Power transmission line icing thickness staged prediction method, system, equipment and medium

Abstract

The invention relates to the technical field of disaster prevention and reduction of power systems, and discloses a power transmission line icing thickness staged prediction method, system and device and a medium. The method comprises the following steps: acquiring initial icing thickness data and future micro-meteorological data, and preprocessing the future micro-meteorological data to obtain an icing growth rate data set; screening the icing growth rate data set based on a gradient aggregation optimization tree algorithm to obtain a training set and a test set; performing parameter optimization based on the expectation improvement convergence framework and the adaptive core fusion machine, and determining optimal parameters of the model; performing training verification according to the model optimal parameters, the training set and the test set to establish a target icing growth rate prediction model; and based on the target icing growth rate prediction model and the initial icing thickness, calculating the icing thickness at the future moment in stages. According to the method, stage-by-stage differential prediction of icing growth, stabilization and melting stages is realized by constructing a cooperation mechanism, and the engineering practicability of prediction of the icing thickness of the power transmission line is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system disaster prevention and mitigation, and particularly relates to a transmission line icing thickness stage-by-stage prediction method, system, device and medium. BACKGROUND

[0002] The power industry is a meteorologically highly sensitive and demanding industry, and its production, construction and operation are greatly affected by environmental and meteorological factors. With the large-scale construction and development of power grids, long-distance and large-capacity transmission lines inevitably pass through micro-meteorological areas caused by special micro-terrain such as high-cold, high-altitude, high-humidity, heavy-rainfall areas or basins, watershed, etc., which are prone to cause transmission line icing disasters. The frequent occurrence of extreme abnormal climate caused by global warming has further aggravated the line icing disaster, which seriously threatens the safe and stable operation of the power system. Therefore, accurate prediction of icing thickness is crucial for power grid disaster prevention and mitigation. SUMMARY

[0003] The main purpose of the present application is to provide a transmission line icing thickness stage-by-stage prediction method, system, device and medium, which aims to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, the present application provides a transmission line icing thickness stage-by-stage prediction method, comprising:

[0005] Obtaining initial icing thickness data and future micro-meteorological data, and preprocessing the future micro-meteorological data to obtain an icing growth rate data set;

[0006] Filtering the icing growth rate data set based on a gradient aggregation optimization tree algorithm to obtain a training set and a test set;

[0007] Optimizing parameters based on an expected improvement convergence framework and an adaptive kernel fusion machine to determine the optimal parameters of the model;

[0008] Training and verifying according to the optimal parameters of the model and the training set and the test set to establish a target icing growth rate prediction model;

[0009] Based on the target icing growth rate prediction model and the initial icing thickness, the future icing thickness is calculated in stages.

[0010] In some embodiments, the obtaining initial icing thickness data and future micro-meteorological data, and preprocessing the future micro-meteorological data to obtain an icing growth rate data set comprises:

[0011] Obtaining initial icing thickness data and future micro-meteorological data;

[0012] Preprocess the future microclimate data based on icing physical conditions to eliminate abnormal data, and obtain processed equivalent icing thickness data;

[0013] Differentially process the processed equivalent icing thickness data to obtain an average icing growth rate;

[0014] Take the average of the microclimate data corresponding to the time period of the average icing growth rate as the microclimate input feature corresponding to the time period to generate an initial data set;

[0015] Normalize the data of the initial data set to generate an icing growth rate data set.

[0016] In some embodiments, the gradient boosting optimization tree algorithm filters the icing growth rate data set to obtain a training set and a test set, including:

[0017] The gradient boosting optimization tree algorithm calculates the importance score of each microclimate feature of the icing growth rate data set;

[0018] According to the importance score, the microclimate features are arranged in descending order, and the features with a cumulative contribution rate reaching a set threshold are selected to form a core feature set;

[0019] The core feature set is divided into a training set and a test set according to a stratified sampling method.

[0020] In some embodiments, the expected improvement convergence framework and adaptive kernel fusion are used for parameter optimization to determine the optimal parameters of the model, including:

[0021] According to the linear combination of the Gaussian kernel function and the polynomial kernel function, a combined kernel function is obtained;

[0022] An adaptive kernel fusion machine is constructed based on the combined kernel function;

[0023] The optimization parameter vector of the adaptive kernel fusion machine is determined, and the value range of each optimization parameter vector is set through exponential mapping;

[0024] The expected improvement convergence framework and the value range are used for parameter optimization iteration of the optimization parameter vector to obtain an optimal parameter combination;

[0025] The optimal parameter combination is used as the optimal parameters of the model.

[0026] In some embodiments, the expected improvement convergence framework and the value range are used for parameter optimization iteration of the optimization parameter vector to obtain an optimal parameter combination, including:

[0027] Optimize the weight of the combined kernel function and the parameter of each kernel function in the parameter vector to be optimized based on the expected improvement convergence framework and the value range;

[0028] The fitness function value obtained by cross-validation is used as the evaluation standard of the parameter, and the optimal parameter combination is obtained by iterative optimization.

[0029] In some embodiments, the optimization of the weight of the combined kernel function and the parameter of each kernel function in the parameter vector to be optimized based on the expected improvement convergence framework and the value range comprises:

[0030] Initialize the expected improvement convergence framework to obtain the optimization architecture of multiple sub-populations, and randomly generate multiple sets of parameter solutions as initial search points for each sub-population;

[0031] Set the maximum number of iterations;

[0032] Based on the initial search point, the value range and the maximum number of iterations, search in parallel through multiple sub-populations;

[0033] During the iteration process, the sub-population information interaction operation is performed once every preset number of iterations;

[0034] By calculating the Euclidean distance between the optimal solutions of each sub-population, the sub-populations whose difference degree meets the preset condition are selected for information sharing, and the excellent parameter solutions in the sub-population are transmitted to other populations.

