Icing condition prediction method and device, electronic equipment and storage medium

Abstract

The invention provides an icing condition prediction method and device, electronic equipment and a storage medium. An accurate prediction model is obtained through pre-training in an incremental learning mode, after environment data at the current moment and icing data at historical moments are obtained, the accurate icing data at the current moment are obtained according to the prediction model, and then the icing data at the current moment serve as the icing condition to be output. Therefore, a corresponding deicing strategy can be executed subsequently according to the icing condition. Under the condition that the sample size of a new sample is small, an accurate prediction model can be obtained through training, and then the accurate icing condition is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-icing of power transmission lines, and in particular to an icing situation prediction method, device, electronic equipment and storage medium. Background Art

[0002] Icing is a widespread natural phenomenon. However, for transmission lines, severe icing may lead to failures and even catastrophic accidents such as large-scale power outages. At present, as more and more AC and DC UHV transmission lines pass through icing areas, icing has become one of the main factors restricting the safe and stable operation of power systems. Therefore, how to accurately predict the icing of transmission lines has become one of the technical problems that need to be solved urgently.

[0003] In general, the existing technology usually uses historical icing data to train a prediction model. After the prediction model training is completed, the icing condition of the transmission line is predicted based on temperature, humidity, wind direction and other data, so as to execute the corresponding deicing strategy according to the predicted icing condition.

[0004] However, during the use of the model, you may encounter the problem of insufficient accuracy of the trained prediction model. The usual practice is to re-acquire more training samples and use them to retrain in order to obtain a more accurate prediction model. However, in practical applications, the collection of new data may be limited by factors such as cost, time, and availability, resulting in a limited amount of new samples. At the same time, retraining the entire model requires a lot of computing resources, and frequent training of prediction models will result in a waste of computing resources. Summary of the invention

[0005] In view of this, the embodiments of the present invention are committed to providing an ice cover prediction method, device, electronic device and storage medium, which use a prediction model trained by incremental learning to solve the problem in the prior art that it takes a long time to collect samples when training a new model, and a large amount of computing resources are required to train an accurate prediction model.

[0006] In one aspect, the present invention provides a method for predicting icing conditions of a power transmission line, the method comprising:

[0007] Determine the ice coverage data of the target line at the historical moment and the environmental data at the current moment as input data;

[0008] Input the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data;

[0009] The icing condition of the target line is determined according to the icing data at the current moment.

[0010] In one embodiment, the prediction model is trained in the following manner:

[0011] Determining the confidence of the prediction model for the input data, and when the confidence is lower than a preset confidence threshold, determining that the input data is a target sample, and determining a label for the target sample;

[0012] A prediction model pre-trained based on training samples and their annotations is obtained, and parameters of the prediction model are adjusted based on the target samples and their annotations.

[0013] In one embodiment, the prediction model pre-trained based on the training samples and their annotations is trained in the following manner:

[0014] For each historical moment, according to the historical ice data of the sample line and the environmental data of the historical ice data, determine the environmental data and ice data of the sample line at the historical moment, and determine the training sample and its annotation according to the environmental data and ice data of the sample line at the historical moment;

[0015] According to each training sample and its annotation, the prediction model to be trained is trained to obtain a trained prediction model.

[0016] In one embodiment, the structure of the prediction model is an extreme learning machine structure;

[0017] According to the training samples and their annotations, the prediction model to be trained is trained, specifically including:

[0018] Determine a kernel function corresponding to the prediction model and a coefficient to be adjusted corresponding to the kernel function;

[0019] Determining a kernel matrix to be adjusted according to the coefficient to be adjusted and the kernel function;

[0020] Inputting each training sample into the prediction model to obtain prediction results of each sample output by the prediction model;

[0021] Taking minimizing the gap between the prediction results of each sample and the annotations of its corresponding training samples as the optimization goal, the matrix parameters in the kernel matrix to be adjusted are adjusted to obtain a target matrix; the target matrix is ​​used to predict ice cover data according to the input data of the prediction model.

