains the icing prediction correction value under the corresponding sub-interval of the comprehensive impact factor.

[0069] The correction and evaluation module corrects the icing prediction data based on the icing prediction correction value to obtain the final icing prediction data.

[0070] Based on a transmission line icing prediction system, an early warning module can be included to form a transmission line icing disaster early warning system. The early warning module is used to obtain a preset icing disaster assessment range. When the icing prediction data exceeds the preset threshold, a multi-level early warning mechanism is triggered, and an early warning signal containing location information is sent to the operation and maintenance terminal through a wireless communication module.

[0071] The following example illustrates the workflow and implementation method of a power transmission line icing disaster early warning system. Figure 3 As shown, the specific steps include:

[0072] Step 1: Obtain historical satellite data and historical forecast meteorological data calculated based on historical satellite data as training samples. Construct a weather forecast model based on a convolutional neural network. Obtain current meteorological data and input it into the trained weather forecast model to output meteorological forecast data.

[0073] Step 2: Set a historical judgment period based on the current forecast period, obtain historical meteorological data within the historical judgment period, extract extreme meteorological characteristics based on the historical meteorological data, determine whether the meteorological forecast data is in an extreme weather state based on the preset meteorological judgment criteria, and extract and construct a comprehensive impact factor of extreme weather.

[0074] Step 3: Acquire real-time status data on the transmission line through multi-source sensors. The real-time status data includes meteorological data and image data. Based on the temperature, humidity and wind speed parameters in the meteorological data, calculate the icing thickness of the transmission line through a physical model. Use deep learning algorithms to identify the icing morphology of the image data to obtain the icing type and local icing density.

[0075] Step 4: Obtain the icing type and local icing density. Input the icing thickness, icing type, local icing density, and meteorological forecast data into the preset icing prediction model and output the icing prediction data.

[0076] Step 5: Calculate the icing trend coefficient based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, establish the icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive impact factor, thereby obtaining the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive impact factor, and obtaining the icing prediction correction value under the corresponding sub-interval of the comprehensive impact factor.

[0077] Step 6: Correct the icing prediction data according to the icing prediction correction value to obtain accurate icing prediction data (final icing prediction data), obtain the preset icing disaster assessment range, and when the final icing prediction data exceeds the preset threshold, trigger a multi-level early warning mechanism and send an early warning signal containing location information to the operation and maintenance terminal through the wireless communication module.

[0078] Specifically, the process of obtaining meteorological forecast data is as follows:

[0079] S101. Obtain historical satellite data and historical forecast meteorological data calculated based on historical satellite data. The historical forecast meteorological data includes wind speed, wind direction, temperature, humidity, air pressure, as well as rainfall, snowfall, and icing area. Establish a data fluctuation curve based on the time axis as a training sample.

[0080] S102. The generated samples are divided into training set, validation set and test set according to the ratio of 7:2:1. The training set is used for network training, the validation set is used for cross-validation to avoid overfitting, and the test set is used as new data that has not been learned by the network to evaluate the actual accuracy of the network after training.

[0081] S103. Construct a weather prediction model based on a BP neural network, download the weight file and load it onto the corresponding network, initialize the BP neural network, and obtain the initialized BP neural network.

[0082] S104. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the weather forecast data, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use fixed step size decay to optimize the training parameters, and retrain the entire network.

[0083] S105. During the training process, small batches of data fluctuation curves are randomly and non-repeatedly extracted from the training set for training. One training cycle is completed after all data fluctuation curves are extracted. The training is completed after a certain number of cycles, and the trained weather prediction model is obtained.

[0084] S106. Use the test set to evaluate the model's performance. Once the model's performance reaches the preset evaluation criteria, obtain the current meteorological data and input it into the trained weather prediction model to output meteorological prediction data.

[0085] Specifically, the process of extracting and constructing the comprehensive impact factors of extreme weather is as follows:

[0086] S201. Obtain the current forecast period T1, set the historical judgment period T2 with the year as the retrospective unit, and obtain the historical meteorological data within the historical judgment period. The historical meteorological data includes wind speed, wind direction, temperature, humidity, air pressure, as well as rainfall, snowfall, and icing area.

[0087] S202. Extract extreme weather features from historical meteorological data. Extreme weather features are used to characterize extreme weather, that is, the significant features of extreme weather extracted from a large amount of historical meteorological data. Extreme weather features include wind speed increment, temperature drop difference, wind direction difference, humidity difference, air pressure change difference, rainfall increment, snowfall increment, and ice cover area increment within the historical judgment period.