[0035] In some embodiments, the future icing thickness is calculated in stages based on the target icing growth rate prediction model and the initial icing thickness, comprising:

[0036] Divide the future prediction time length into multiple equal interval stages;

[0037] Based on each equal interval stage, obtain the predicted growth rate function according to the target icing growth rate prediction model and the current microclimate data;

[0038] According to the initial icing thickness and the predicted growth rate function, the icing thickness is iteratively calculated by integral operation to obtain the icing thickness prediction value at different stages.

[0039] In addition, in order to achieve the above purpose, the application also provides a power transmission line icing thickness stage prediction system, comprising:

[0040] The data processing module is used for acquiring initial icing thickness data and future microclimate data, and preprocessing the future microclimate data to obtain an icing growth rate data set;

[0041] The data screening module is configured to screen the icing growth rate data set based on a gradient aggregation optimization tree algorithm to obtain a training set and a test set.

[0042] The parameter optimization module is configured to perform parameter optimization based on an expected improvement convergence framework and an adaptive kernel fusion machine to determine optimal parameters of the model.

[0043] The model training module is configured to perform training verification according to the optimal parameters of the model and the training set and the test set to establish a target icing growth rate prediction model.

[0044] The icing prediction module is configured to calculate the icing thickness at a future time in stages based on the target icing growth rate prediction model and the initial icing thickness.

[0045] In addition, to achieve the above object, the application further provides an electronic device, which comprises a memory, a processor, and a power line icing thickness staged prediction program stored in the memory and capable of running on the processor, and the power line icing thickness staged prediction program is configured to implement the power line icing thickness staged prediction method as described above.

[0046] In addition, to achieve the above object, the application further provides a storage medium, which stores a power line icing thickness staged prediction program, and the power line icing thickness staged prediction program is used to enable a processor to implement the power line icing thickness staged prediction method as described above when the processor executes the power line icing thickness staged prediction program.

[0047] The application provides a power transmission line icing thickness phased prediction method, comprising the following steps: obtaining initial icing thickness data and future microclimate data, and preprocessing the future microclimate data to obtain an icing growth rate data set; screening the icing growth rate data set based on a gradient aggregation optimization tree algorithm to obtain a training set and a test set; performing parameter optimization based on an expected improvement convergence framework and an adaptive kernel fusion machine to determine model optimal parameters; training and verifying according to the model optimal parameters and the training set and the test set to establish a target icing growth rate prediction model; and calculating the future time icing thickness in stages based on the target icing growth rate prediction model and the initial icing thickness. In the application, the gradient aggregation optimization tree algorithm is used to evaluate the feature importance of the microclimate data to screen out core features, the expected improvement convergence framework is used to globally optimize the parameters of the adaptive kernel fusion machine to improve the model generalization ability, and the initial icing thickness and the real-time predicted growth rate are iteratively accumulated to consider the diachronic cumulative effect, so that the icing growth, stability and melting stages are differentiated and predicted, a collaborative mechanism of 'feature dimension reduction-model optimization-dynamic accumulation' is constructed, and the problems of insufficient description of the icing dynamic process, low parameter optimization efficiency and poor adaptability of multi-stage prediction of the traditional method are effectively solved, and the engineering practicability of the power transmission line icing thickness prediction is improved. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A structural schematic diagram of an electronic device related to a hardware running environment of an embodiment scheme of the application;

[0049] Figure 2 A flowchart of a power transmission line icing thickness phased prediction method according to an embodiment of the application;

[0050] Figure 3 A whole flowchart of a power transmission line icing thickness phased prediction algorithm considering diachronic cumulative effect related to an embodiment scheme of the application;

[0051] Figure 4 A comparison chart of an icing growth stage prediction related to an embodiment scheme of the application;

[0052] Figure 5 A comparison chart of an icing stability stage prediction related to an embodiment scheme of the application;

[0053] Figure 6 A comparison chart of an icing melting stage prediction related to an embodiment scheme of the application;

[0054] Figure 7 A structural block diagram of a power transmission line icing thickness phased prediction system according to an embodiment of the application.

[0055] The objectives, functional features and advantages of the present application will be further described with reference to the embodiments in combination with the accompanying drawings. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0057] It should be noted that all the directionality indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative position relationship, movement condition, etc. between components in a certain specific posture (as shown in the accompanying drawings), and if the specific posture changes, the directionality indications also change accordingly.

[0058] In addition, the descriptions involving “first”, “second” and the like in the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features limited by “first” and “second” can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of a person of ordinary skill in the art, and when the combination of technical solutions appears to be contradictory or unachievable, it should be considered that the combination of technical solutions does not exist and is not within the protection scope of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0059] Reference Figure 1 , Figure 1 The electronic device structure schematic diagram of the hardware running environment involved in the embodiment of the present application.

[0060] As Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also be a storage device independent of the processor 1001.

[0061] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation to the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0062] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a power transmission line ice thickness stage prediction program.

[0063] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device, and the electronic device calls the transmission line ice thickness stage prediction program stored in the memory 1005 through the processor 1001, and executes the transmission line ice thickness stage prediction method provided by the embodiment of the present invention.