[0022] In one embodiment, the prediction model is trained through multiple iterative processes;

[0023] Taking minimizing the gap between the prediction results of each sample and the annotations of the corresponding training samples as the optimization goal, the matrix parameters in the kernel matrix to be adjusted are adjusted to obtain the target matrix, which specifically includes:

[0024] For each iteration process, the matrix parameters of the kernel matrix adjusted in the previous iteration process are determined as the initial matrix parameters;

[0025] The initial matrix parameters are adjusted by using the Grey Wolf Optimization Algorithm to obtain an adjusted kernel matrix;

[0026] Determining the sample prediction results of each training sample in the iterative process according to the adjusted kernel matrix;

[0027] Whether the iteration termination condition is met is determined according to the sample prediction results of each training sample in the iteration process. If so, the target matrix is ​​determined according to the matrix parameters of the kernel matrix adjusted in the iteration process.

[0028] In one embodiment, the prediction model includes multiple classifiers, the weights of the multiple classifiers are not completely the same, and the structure of the classifier is an extreme learning machine structure;

[0029] According to the training samples and their annotations, the prediction model to be trained is trained to obtain a trained prediction model, which specifically includes:

[0030] Inputting each training sample into each classifier included in the prediction model to obtain classification results output by each classifier;

[0031] Determine the weight of each classifier in the current iteration process, and determine the sample prediction results corresponding to each training sample according to the classification results output by each classifier and the weights corresponding to each classifier;

[0032] The prediction model to be trained is trained according to the sample prediction results corresponding to the training samples and the labels of the training samples.

[0033] In one embodiment, the prediction model to be trained is trained according to the sample prediction results corresponding to the training samples and the labels of the training samples, specifically including:

[0034] In each iteration process, for each training sample, determine the weight of the training sample in the iteration process;

[0035] Determining the iteration difference of each training sample in the iteration process according to the weight of each training sample in the iteration process, the sample prediction results corresponding to each training sample, and the label of each training sample;

[0036] The prediction model to be trained is trained by minimizing the difference between the iteration differences corresponding to each iteration process as the optimization goal;

[0037] The method further comprises:

[0038] For each training sample, determine the number of target classification results, where the target classification results are classification results whose difference from the annotation of the training sample is greater than a preset difference threshold;

[0039] The weight of the training sample in the next iteration process is determined according to the number of the target classification results.

[0040] This specification provides an icing situation prediction device, the device comprising:

[0041] An acquisition module is used to determine ice coverage data of a target line at a historical moment and environmental data at a current moment as input data;

[0042] A prediction module, used for inputting the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data;

[0043] The determination module is used to determine the ice coverage of the target line according to the ice coverage data at the current moment.

[0044] The present specification provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned ice cover prediction method is implemented.

[0045] The present specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned ice cover prediction method when executing the program.

[0046] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0047] In the ice cover prediction method provided in this specification, an accurate prediction model is pre-trained by using incremental learning. After obtaining the current environmental data and the ice cover data at the historical time, the accurate ice cover data at the current time is obtained according to the prediction model, and then the accurate ice cover situation is obtained according to the ice cover data at the current time, so as to perform the corresponding deicing strategy according to the ice cover situation. Even when the number of new samples is small, an accurate prediction model can be trained to obtain the accurate ice cover situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The figure shows a flow chart of the ice cover prediction method provided in this specification;

[0049] Figure 2 The figure shows a schematic diagram of the structure of the ice cover prediction device provided in this specification;

[0050] Figure 3 The corresponding Figure 1 electronic equipment. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] It should be noted that all actions to obtain signals, information or data in this manual are performed in compliance with the relevant local data protection laws and policies and with the authorization given by the owner of the corresponding device.

[0053] An embodiment of the present invention provides a method for predicting the icing condition of a power transmission line. An accurate prediction model is obtained by pre-training using an incremental learning method. After obtaining environmental data at the current moment and icing data at historical moments, accurate icing data at the current moment is obtained according to the prediction model, and then the icing data at the current moment is output as the icing condition, so as to facilitate the subsequent execution of a corresponding deicing strategy according to the icing condition. Even when the number of new samples is small, an accurate prediction model can be trained to obtain an accurate icing condition.

[0054] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0055] Figure 1 The following is a flow chart of an ice cover prediction method in this specification, which specifically includes the following steps:

[0056] S100: Determine ice coverage data of a target line at a historical moment and environmental data at a current moment.