[0088] S203. Obtain preset meteorological judgment standards, namely wind speed increment judgment threshold, temperature drop difference judgment threshold, wind direction difference judgment threshold, humidity difference judgment threshold, air pressure change difference judgment threshold, rainfall increment judgment threshold, snowfall increment judgment threshold and ice cover area increment judgment threshold.

[0089] S204. Obtain meteorological forecast data, extract extreme meteorological characteristics under the current state based on meteorological forecast data and current meteorological data, and if any characteristic quantity of extreme meteorological characteristics under the current state exceeds the meteorological judgment standard, then the meteorological forecast data is established to be under extreme weather conditions.

[0090] S205. Construct the comprehensive impact factor of extreme weather based on the following formula:

[0091] , where i = 1, 2, 3, ..., n, and n is the specific number of days in the current prediction period;

[0092] Among them, Px i Let ΔPx be the comprehensive impact factor of extreme weather on day i under extreme weather characteristic x. i，max Let ΔPx be the maximum fluctuation value among all characteristic quantities on day i under extreme weather condition x. i，avg Let Px be the average fluctuation value of each characteristic quantity on the i-th day under the x-th extreme weather characteristic. i，avg It represents the average value of each characteristic quantity on the i-th day under the x-th extreme weather characteristic.

[0093] Specifically, the process for calculating the icing thickness of transmission lines and obtaining the icing type and local icing density is as follows:

[0094] S301. Obtain the icing time t, and calculate the radius R of the transmission line after icing based on the Imai model: Where C is a constant preset according to the material of the transmission line, and T is the absolute temperature;

[0095] S302. Under ideal conditions, all supercooled water droplets freeze immediately upon impact with the power line. Calculate the icing mass M per unit length based on the dry growth model: Where R0 is the radius of the transmission line, v is the impact velocity of the supercooled water droplets (vertical component of wind speed or effective collision velocity), and Wt is the liquid water content in the air;

[0096] S303. Calculate the icing thickness D using the following formula: , where ρ is the density of ice;

[0097] S304. Collect images of icing on power transmission lines using drones, including image data of different icing types, icing densities, and environmental conditions. Manually annotate the image data, marking the boundaries of icing areas and icing type labels. Simultaneously, expand the target dataset by rotating, flipping, scaling, and adding noise.

[0098] S305. Extract texture features, shape features and color features from the segmented ice-covered areas based on the target dataset. Construct an ice-covered classification model using a deep learning algorithm. Input the features into the trained ice-covered classification model and output the probability distribution of ice-covered types. Select the category with the highest probability as the final result to obtain the classification result.

[0099] The number of pixels in the icing area is counted in the segmentation mask based on the pixel statistics method, and the icing cross-sectional area is calculated by combining the wire diameter. Then, it is converted into a density level by empirical formula. After integration, an icing density heat map is obtained, and the local icing density is extracted based on the icing density heat map.

[0100] Specifically, the process of outputting icing prediction data is as follows:

[0101] S401. Obtain the structural data of the transmission line and the ice thickness, ice type and local ice density under historical icing conditions, and calculate the total load of the distribution network line under icing conditions based on the structural data.

[0102] S402. Based on the overlapping relationship between the total load and the bearing capacity of the distribution network line, a fault prediction model is established, and the AHP-entropy weight method is used to determine the parameter weights in the fault prediction model in order to calculate the fault probability of the distribution network line when it is covered with ice. Based on the ice thickness, ice type and local ice density, and based on the ice physical mechanism, an ice impact model is constructed, and the corresponding fault probability is imported into the ice impact model to obtain multiple sets of ice disaster impact maps.

[0103] The specific process of establishing the fault prediction model is as follows:

[0104] Acquire structural and meteorological data of power distribution network lines, including line type, tower type, conductor specifications, and meteorological data including wind speed, temperature, humidity, and snowfall. Clean the structural and meteorological data to remove outliers and missing values, and normalize data of different dimensions.

[0105] Based on meteorological data, the existing icing growth model is used to predict the icing thickness. Then, based on the icing thickness and line structure data, the total load on the line when it is covered with ice is calculated. The total load includes the line's self-weight, the weight of the ice, and the wind load.