[0064] At present, the research progress on the prediction of ice thickness on transmission lines is as follows:

[0065] From the characterization of the icing dynamic process, the existing technology has systematic defects. ①The historical accumulation characteristics of the icing thickness are not considered, resulting in the prediction error of the continuous icing process being amplified over time; ②The nonlinear mutation of different stages of icing is ignored, and fixed model parameters are used, so that the prediction error of each stage is different. ③The data preprocessing mechanism of the existing technology has defects. The abnormal data processing is simple, and the data is not filtered in combination with the icing physical conditions, resulting in non-icing state data polluting the training set.

[0066] In summary, the icing growth is a dynamic process that accumulates over time, and existing models often ignore the influence of historical data on the current state. Due to the fact that the existing model does not fully consider the "time accumulation effect", the prediction error is large in different stages of icing development.

[0067] Based on the above analysis, in order to fully solve the problems of the existing research, the embodiment of the present application proposes a power transmission line icing thickness stage prediction method considering the time accumulation effect, aiming to realize the prediction of the line icing thickness by establishing a growth rate combination prediction model and combining the time accumulation effect, and to provide strong support for the safe operation of the power grid.

[0068] Therefore, the present application proposes a power transmission line icing thickness stage prediction method, system, device and medium.

[0069] The embodiment of the present application provides a power transmission line icing thickness stage prediction method, referring to Figure 2 , Figure 2 The flowchart of an embodiment of the power transmission line icing thickness stage prediction method of the present application.

[0070] As shown in Figure 2 , the power transmission line icing thickness stage prediction method comprises:

[0071] Step S100: acquiring initial icing thickness data and future microclimate data, and preprocessing the future microclimate data to obtain an icing growth rate data set;

[0072] Step S200: filtering the icing growth rate data set based on a gradient aggregation optimization tree algorithm to obtain a training set and a test set;

[0073] Step S300: parameter optimization based on an expected improvement convergence framework and an adaptive kernel fusion machine to determine the optimal parameters of the model;

[0074] Step S400: training and verification according to the optimal parameters of the model and the training set and the test set to establish a target icing growth rate prediction model;

[0075] Step S500: Based on the target ice accretion rate prediction model and the initial ice thickness, the ice thickness at future time is calculated in stages.

[0076] It should be noted that the execution subject in the embodiment can be an electronic device, which can be a computer device with data processing function, and can also be other devices that can realize the same or similar functions, and the embodiment does not limit this. In the embodiment, the computer device is taken as an example for description.

[0077] It can be understood that the method provided in the embodiment is a power transmission line ice thickness prediction method considering the chronological cumulative effect, which aims to realize the prediction of line ice thickness by establishing a growth rate combination prediction model and combining the chronological cumulative effect, and provide strong support for the safe operation of power grid. The improvement ideas of the embodiment include: using gradient aggregation optimization tree algorithm (GAOT algorithm) instead of traditional RF for feature selection, improving the accuracy of feature importance evaluation; using expected improvement convergence framework (EICF) instead of grid search to efficiently optimize the kernel parameters λ, r g , r1, r2, r3 of adaptive kernel fusion machine (AKF) in small sample scenarios, reducing the parameter optimization cost; designing a chronological cumulative prediction model, calculating the ice accretion rate by difference and updating the ice thickness iteratively combined with historical data, solving the problem of insufficient description of dynamic process of existing models.

[0078] Specifically, as shown in Figure 3 , first, the initial ice thickness and microclimate data are collected by using a laser ranging sensor and a meteorological monitoring station, and the data is preprocessed to generate an ice accretion rate dataset. Then, the gradient aggregation optimization tree (GAOT) algorithm is used to screen the core microclimate features affecting ice accretion, reduce the data dimension, and improve the accuracy of feature importance evaluation. Then, the kernel parameters of the adaptive kernel fusion machine (AKF) are optimized based on the expected improvement convergence framework (EICF) to construct an ice accretion rate prediction model. The time accumulation effect iteration mechanism is introduced, the initial ice thickness and the stage-by-stage growth rate prediction results are combined, and the dynamic calculation of the ice thickness at future time is realized. Finally, through the stage-by-stage weight adjustment strategy, the characteristics of different stages such as ice accretion, stability, and melting are adapted to ensure the reliability of the prediction results. The method of the embodiment effectively solves the problems of insufficient description of ice dynamic process and low parameter optimization efficiency of existing models, improves the precision and engineering applicability of power transmission line ice thickness prediction through multi-algorithm fusion, and provides technical support for power grid icing disaster warning and operation. The specific steps are described below.

[0079] In an embodiment, initial ice thickness data and future microclimate data are obtained, and the future microclimate data is preprocessed to obtain an ice growth rate data set, including: obtaining initial ice thickness data and future microclimate data; preprocessing the future microclimate data based on ice physical conditions to eliminate abnormal data to obtain processed equivalent ice thickness data; performing difference processing on the processed equivalent ice thickness data to obtain an average ice growth rate; taking the average of the microclimate data of the time period corresponding to the average ice growth rate as the microclimate input feature corresponding to the time period to generate an initial data set; and performing normalization processing on the data of the initial data set to generate an ice growth rate data set.