[0057] In one or more embodiments provided in this specification, the ice cover prediction method can be applied to any electronic device, which can be a server or a smart terminal. For the convenience of description, the following description will take the execution process of the ice cover prediction method executed by the server as an example.

[0058] Generally, the ice thickness of a transmission line at the current moment is usually affected by the ice thickness at the historical moment and the environmental data at the current moment. There are usually multiple transmission lines, and the environmental data of each transmission line in different line sections is different. Obviously, the corresponding ice data is not completely the same. Therefore, the server can first determine the target line, and determine the environmental data of the target line and the ice data at the historical moment, so as to predict the ice data of the target line at the current moment based on the acquired environmental data and ice data.

[0059] Specifically, the server may receive the line identification of the target line and the environmental data corresponding to the line identification sent by the collection device, and parse the environmental data to obtain the specific content corresponding to the environmental data. The environmental data may be at least one of temperature, humidity, wind direction and wind speed. The collection device may be a sensor arranged around the target line, or a device that obtains environmental data from channels such as weather forecasts.

[0060] Then, the server can determine the ice coverage data of the target line at a historical moment from the ice coverage data of the historical moments corresponding to each line stored in advance according to the line identifier of the target line. The ice coverage data of the target line at a historical moment may only include the ice coverage data of the target line at the previous moment, or may include ice coverage data of several consecutive historical moments. Of course, the ice coverage data of several consecutive historical moments include ice coverage data of the previous moment. The ice coverage data may only include ice coverage thickness, or may include ice coverage thickness and ice coverage thickness growth rate, and the ice coverage thickness may be ice coverage thickness obtained by direct measurement, or may be equivalent ice coverage thickness.

[0061] The specific data types included in the environmental data and the ice cover data and the specific time to which the historical moment corresponds can be set as needed, and this manual does not impose any restrictions on this.

[0062] Finally, the server may store the determined ice coverage data of the target line at historical moments and the environmental data at the current moment as input data.

[0063] S102: Input the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data corresponding to the historical ice cover data.

[0064] In one or more embodiments provided in the present specification, as mentioned above, the inventive point of the ice cover prediction method is to use a prediction model pre-trained based on incremental learning to predict accurate ice cover data at the current moment based on environmental data at the current moment and ice cover data at historical moments.

[0065] Therefore, the server can use the input data as input into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment.

[0066] The specific input data of the prediction model may be a vector composed of the environmental data at the current moment, the ice thickness at the historical moment, and the ice growth rate at the historical moment. The specific output data of the prediction model may be only the ice thickness at the current moment, or it may be a vector composed of the ice thickness at the current moment, the environmental data at the current moment, the ice thickness at the historical moment, and the ice growth rate at the historical moment.

[0067] The prediction model can be trained in the following way:

[0068] First, the server used to train the prediction model can obtain historical ice coverage data of a sample line and historical environmental data corresponding to the historical ice coverage data. The sample line can be the target line, or any transmission line or any line segment of a transmission line.

[0069] Then, for the ice cover data corresponding to each historical moment respectively contained in the historical ice cover data, the server may determine the environmental data corresponding to the ice cover data from the historical environmental data as the environmental data corresponding to the historical moment.

[0070] Afterwards, the server may determine a training sample based on the environmental data corresponding to the historical moment, and use the ice cover data corresponding to the historical moment as a label for the training sample.

[0071] Finally, based on the training samples and their annotations, an accurate prediction model is trained.

[0072] During the use of the prediction model, the server class of the prediction model is used to monitor the prediction results of the prediction model. When the prediction results are different from the actual ice cover conditions, the target samples and their annotations, that is, new samples and their annotations, are determined.

[0073] Therefore, the server for training the model can fine-tune the model parameters of the prediction model according to the target sample and its annotation to obtain an adjusted prediction model. In this way, even when the sample size of the target sample is small, an accurate prediction model can be trained.

[0074] It should be noted that the server for training the prediction model and the server for using the prediction model can be the same server or different servers. The specific training and deployment of the prediction model can be set as needed, and this manual does not impose any restrictions on this.