[0106] Based on structural data and material properties, calculate the load-bearing capacity of the lines and towers, specifically including tensile strength and bending strength;

[0107] Further analysis of the overlap relationship between total load and bearing capacity was conducted. Based on the size and location of the overlapping area, machine learning algorithms were used to predict the probability of line failure when covered by ice, and the fault prediction area was identified. The fault prediction area is the area where the total load exceeds the bearing capacity.

[0108] The specific process of constructing the icing impact model is as follows:

[0109] Based on the Ima model, a classic icing growth model is established to obtain the icing mass per unit length of conductor, and then the current icing thickness is calculated using the mass and volume conversion formula.

[0110] An icing impact model is established based on a convolutional neural network. Historical icing thickness and historical line load data are used as training data for the model. The current icing thickness is used as the input layer, and the output layer is the line load data under the current icing thickness.

[0111] An icing prediction model is constructed based on LSTM time series. The icing disaster impact map is used as the training sample. The LSTM layer in the LSTM time series is a multi-layer bidirectional LSTM, with each layer including multiple units to capture long-term dependencies. Its fully connected layer is used to output icing prediction data within the prediction period.

[0112] Specifically, the process for obtaining the icing prediction correction value is as follows:

[0113] S501. Obtain icing prediction data, specifically the predicted icing location and corresponding icing thickness. Select analysis points at each icing location and obtain real-time status data for those points. The real-time status data is the radius R of the transmission line after icing. Calculate the icing trend coefficient using the following formula. Ui : , where H is the predicted ice thickness, α is the preset proportional coefficient, and the ice trend coefficient is used to assess the degree of ice accumulation under the current meteorological forecast data;

[0114] S502. Divide the meteorological data for constructing the comprehensive impact factor of extreme weather into n intervals, as shown below:

[0115]

[0116] in, To construct the maximum value of meteorological data for the comprehensive impact factor of extreme weather, To construct the minimum value of meteorological data for the comprehensive impact factors of extreme weather, The unit for dividing intervals;

[0117] The prediction accuracy loss is calculated based on real-time status data and predicted ice thickness, and the prediction accuracy loss is divided into corresponding intervals.

[0118] S503. Based on meteorological data from each interval, the accuracy of icing prediction can be calculated using nonparametric estimation methods. That is, the probability density function of the difference between actual icing data and predicted icing data. For the prediction accuracy loss corresponding to the comprehensive impact factors of extreme weather, its probability density function As shown below:

[0119]

[0120] in To predict the accuracy loss, N is The number of elements, h is the preset smoothing parameter, and G is the kernel function;

[0121] S504. The probability density function of the prediction accuracy loss under different comprehensive impact factor intervals is obtained by fitting, and its cumulative probability distribution function is obtained by integration. Thus, the icing prediction correction value under different sub-intervals of the comprehensive impact factors of each extreme weather is obtained.

[0122] Specifically, the process for obtaining the assessment and prediction results of ice accumulation disasters is as follows:

[0123] S601. Obtain the icing prediction correction value, which is the error value of the icing thickness prediction, and obtain the corrected icing prediction data H′.

[0124] S602. Obtain the preset icing disaster assessment range (Hmin, Hmax), where Hmin is the minimum value of icing disaster assessment and Hmax is the maximum value of icing disaster assessment. If the icing hazard coefficient is less than or equal to Hmin, no signal is generated.

[0125] If the icing hazard factor is less than Hmax and greater than Hmin, a Level 1 warning signal is generated.

[0126] If the icing hazard factor is greater than or equal to Hmax, a level 2 warning signal will be generated.

[0127] It should be noted that the range and threshold size are set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0128] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0129] In another embodiment, a computer storage medium stores a computer program thereon, wherein when a computer executes the computer program, it implements the transmission line icing disaster early warning method described in any of the above embodiments.

[0130] In another embodiment, an electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program stored in the memory, wherein the computer program, when executed, implements the transmission line icing disaster early warning method described in any of the preceding embodiments.