[0080] Specifically, past microclimate data, corresponding ice thickness data, and future microclimate data are obtained. The initial ice thickness can be obtained by deploying a laser ranging sensor on a power transmission line tower, and the microclimate data can be obtained by a weather monitoring station. The obtained microclimate data includes, but is not limited to, temperature, relative humidity, wind speed, wind direction, precipitation, atmospheric pressure, and solar radiation intensity.

[0081] Illustratively, initial ice thickness (initial ice thickness data) collection: the initial ice thickness H(t0) at the current time can be obtained by deploying a laser ranging sensor on a power transmission line tower. The measurement accuracy calibration formula of the sensor is:

[0082]

[0083] wherein H calibrated (t0) is the calibrated initial ice thickness, H raw (t0) is the original measured value of the sensor, k is the calibration coefficient, and b is the offset correction.

[0084] Illustratively, microclimate data collection: the future T hours of microclimate data are obtained by a weather monitoring station, including but not limited to temperature T(t), relative humidity RH(t), wind speed V(t), and other parameters. The data collection frequency can be set as Δt data , and a time series set is formed: .

[0085] Specifically, the icing and microclimate data are preprocessed, the abnormal data are removed, the icing growth rate dataset is generated, and the data are normalized. The abnormal data removal can be based on icing physical conditions, including temperature, relative humidity and wind speed satisfying certain conditions. When generating the icing growth rate dataset, the icing thickness data are differentially calculated to obtain the icing growth rate, and the microclimate data corresponding to the time period are averaged as input features. When normalizing the data, the standard deviation normalization method can be used to map features of different dimensions to a preset interval, eliminating the influence of dimensions on the model.

[0086] For example, the abnormal data are removed from the set based on the icing physical conditions, and the removal process is implemented by the following inequality group:

[0087]

[0088] where T(t) is the temperature, RH(t) is the relative humidity, and V(t) is the wind speed. Based on the inequality group, data points that do not meet the above conditions will be removed.

[0089] For example, the equivalent icing thickness data after removing the abnormal data are first-order differential processed, the icing thickness increment Δh in the time interval l is calculated, and the average icing growth rate Δh / l is obtained. At the same time, the microclimate data corresponding to the time interval are averaged as the microclimate input features of the period, forming the “microclimate average-icing growth rate” corresponding dataset (initial dataset).

[0090] For example, the standard deviation normalization method is used to map features of different dimensions to the [0, 1] interval, and the normalization formula is:

[0091]

[0092] where x norm (t) is the normalized feature value; x(t) is the original feature value; x min and x max are the historical minimum and maximum values of the feature x, respectively.

[0093] In an embodiment, the icing growth rate dataset is filtered based on the gradient boosting optimization tree algorithm to obtain a training set and a test set, including: calculating the importance score of each microclimate feature of the icing growth rate dataset based on the gradient boosting optimization tree algorithm; arranging the microclimate features in descending order according to the importance score, selecting features with a cumulative contribution rate reaching a set threshold to form a core feature set; and dividing the core feature set into a training set and a test set according to a stratified sampling method.

[0094] Specifically, the embodiment adopts a Gradient Aggregative Optimization Tree (GAOT) algorithm to screen core microclimate features affecting the icing growth rate, and then divides the data set into a training set and a test set. Among them, the Gradient Aggregative Optimization Tree algorithm is used to screen the core features, specifically by calculating the feature importance score, and selecting the features with a cumulative contribution rate reaching a set threshold. When dividing the data set, 30% of the samples can be extracted from the icing growth rate data set as the test set by stratified sampling method, and the remaining 70% of the samples are used as the training set.

[0095] Exemplarily, the Gradient Aggregative Optimization Tree algorithm (GAOT model) is initialized: the initialization process of the GAOT model sets the number of trees n estimators , the learning rate , and the maximum tree depth max depth parameters, and defines the mean square error (MSE) as the loss function, and the loss function formula is:

[0096]

[0097] Wherein, MSE is the mean square error, m is the number of training samples, is the predicted value of the model for the i-th sample, y i is the true value of the i-th sample.

[0098] Exemplarily, feature importance calculation: after the model (GAOT model) is trained, the importance score of each microclimate feature can be calculated by the information gain method, and the calculation formula is:

[0099]

[0100] Wherein, Importance(f) is the importance score of feature f, Gain(t,f) represents the information gain of feature f in the t-th decision tree, n estimators is the number of decision trees, and F is the total number of features.

[0101] Exemplarily, feature importance screening: all features are arranged in descending order according to the feature importance score, and features with a cumulative contribution rate reaching a set threshold, for example, 90%, are selected to form a core feature set .

[0102] Exemplarily, data set division: the preprocessed data set (core feature set) is divided into a training set (for example, 70%) and a test set (for example, 30%) by stratified sampling method, 30% of the samples are extracted from the icing growth rate data set as the test set by stratified sampling method, and the remaining 70% of the samples are used as the training set to avoid introducing bias due to data division.

[0103] In an embodiment, the parameter optimization is performed based on the expected improvement convergence framework and the adaptive kernel fuser to determine the optimal parameters of the model; and the training set and the test set are used for training and verification according to the optimal parameters of the model to establish the target ice accretion rate prediction model.

[0104] Specifically, in the embodiment, the expected improvement convergence framework (EICF) is used to optimize the parameters of the adaptive kernel fuser (AKF) with the fitness function as the optimization target, the optimal parameters of the model are determined, the model is trained according to the training set and verified on the test set, and the ice accretion rate prediction model (target ice accretion rate prediction model) is established.