[0075] S104: Determine the ice coverage of the target line according to the ice coverage data at the current moment.

[0076] In one or more embodiments provided in this specification, as mentioned above, this method is an ice coverage prediction method, and the ice coverage can be the ice coverage thickness of the target line at the current moment, the ice coverage growth rate from the previous moment to the current moment, or the ice coverage distribution of each transmission line determined according to the ice coverage thickness corresponding to each transmission line and each line section of each transmission line. Therefore, after obtaining the ice coverage data at the current moment, the server can also determine the ice coverage of the target line.

[0077] Specifically, the server may determine the ice thickness of the target line at the current moment according to the ice coverage data of the target line at the current moment as the ice coverage condition of the target line.

[0078] Alternatively, the server may determine the ice coverage growth rate of the target line from the last moment to the current moment according to the ice coverage thickness of the target line at the current moment and the ice coverage thickness at the last moment, as the ice coverage condition of the target line.

[0079] Alternatively, the server may preset the environmental data at the current moment as the environmental data for the next moment or for several consecutive moments starting from the current moment, and determine the ice coverage data of the target line at the next moment or for several consecutive moments starting from the current moment in sequence according to the sequence of each moment, and then predict the ice coverage conditions at the next moment or for several consecutive moments starting from the current moment.

[0080] Alternatively, the server may also store the ice thickness and ice growth rate of the target line at the current moment as the ice data at the current moment after determining the ice thickness and ice growth rate of the target line at the current moment, and predict the ice distribution conditions corresponding to each transmission line based on the ice data corresponding to each transmission line at each historical moment and the current moment respectively stored in the server as the ice conditions corresponding to each transmission line.

[0081] That is to say, the ice coverage condition includes at least one of the ice coverage thickness at the current moment, the ice coverage growth rate from the previous moment to the current moment, the ice coverage condition at the next moment or several consecutive moments starting from the current moment, and the ice coverage distribution condition corresponding to each transmission line.

[0082] based on Figure 1 The ice cover prediction method uses incremental learning to pre-train an accurate prediction model. After obtaining the current environmental data and the ice cover data at the historical time, the accurate ice cover data at the current time is obtained according to the prediction model, and then the ice cover data at the current time is output as the ice cover situation, so as to perform the corresponding deicing strategy according to the ice cover situation. Even when the number of new samples is small, an accurate prediction model can be trained to obtain the accurate ice cover situation.

[0083] Furthermore, the above-mentioned method of pre-training to obtain an accurate prediction model and fine-tuning the prediction model when there are target samples is usually seen when the prediction model is a large model. If the prediction model is a large model, its corresponding accuracy is indeed higher, but the requirements for computing power also increase accordingly. Therefore, limited by the requirements for computing power, how to train an accurate prediction model when the computing power is small? This specification provides an answer: setting the structure of the prediction model to an extreme learning machine structure.

[0084] Therefore, the training process of the prediction model can be as follows:

[0085] Specifically, the server for training the model may first determine the kernel function of the extreme learning machine corresponding to the prediction model and the coefficient to be adjusted corresponding to the kernel function. The kernel function may be at least one of a linear kernel function, a polynomial kernel function, a Gaussian radial basis function kernel function, and a sigmoid kernel function.

[0086] Then, the server may determine the kernel matrix to be adjusted according to the kernel function and the coefficients to be adjusted.

[0087] Afterwards, the server may input each training sample as an input into the prediction model to obtain prediction results of each sample output by the prediction model based on the kernel matrix to be adjusted.

[0088] Finally, the server can minimize the gap between the prediction results of each sample and the annotations of the corresponding training samples as the optimization goal, and obtain the adjusted kernel matrix as the target matrix for the matrix parameters in the kernel matrix to be adjusted. Thus, a trained prediction model is obtained. The target matrix is ​​used to predict ice cover data based on the input data of the prediction model.