[0131] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method of predicting icing on a power transmission line, characterized by, Includes the following steps: An icing prediction model is constructed, and icing prediction data is obtained based on the constructed icing prediction model. The icing prediction data includes the icing location and the corresponding icing thickness prediction value. The construction of the icing prediction model, and the acquisition of icing prediction data based on the constructed icing prediction model, include: Obtain the structural data of the transmission line and the ice thickness, ice type and local ice density under historical icing conditions, and calculate the total load of the distribution network line under icing conditions based on the structural data; Based on the overlapping relationship between the total load and the bearing capacity of the distribution network line, a fault prediction model is established, and the AHP-entropy weight method is used to determine the parameter weights in the fault prediction model. The fault probability of the distribution network line under icing is calculated. Based on the icing thickness, icing type and local icing density, and based on the icing physical mechanism, an icing impact model is constructed. The fault probability is imported into the icing impact model to obtain multiple sets of icing disaster impact maps. An icing prediction model is constructed based on LSTM time series. The icing disaster impact map is used as a training sample. The LSTM layer in the LSTM time series is a multi-layer bidirectional LSTM, with each layer including multiple units to capture long-term dependencies. Its fully connected layer is used to output icing prediction data within the prediction period. The icing trend coefficient is calculated based on the icing prediction data and the radius of the transmission line after icing at the current icing location. Based on the icing trend coefficient, the icing prediction accuracy density function under each sub-interval of the comprehensive impact factor of extreme weather is established. The corresponding icing prediction accuracy under each sub-interval of the comprehensive impact factor of extreme weather is obtained, and the icing prediction correction value under each sub-interval of the comprehensive impact factor is obtained. The icing prediction data is corrected based on the icing prediction correction value to obtain the final icing prediction data.

2. The power transmission line icing prediction method according to claim 1, characterized by, Methods for obtaining icing thickness, icing type, and local icing density include: Real-time status data on the power transmission line is acquired through multi-source sensors, including meteorological data and image data. Based on the temperature, humidity and wind speed parameters in the meteorological data, the icing thickness of the power transmission line is calculated through a physical model. The icing morphology is identified by deep learning algorithm on the image data to obtain the icing type and local icing density.

3. The power transmission line icing prediction method according to claim 2, characterized by, The methods for calculating icing thickness, icing type, and local icing density include: The icing time t is acquired, and the radius R of the power transmission line after icing is calculated: where C is a constant preset according to the material of the power transmission line, and T is an absolute temperature. Calculating icing mass per unit length based on dry growth model where R0 is the radius of the power transmission line, v is the impact velocity of the supercooled water droplets, and Wt is the liquid water content in the air; Calculating ice thickness wherein ρ is the ice density; A deep learning algorithm is used to construct an ice classification model. Features are input into the trained ice classification model, and the probability distribution of ice types is output. The category with the highest probability is selected as the final result to obtain the classification result. The number of pixels in the icing area is counted in the segmentation mask based on the pixel statistics method, and the icing cross-sectional area is calculated by combining the wire diameter. Then, it is converted into a density level by empirical formula. After integration, an icing density heat map is obtained, and the local icing density is extracted based on the icing density heat map.

4. The power transmission line icing prediction method according to claim 1, characterized by, Methods for constructing comprehensive impact factors of extreme weather include: Based on the current forecast cycle, a historical judgment cycle is set, historical meteorological data within the historical judgment cycle is obtained, extreme meteorological characteristics are extracted based on the historical meteorological data, and the meteorological forecast data is judged to be in an extreme weather state based on the preset meteorological judgment criteria. The comprehensive impact factor of extreme weather is extracted and constructed.

5. The power transmission line icing prediction method according to claim 4, characterized by, The comprehensive impact factors of extreme weather were extracted and constructed, including: Obtaining the current prediction period T1, setting the historical judgment period T2 in the year as the backtracking unit, and obtaining the historical meteorological data in the historical judgment period; According to the historical meteorological data, the extreme weather feature is extracted, and the extreme weather feature is used to represent the extreme weather; Obtaining meteorological prediction data, extracting the extreme weather feature under the current state according to the meteorological prediction data and the current meteorological data, and if the extreme weather feature under the current state exceeds the preset meteorological judgment standard, it is judged that the meteorological prediction data is in the extreme weather state; According to the following formula, the extreme weather comprehensive influence factor is constructed: where i = 1, 2, 3,..., n, n is the specific number of days of the current prediction period; wherein, Px i is the integrated extreme weather impact factor of the i-th day under the x-th extreme weather characteristic, ΔPx i,max is the maximum fluctuation value of the i-th day under the x-th extreme weather characteristic, ΔPx i,avg is the average fluctuation value of the i-th day under the x-th extreme weather characteristic, Px i，avg is the average value of the i-th day under the x-th extreme weather characteristic.