[0105] In an embodiment, the parameter optimization is performed based on the expected improvement convergence framework and the adaptive kernel fuser to determine the optimal parameters of the model, which includes: obtaining a combined kernel function by linearly combining a Gaussian kernel function and a polynomial kernel function; constructing an adaptive kernel fuser based on the combined kernel function; determining a to-be-optimized parameter vector of the adaptive kernel fuser and setting a value range of each to-be-optimized parameter vector through exponential mapping; performing parameter optimization iteration on the to-be-optimized parameter vector based on the expected improvement convergence framework and the value range to obtain an optimal parameter combination; and taking the optimal parameter combination as the optimal parameters of the model.

[0106] In an embodiment, the parameter optimization is performed based on the expected improvement convergence framework and the value range on the to-be-optimized parameter vector to obtain an optimal parameter combination, which includes: optimizing the weight of the combined kernel function and the parameters of each kernel function in the to-be-optimized parameter vector based on the expected improvement convergence framework and the value range; taking the fitness function value obtained through cross-validation as an evaluation standard of the parameters, and obtaining the optimal parameter combination through iterative optimization.

[0107] In an embodiment, the weight of the combined kernel function and the parameters of each kernel function in the to-be-optimized parameter vector are optimized based on the expected improvement convergence framework and the value range, which includes: initializing the expected improvement convergence framework to obtain the optimization architecture of a plurality of sub-populations and randomly generate a plurality of groups of parameter solutions as initial search points in each sub-population; setting a maximum number of iterations; performing parallel search through a plurality of sub-populations based on the initial search points, the value range and the maximum number of iterations; performing a sub-population information interaction operation every preset number of iterations in the iteration process; filtering out sub-populations with a difference degree meeting a preset condition through calculation of the Euclidean distances between the optimal solutions of each sub-population to share information, and transferring the excellent parameter solutions in the sub-populations to other populations.

[0108] Exemplarily, the combination kernel function is constructed: the adaptive kernel fusion machine AKF adopts a combination kernel function, which in this embodiment can be linearly combined by a Gaussian kernel function and a polynomial kernel function. The combination kernel function of the adaptive kernel fusion machine AKF has the following specific formula:

[0109]

[0110] wherein λ is used to adjust the weight proportion of the Gaussian kernel function K G and the polynomial kernel function K P , r g is a bandwidth parameter of the Gaussian kernel function, which determines the locality of the function; r1, r2 and r3 are coefficients and orders of the polynomial kernel function, which jointly affect the nonlinear fitting ability of the function. The combination mode adopted in this embodiment can fully integrate the advantages of the two kernel functions and effectively handle complex nonlinear data relationships.

[0111] Exemplarily, the parameter space is defined: the parameter vector S to be optimized is determined as [λ, r g , r1, r2, r3], and the value range of each parameter is set through exponential mapping:

[0112]

[0113] The value range set in this embodiment not only ensures the comprehensiveness of parameter search, but also avoids unreasonable parameter values in combination with actual experience.

[0114] Specifically, the EICF algorithm (expected improvement convergence framework) is used to optimize the parameters of the adaptive kernel fusion machine AKF, including the weights of the combination kernel function and the parameters of each kernel function. When the EICF algorithm (expected improvement convergence framework) is used to optimize the parameters of the adaptive kernel fusion machine AKF, the fitness function value obtained by cross-validation is used as the evaluation standard for the quality of the parameters, and the optimal parameter combination is obtained through iterative optimization.

[0115] Exemplarily, adaptive optimization iteration: the EICF algorithm is initialized as an optimization framework containing multiple sub-populations, for example, 8 sub-populations. Each sub-population randomly generates multiple sets of parameter solutions, for example, 5 sets of parameter solutions, as initial search points. The maximum number of iterations can be set to 500. This embodiment expands the search space through parallel search of multiple sub-populations, thereby improving the probability of finding the global optimal solution. During the iteration process, the sub-population information exchange operation can be performed every 20 iterations. By calculating the Euclidean distance between the optimal solutions of each sub-population, the sub-populations with larger differences are selected for information sharing, and the excellent parameter solutions in the sub-populations are transmitted to other populations, thereby breaking the local optimal limit and enhancing the global search ability of the algorithm.

[0116] Exemplarily, the root mean square error (RMSE) obtained by 10-fold cross-validation is taken as the fitness function for evaluating the pros and cons of each set of parameter solutions, and the specific formula is:

[0117]

[0118] wherein, m j is the number of samples in the jth-fold cross-validation, y j,i is the actual ice accretion rate of the ith sample in the jth-fold, y' j,i is the predicted value of the corresponding sample. This function can more reliably evaluate the generalization ability of the model by verifying multiple times by dividing the data set.

[0119] In each iteration process, the parameter solutions in each sub-population are dynamically updated according to the formula. The specific formula is:

[0120]

[0121] wherein, q1, q2 are random numbers subject to uniform distribution U(0, 1), S best and S worst represent the current global optimal solution and the worst solution respectively. This strategy guides the parameter solutions to search in a better area by learning from the optimal solution and moving away from the worst solution.

[0122] Exemplarily, the convergence criterion is: continuously monitor the algorithm iteration process, when the change rate of fitness value in 10 consecutive iterations is less than 0.001, it is determined that the algorithm reaches the convergence state, the iteration is stopped and the optimal parameter combination S * =[λ * ,r g * ,r1 * ,r2 * ,r3 * ] is output, which is used to construct the final high-precision ice accretion rate prediction model (target ice accretion rate prediction model).