[0089] The kernel function used by the extreme learning machine in this specification is the polynomial kernel function K poly and Gaussian radial basis function kernel function K rbf For example, since the expression of the polynomial kernel function can usually be The expression of Gaussian radial basis function kernel function can usually be Therefore, the coefficients to be adjusted corresponding to the kernel function of the extreme learning machine in this specification may include the polynomial kernel function K poly The constant term coefficient q and the exponential term coefficient z, the parameter g of the Gaussian radial basis function kernel function, and the kernel weights w1 and w2 corresponding to the two kernel functions. However, since the weights are usually set by means of w1+w2=1, it is sufficient to set the weight of only one kernel function to w and the weight of the other kernel function to 1-w. Among them, x can usually be an input vector, and y can usually be an output vector, that is, the input vector can be converted into an output vector through the kernel function.

[0090] Then, according to the kernel function and the coefficients to be adjusted, the server can determine the kernel matrix.

[0091] Afterwards, the server can input each training sample into the prediction model, and obtain the sample prediction result output by the prediction model through the kernel matrix to be adjusted set in the prediction model.

[0092] Finally, according to the difference between the prediction results of each sample and the annotations of the corresponding training samples, the matrix parameters in the kernel matrix are adjusted to obtain the adjusted kernel matrix. When the iteration termination condition is met, the adjusted kernel matrix is ​​used as the target matrix. In this way, a trained prediction model is obtained.

[0093] Of course, since there are usually errors in the model training process, in order to control the degree of penalty for misclassification during the model training process, the penalty coefficient C can also be adjusted in the above prediction model training procedure.

[0094] Based on the above method, during the model training process, only the parameters to be adjusted that affect the model output results need to be adjusted, avoiding the current situation in which massive parameters need to be adjusted during the model training process, the parameters have poor interpretability and require more computing resources to adjust to an accurate model. This ensures the training efficiency of the model and enables accurate prediction models to be trained in scenarios with limited computing power.

[0095] Furthermore, in the process of adjusting the model parameters, evolutionary algorithms such as the grey wolf algorithm, the whale algorithm, and the mongoose algorithm can also be used to adjust the model parameters.

[0096] Specifically, taking the Gray Wolf Optimization Algorithm as an example, the prediction model in this specification can be obtained by training through multiple iterative processes. The training process of each iterative process can be the following process:

[0097] First, the server for training the model may determine the matrix parameters of the kernel matrix adjusted in the previous iteration process as the initial matrix parameters of the current iteration process.

[0098] Then, the server can adjust the initial matrix parameters through the Grey Wolf Optimization Algorithm to obtain an adjusted kernel matrix.

[0099] Afterwards, the server may determine the sample prediction result of each training sample in the current iteration process according to the adjusted kernel matrix.

[0100] Finally, the server may determine the gap between each sample prediction result and the corresponding annotation of each training sample, and determine whether the iteration termination condition is met based on the gap.

[0101] If satisfied, the server can use the kernel matrix adjusted in the current iteration as the target matrix.

[0102] If not satisfied, the server may proceed to the next iteration.

[0103] Using the gray wolf algorithm to update the matrix parameters usually includes the following steps:

[0104] (1) Initialize the gray wolf population and set parameters such as population size N, maximum number of iterations M, spatial search dimension D, and iteration coefficient a.

[0105] (2) Calculate the fitness value of each gray wolf in the initial population and sort them. The first three individuals with the best fitness value are named α, β, and δ wolves, and the remaining wolves are all w wolves. The initial population includes N sets of solutions of matrix parameters. For each set of solutions, the fitness value corresponding to the set of solutions can be determined by the difference between the sample prediction results output by the prediction model and the annotations of the training samples after the training samples are input into the prediction model whose target matrix is ​​the kernel matrix corresponding to the set of solutions. The smaller the difference, the higher the fitness value.

[0106] (3) For each w wolf in the population, the following formula is used to update the current gray wolf position:

[0107]

[0108] X(t+1)=[X1(t)+X2(t)+X3(t)] / 3

[0109] Among them, X1, X2, and X3 are the step lengths of the w wolf moving towards the α, β, and δ wolf directions respectively. ( t) and X( t +1) is the position of the wolf pack before and after the tth update, A is the encirclement step length, which is used to characterize the range of movement of the w wolves in the search space, and C is the moving direction, which is used to describe the direction in which the w wolves move in the search space. α is the position of wolf α, X β is the position of β wolf, X δ is the position of δ wolf. α is the distance vector between the w wolf and the α wolf, D β is the distance vector between the w wolf and the β wolf, D δ is the distance vector between the w wolf and the δ wolf, and X is the current position of the w wolf.