6. The power line icing prediction method according to claim 1, characterized by, The calculation of the icing trend coefficient includes: Select analysis site one by one at icing site, get transmission line radius R after icing at analysis site, calculate icing tendency coefficient according to following formula Ui : Wherein, H is the icing thickness prediction value, a is a preset proportion coefficient, and R0 is the radius of the power transmission line.

7. The power line icing prediction method of claim 1, wherein Obtaining the icing prediction correction value under the sub-interval of the corresponding comprehensive influence factor includes: The meteorological data of the extreme weather comprehensive influence factor is divided into n intervals, as follows: wherein, maxi is the maximum value of the meteorological data for constructing the extreme weather comprehensive influence factor, mini is the minimum value of the meteorological data for constructing the extreme weather comprehensive influence factor, is the unit for dividing the interval; According to the real-time state data and the icing thickness prediction value, the prediction accuracy loss is calculated, and the prediction accuracy loss is divided into corresponding intervals; Based on the meteorological data of each interval, the icing prediction accuracy is solved by a non-parametric estimation method , and the probability density function of the difference between the actual icing data and the icing prediction data is obtained : wherein, N is the number of prediction accuracy loss, h is a preset smoothing parameter, and G is a kernel function, Ui is an icing trend coefficient. The icing prediction accuracy density function under each sub-interval of different extreme weather comprehensive influence factors is fitted, the cumulative probability distribution function is obtained by integration, and the icing prediction correction value under each sub-interval of each extreme weather comprehensive influence factor is obtained.

8. The power line icing prediction method of claim 1, wherein After obtaining the final icing prediction data, it further includes: Obtaining a preset icing disaster evaluation interval (Hmin, Hmax), Hmin is the minimum value of the icing disaster evaluation, and Hmax is the maximum value of the icing disaster evaluation. If the icing hazard coefficient is less than Hmax and greater than Hmin, a first-level warning signal is generated; If the icing hazard coefficient is greater than or equal to Hmax, a second-level warning signal is generated; Through the wireless communication module, the warning signal containing the position information is sent to the operation and maintenance terminal.

9. A power line icing prediction system characterized by, It includes: An icing prediction module, which constructs an icing prediction model, obtains icing prediction data based on the constructed icing prediction model, and the icing prediction data includes an icing position and a corresponding icing thickness prediction value; The construction of the icing prediction model and the obtaining of the icing prediction data based on the constructed icing prediction model include: Obtaining the structure data of the power transmission line and the icing thickness, the icing type and the local icing density under the history icing, calculating the total load of the distribution network line under the icing according to the structure data; According to the overlapping relationship between the total load and the bearing strength of the distribution network line, a fault prediction model is established, the AHP-entropy weight method is used to determine the parameter weight in the fault prediction model, the fault probability of the distribution network line under the icing is calculated, the icing thickness, the icing type and the local icing density are obtained, and the icing influence model is constructed based on the icing physical mechanism, the fault probability is introduced into the icing influence model, and a plurality of icing disaster influence graphs are obtained; Based on the LSTM time sequence, the icing prediction model is constructed, the icing disaster influence graph is taken as a training sample, the LSTM layer in the LSTM time sequence is a multi-layer bidirectional LSTM, each layer includes a plurality of units, the long-term dependence relationship is captured, and the full connection layer is used to output the icing prediction data in the prediction period. The icing prediction correction module calculates an icing trend coefficient according to the icing prediction data and the radius of the transmission line after icing at the current icing site, establishes an icing prediction accuracy density function under each sub-interval of the extreme weather comprehensive influence factor according to the icing trend coefficient, obtains the corresponding icing prediction accuracy under the sub-interval of the extreme weather comprehensive influence factor, and obtains the icing prediction correction value under the sub-interval of the corresponding comprehensive influence factor. The correction evaluation module corrects the icing prediction data according to the icing prediction correction value to obtain the final icing prediction data.

10. A computer storage medium having stored thereon a computer program, characterized in that The computer executes the computer program to implement the transmission line icing prediction method in any one of claims 1-8.

11. An electronic device comprising a memory and a processor, characterized in that, The memory stores the computer program, and the processor runs the computer program stored on the memory, and the computer program is executed to implement the transmission line icing prediction method in any one of claims 1-8.

Images