[0123] In an embodiment, based on the target ice accretion rate prediction model and the initial ice thickness, the ice thickness at future time is calculated in stages, including: dividing the future prediction time length into multiple equal interval stages; based on each equal interval stage, obtaining a predicted growth rate function according to the target ice accretion rate prediction model and the current microclimate data; according to the initial ice thickness and the predicted growth rate function, performing iterative calculation of ice thickness by integral operation to obtain ice thickness prediction values at different stages.

[0124] Specifically, the icing growth rate is combined with the diachronic cumulative effect and the initial icing thickness to calculate the icing thickness at future time in stages. In the calculation of the icing thickness in stages, the prediction time length is divided into multiple stages, the core microclimate characteristics of each stage are input to predict the icing growth rate, and the icing thickness of each stage is calculated in combination with the initial icing thickness. In the calculation in stages, the feature weight is dynamically adjusted according to different stages of icing.

[0125] For example, the prediction stage is divided into N equal-interval stages, and the time length ΔT of each stage can be calculated by the formula ΔT=T / N. The time node t of each stage is calculated by the formula t=t0+kΔT, where k is the stage number. k The formula is:

[0126]

[0127] Wherein, t k is the starting time point of the kth prediction stage; t0 is the prediction starting time; k is the stage number, and the value range is from 1 to N; ΔT is the time length of each prediction stage; T is the total prediction time length; and N is the total number of prediction stages.

[0128] For example, the icing thickness is iteratively calculated by combining the initial icing thickness H(t0) and the growth rate function v(τ) predicted by the adaptive kernel fusion machine (AKF), and the iterative calculation of the icing thickness is realized by integral operation, and the formula is:

[0129]

[0130] Wherein, H(t) is the icing thickness prediction value at the target time t; H(t0) is the initial icing thickness at the initial time t0; v(τ) is the icing growth rate predicted by the AKF model at time τ, and τ is the integral variable, and the value range is from the initial time t0 to the target time t.

[0131] In the actual calculation process, the trapezoidal integral method is used to approximately calculate the integral, and Δτ=10 minutes is selected as the time step, which effectively controls the calculation complexity while ensuring the calculation accuracy.

[0132] For example, the weight is adjusted in stages. The three-stage state recognition is determined by the temperature T and the growth rate v, and the specific logic is: when the environmental temperature satisfies T≤0°C and the icing growth rate v>0, it is determined as the growth stage; if the absolute value |v| of the growth rate is ≤0.05 mm / h, it is the stable stage; when T>0°C and v<0, it is determined as the melting stage, and the mathematical expression is:

[0133]

[0134] The weight adjustment function formula is:

[0135]

[0136] wherein ω i (t) is the weight of the ith feature at time t; is the basic weight of the ith feature, and δ(s(t)) is an adjustment factor related to the icing stage state s(t) at time t, which changes according to the different stages of icing and is used to adjust the basic weight to adapt to the difference in the influence degree of different stages of features on the icing thickness. δ(s) is set according to the difference in stage characteristics, and the formula is:

[0137]

[0138] The mechanism of this embodiment realizes adaptive fitting of different icing stages through formulaic stage recognition and weight correction.

[0139] The method proposed in this embodiment has the following technical effects:

[0140] 1. Efficient screening of core features based on GAOT: The gradient aggregation optimization tree (GAOT) algorithm is used to evaluate the feature importance of microclimate data, capture the correlation between temperature, humidity, wind speed and icing growth, and automatically select core features with a cumulative contribution rate of ≥90%. This solves the problem of low efficiency and redundant features in traditional feature engineering that interfere with prediction accuracy.

[0141] 2. EICF improves the generalization ability of AKF model: The expected improvement convergence framework (EICF) is used to globally optimize the kernel parameters λ, r g , r1, r2, r3 of the adaptive kernel fusion machine (AKF), avoiding the blindness of traditional grid search and realizing high-precision fitting of the AKF model to the icing growth rate.

[0142] 3. Dynamic prediction of the duration of the cumulative effect: The historical icing thickness and the real-time predicted growth rate are iteratively calculated to build an icing thickness prediction model that considers the duration of the cumulative effect, supporting differential prediction of the icing growth, stable, and melting stages. By inputting the initial thickness and microclimate data of different stages, the dynamic process can be accurately described.

[0143] 4. Prediction performance optimization of multi-algorithm fusion: Through the three-layer architecture design of GAOT feature selection, EICF optimization of AKF parameters, and duration of cumulative effect fusion, a collaborative mechanism of "feature dimension reduction-model optimization-dynamic accumulation" is formed, effectively solving the problems of insufficient description of the dynamic process of icing, low parameter optimization efficiency, and poor adaptability of multi-stage prediction of existing models, and improving the engineering practicability of the icing thickness prediction of the transmission line.

[0144] For example Figure 4The ice growth stage prediction contrast chart shown in the figure, the coincidence degree of the algorithm model prediction value and the real value of the ice thickness proposed in the embodiment of the present application is the highest; as Figure 5 The ice stable stage prediction contrast chart shown in the figure, the coincidence degree of the algorithm model prediction value and the real value of the ice thickness proposed in the embodiment of the present application is the highest; as Figure 6 The ice melting stage prediction contrast chart shown in the figure, the coincidence degree of the algorithm model prediction value and the real value of the ice thickness proposed in the embodiment of the present application is the highest. The embodiment realizes the differential prediction of the ice growth, stable and melting stages, and has high precision.