[0110] (4) Check whether the updated position of the w wolf is out of bounds. If so, re-initialize the position of the w wolf randomly. Calculate the fitness value of each wolf in the wolf pack and update α, β, and δ.

[0111] Among them, for each parameter that needs to be adjusted during the iteration of the prediction model, the boundary corresponding to the parameter can be pre-set, and then the boundary of the solution of the gray wolf search algorithm can be determined. Therefore, after obtaining the updated position of the w wolf, the server can determine whether it has crossed the boundary of the solution of the gray wolf search algorithm based on the position of the w wolf. If it has crossed the boundary, the position of the w wolf is randomly generated within the boundary of the solution of the gray wolf search algorithm.

[0112] (5) Finally, when the evolution termination condition is reached, the server can use the final position of the α wolf as the optimal solution of the matrix parameters.

[0113] In addition, when adjusting the matrix parameters of the core matrix according to the evolutionary algorithm, the server may also use an evolutionary algorithm that improves the grey wolf algorithm to adjust the matrix parameters.

[0114] The matrix parameters are updated using an evolutionary algorithm that improves the gray wolf algorithm, which usually includes the following steps:

[0115] (1) Initialize the gray wolf population and set parameters such as population size N, maximum number of iterations M, spatial search dimension D, and iteration coefficient a.

[0116] (2) Calculate the fitness values ​​of each gray wolf in the initial population and sort them. The top three individuals with the best fitness values ​​are named α, β, and δ wolves, respectively, and the rest of the wolf group is named w wolves.

[0117] (3) Introduce inertia weight coefficient: For each w wolf in the population, the following formula is used to update the current gray wolf position:

[0118]

[0119] At the same time, the Sigmoid function is introduced to adaptively update the population position:

[0120]

[0121] X i (t+1)=X i (t)·φ

[0122] Among them, f i and f ave are the fitness value of the i-th individual in the wolf pack and the average fitness of the wolf pack; θ is the exponential coefficient. (t) and X (t+1) is the position of the wolf pack before and after the tth update. The specific meanings of other characters in the formula are the same as those in the above gray wolf algorithm.

[0123] (4) Check whether the updated position of the w wolf is out of bounds. If so, re-initialize the position of the w wolf randomly. Calculate the fitness value of each wolf in the wolf pack and update α, β, and δ.

[0124] (5) Finally, when the evolution termination condition is reached, the server can use the final position of the α wolf as the optimal solution of the matrix parameters.

[0125] Furthermore, the above description only takes the structure of the prediction model as an extreme learning machine structure as an example, but in fact, the prediction model may include multiple classifiers, and the structure of each classifier is an extreme learning machine structure.

[0126] Therefore, during the training process of the model, for each training sample, the server used for training can use the training sample as input and input it into each classifier included in the prediction model to obtain the classification results output by each classifier.

[0127] Next, the server may determine the weight of each classifier in the current iteration process, and determine the sample prediction result of the training sample according to the classification results respectively output by each classifier and the weights respectively corresponding to each classifier.

[0128] Finally, according to the prediction results of each sample and the annotations of the training samples corresponding to the prediction results of each sample, the server can train the prediction model to be trained.

[0129] The server may preset the weights corresponding to the classifiers, or determine the weight of the classifier in the next iteration according to the accuracy of the classifier in the iteration for each iteration. The weight may be This formula is calculated. m is the weight of the classifier, εm is the prediction error, which can be specifically the sum of the differences between the classification results of each training sample output by the classifier during the iteration and its corresponding label.

[0130] In addition, as the number of iterations continues to increase, the accuracy of the prediction model continues to improve. In the later stages of the iterations, the gap between the sample prediction results and their labels of most training samples may be small, but there are still some training samples with a large gap between the sample prediction results and their labels. In this case, the gaps corresponding to the training samples with a large gap between the sample prediction results and their labels can be reduced, which can enable the model to be optimized more specifically, thereby enabling the accurate model to be trained faster.

[0131] Therefore, the server can also determine the weights corresponding to each training sample.