[0145] In the embodiment, the gradient aggregation optimization tree (GAOT) algorithm is used to evaluate the feature importance of the microclimate data to filter out the core features, the expectation improvement convergence framework (EICF) is used to globally optimize the kernel parameters λ, r g , r1, r2, r3 of the adaptive kernel fusion machine (AKF) to improve the model generalization ability, and the historical ice thickness and the real-time predicted growth rate are combined to iteratively accumulate and calculate to consider the diachronic cumulative effect, so as to realize the differential prediction of the ice growth, stable and melting stages. Through this series of methods, a collaborative mechanism of "feature dimension reduction-model optimization-dynamic accumulation" is constructed, effectively solving the problems of insufficient description of the dynamic process of the existing model, low parameter optimization efficiency, poor adaptability of multi-stage prediction, and improving the engineering practicability of the ice thickness prediction of the power transmission line.

[0146] In addition, the embodiment of the present application also proposes a storage medium, and the storage medium stores a power transmission line ice thickness staged prediction program. When the power transmission line ice thickness staged prediction program is executed by a processor, the steps of the power transmission line ice thickness staged prediction method described above are realized.

[0147] Referring to Figure 7 , Figure 7 is a structural block diagram of an embodiment of the power transmission line ice thickness staged prediction system of the present application.

[0148] As Figure 7 shown, the power transmission line ice thickness staged prediction system comprises:

[0149] The data processing module 10 is configured to obtain initial ice thickness data and future microclimate data, and pre-process the future microclimate data to obtain an ice growth rate data set;

[0150] The data filtering module 20 is configured to filter the ice growth rate data set based on the gradient aggregation optimization tree algorithm to obtain a training set and a test set;

[0151] The parameter optimization module 30 is configured to perform parameter optimization based on the expected improvement convergence framework and the adaptive kernel fusion machine to determine optimal parameters of the model.

[0152] The model training module 40 is configured to perform training verification according to the optimal parameters of the model and the training set and the test set to establish a target ice accretion rate prediction model.

[0153] The ice prediction module 50 is configured to calculate the ice thickness at future time points in stages based on the target ice accretion rate prediction model and the initial ice thickness.

[0154] Specifically, the initial ice thickness and microclimate data are collected by a multi-source sensor, the core features are screened by using a gradient aggregation optimization tree (GAOT) after preprocessing, the kernel parameters of the adaptive kernel fusion machine (AKF) are optimized by combining an expected convergence improvement framework (EICF), an ice accretion rate prediction model is constructed, and the thickness at future time periods is predicted in stages based on the diachronic cumulative effect. The core of the technical solution of the embodiment is that the feature importance evaluation mechanism of the GAOT is used to improve the microclimate feature screening efficiency, the global search problem of the AKF parameter optimization is solved by the EICF, and the diachronic cumulative effect is used to iteratively calculate the ice thickness to adapt to the dynamic changes in different stages. The technical solution of the embodiment breaks through the technical bottleneck that the existing model is insufficient to describe the dynamic process of ice, realizes differential prediction of the ice growth, stable, and melting stages, supports engineering application of multi-voltage grade lines and complex terrain scenes, and provides an efficient solution for precise early warning and operation optimization of power grid ice disasters.

[0155] The embodiment proposes a transmission line ice thickness stage-by-stage prediction system. In the system, the gradient aggregation optimization tree (GAOT) algorithm is used to evaluate the feature importance of the microclimate data to screen out core features, the expected convergence improvement framework (EICF) is used to globally optimize the kernel parameters λ, r g , r1, r2, and r3 of the adaptive kernel fusion machine (AKF) to improve the model generalization ability, the historical ice thickness and the iteratively accumulated growth rate are combined to consider the diachronic cumulative effect, thereby realizing stage-by-stage differential prediction of the ice growth, stable, and melting stages. Through the series of methods, a collaborative mechanism of “feature dimension reduction-model optimization-dynamic accumulation” is constructed, and the problems of the existing model, such as insufficient description of the dynamic process of ice, low parameter optimization efficiency, and poor adaptability of multi-stage prediction, are effectively solved, and the engineering practicability of the transmission line ice thickness prediction is improved.

[0156] It should be noted that technical details not described in detail in the transmission line ice thickness stage-by-stage prediction system embodiment can be referred to the transmission line ice thickness stage-by-stage prediction method provided in any embodiment of the application as described above, and will not be described here.

[0157] It should be understood that the above is only illustrative, and does not constitute any limitation on the technical solutions of the present application. In specific applications, those skilled in the art can set up according to the needs, and the present application does not limit this.

[0158] It should be noted that the above-described workflow is only illustrative and does not limit the scope of protection of the present application. In actual application, those skilled in the art can select part or all of them to achieve the purpose of the embodiment scheme according to the actual needs, which is not limited here.

[0159] In addition, it should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the sentence "includes a" does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0160] The above embodiment numbers of the present application are only for description, not representing the advantages and disadvantages of the embodiments.

[0161] From the above description of the embodiments, those skilled in the art can clearly understand that the above embodiment methods can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of software product, which is stored in a storage medium (such as read only memory (Read Only Memory, ROM) / RAM, magnetic disk, optical disk), including a number of instructions to make a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the method described in each embodiment of the present application.

[0162] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A method for predicting ice thickness of transmission lines in stages, characterized in that: include: Acquiring initial ice thickness data and future micrometeorological data, and preprocessing the future micrometeorological data to obtain an ice growth rate data set; The ice growth rate dataset is screened based on the gradient aggregation optimized tree algorithm to obtain a training set and a test set; Optimize the parameters based on the expected improvement convergence framework and adaptive nuclear fusion engine to determine the optimal parameters of the model; Conducting training and verification based on the optimal parameters of the model and the training and test sets to establish a target ice cover growth rate prediction model; Based on the target ice cover growth rate prediction model and the initial ice cover thickness, the ice cover thickness at a future moment is calculated in stages.

2. The method according to claim 1, wherein The obtaining of initial ice thickness data and future micrometeorological data, and preprocessing of the future micrometeorological data to obtain an ice growth rate data set includes: Obtain initial ice thickness data and future micrometeorological data; Preprocessing the future micrometeorological data based on the physical conditions of ice cover to remove abnormal data and obtain processed equivalent ice cover thickness data; Performing differential processing on the processed equivalent ice thickness data to obtain an average ice growth rate; Taking the average of the micrometeorological data of the time period corresponding to the average ice cover growth rate as the micrometeorological input feature corresponding to the time period to generate an initial data set; The data of the initial data set are normalized to generate an ice cover growth rate data set.

3. The method according to claim 1, wherein The ice growth rate dataset is screened based on the gradient aggregation optimization tree algorithm to obtain a training set and a test set, including: Calculate the importance score of each micrometeorological feature of the ice growth rate dataset based on the gradient aggregation optimized tree algorithm; Arrange the micrometeorological features in descending order according to the importance scores, and select features whose cumulative contribution rates reach a set threshold to form a core feature set; The core feature set is divided into a training set and a test set according to a stratified sampling method.

4. The method according to claim 1, wherein The parameter optimization based on the expected improvement convergence framework and the adaptive nuclear fusion machine to determine the optimal parameters of the model includes: The combined kernel function is obtained by linearly combining the Gaussian kernel function and the polynomial kernel function; Constructing an adaptive kernel fusion machine based on the combined kernel function; Determining the parameter vectors to be optimized of the adaptive nuclear fusion machine, and setting the value range of each parameter vector to be optimized by exponential mapping; Performing parameter optimization iteration on the parameter vector to be optimized based on the expected improvement convergence framework and the value range to obtain the optimal parameter combination; The optimal parameter combination is used as the optimal parameter of the model.

5. The method according to claim 4, wherein The performing parameter optimization iteration on the parameter vector to be optimized based on the expected improvement convergence framework and the value range to obtain the optimal parameter combination includes: Optimizing the weight of the combined kernel function and the parameters of each kernel function in the parameter vector to be optimized based on the expected improved convergence framework and the value range; The fitness function value obtained by cross-validation is used as the evaluation standard for the quality of parameters, and the optimal parameter combination is obtained through iterative optimization.

6. The method according to claim 5, wherein The optimizing the weight of the combined kernel function and the parameters of each kernel function in the parameter vector to be optimized based on the expected improved convergence framework and the value range includes: Initialize the expected improvement convergence framework to obtain the optimization architecture of multiple subpopulations and randomly generate multiple sets of parameter solutions for each subpopulation as the initial search points; Set the maximum number of iterations; Based on the initial search point, the value range and the maximum number of iterations, a parallel search is performed through multiple subpopulations; During the iteration process, the subpopulation information interaction operation is performed once every preset number of iterations; By calculating the Euclidean distance between the optimal solutions of each subpopulation, the subpopulations whose differences meet the preset conditions are screened out for information sharing, and the excellent parameter solutions in the subpopulations are transferred to other populations.

7. The method according to any one of claims 1 to 6, characterized in that The step of calculating the ice thickness at a future moment in stages based on the target ice growth rate prediction model and the initial ice thickness includes: Divide the future forecast duration into multiple equally spaced stages; Based on each equally spaced stage, a predicted growth rate function is obtained according to the target ice cover growth rate prediction model and current micrometeorological data; According to the initial ice thickness and the predicted growth rate function, the ice thickness is iteratively calculated through integral operation to obtain predicted values ​​of ice thickness at different stages.

8. A system for predicting ice thickness of transmission lines in stages, characterized in that: include: A data processing module is used to obtain initial ice thickness data and future micro-meteorological data, and pre-process the future micro-meteorological data to obtain an ice growth rate data set; A data screening module, configured to screen the ice growth rate dataset based on a gradient aggregation optimized tree algorithm to obtain a training set and a test set; Parameter optimization module, which is used to optimize parameters based on the expected improvement convergence framework and adaptive nuclear fusion engine to determine the optimal parameters of the model; A model training module is used to perform training and verification based on the optimal parameters of the model and the training set and test set to establish a target ice cover growth rate prediction model; The ice cover prediction module is used to calculate the ice cover thickness at a future moment in stages based on the target ice cover growth rate prediction model and the initial ice cover thickness.

9. An electronic device, characterized in that: The electronic device includes: a memory, a processor, and a transmission line ice thickness stage-by-stage prediction program stored in the memory and executable on the processor, wherein the transmission line ice thickness stage-by-stage prediction program is configured to implement the transmission line ice thickness stage-by-stage prediction method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a program for predicting ice thickness of a transmission line in stages, and the program for predicting ice thickness of a transmission line in stages is used to enable a processor to implement the method for predicting ice thickness of a transmission line in stages according to any one of claims 1 to 7 when executed.

Images