[0132] Specifically, for each training sample, the server may determine the number of target classification results, wherein the target classification result is a classification result whose difference with the annotation of the training sample is greater than a preset difference threshold.

[0133] Then, the server can determine the weight of the training sample in the next iteration process according to the number of the target classification results. The server can directly use the number as the weight, or calculate it according to the following formula: E m (k) represents the weight of the k-th training sample in the current iteration process, that is, the m-th iteration process, E m+1 (k) represents the weight of the k-th training sample in the next iteration, that is, the m+1-th iteration. m is the sum of weights, ε l is the error limit, ε m is the prediction error. m is the weight of the classifier, which may be a preset value, and may be specifically determined according to the number of classifiers.

[0134] Therefore, for each iteration process, the server can determine the weight of each training sample in this iteration process, and determine the iteration difference in the iteration process based on the weight of each training sample in the current iteration process, the sample prediction results corresponding to each training sample, and the annotations of each training sample.

[0135] Finally, the parameters of the prediction model are adjusted with the optimization goal of minimizing the iteration difference in the specified number of iterations, where the specified number can be 2 or more.

[0136] Furthermore, line icing is conditional. In the process of determining training samples, the server used to train the model can also eliminate environmental data and icing data that do not meet the icing conditions based on the icing conditions.

[0137] Specifically, the conditions for line icing may generally include that the surface temperature of the overhead line is as low as 0°C or below, the relative humidity of the environment is 85% or above, and the wind speed of the environment is not greater than 1m / s. The abnormal data in the icing data set are removed according to the icing conditions.

[0138] Therefore, the server can determine whether the environmental data of the sample line at the historical moment meets the icing condition, that is, if the environmental data meets at least one of the following conditions, it will be removed. The following conditions can be specifically: the surface temperature of the overhead line is greater than 0°C, the relative humidity of the environment is 85% or less, and the wind speed of the environment is greater than 1m / s.

[0139] Based on this, the server can determine the training samples that meet the icing conditions.

[0140] Furthermore, for the sample line, there may be a situation where the environmental data and ice cover data at a certain moment are not saved. Therefore, the server can also use the k-neighbor filling algorithm to construct the environmental data and ice cover data at a certain moment from the data at several relatively close moments. Since the k-neighbor filling algorithm is already a relatively mature algorithm, this specification will not be repeated here.

[0141] In addition, the model training process requires a lot of computing power. If the dimensions of the model input data are inconsistent and the gap is large, a high computing power is required to train an accurate model. In order to further reduce the computing power requirements of model training, the server can also normalize the data before inputting it into the prediction model. The commonly used normalization formula is as follows:

[0142]

[0143] Among them, s is the standard deviation, x′ is the mean of the sample data, and x i is the original data value of the i-th training sample, is the normalized result.

[0144] Furthermore, the model training process in this specification can be guided by a fitness function, and the prediction model in this specification can be trained by using a K-fold training set.

[0145] Furthermore, in the description of the confidence level of the prediction results output by the prediction model, the difference between the prediction results output by the model and the actual ice cover corresponding to the input data is used. However, in the actual use of the model, the confidence level can also be determined based on whether the input data meets the KKT (Karush-Kuhn-Tucker) condition of the prediction model.

[0146] Specifically, the server may determine whether each input data satisfies the KKT condition, and if so, the confidence of the input data is considered to be high. If not, the confidence of the input data is considered to be lower than a preset confidence threshold.

[0147] In addition, the incremental learning of the prediction model trained based on the above structure can be the following process.

[0148] Specifically, at time t, the prediction model satisfies H(t) = K t +C ―1 I, where H(t) is the output matrix, K t is the target matrix, C and I are the penalty coefficient and the unit matrix respectively, and a new sample (x t ,y t ), then H(t+1)=K t+1 +C ―1 I=

[0149]

[0150] Among them, H(t) is the output matrix at time t, K(t,x t+1 ) is the result calculated by the kernel function, which can characterize the similarity between the target sample and the training sample used to train the prediction model. t+1 ,t) and K(t,x t+1 ), which can also be used to characterize the similarity between the target sample and the training sample used to train the prediction model. t+1 ,x t+1 ) is the prediction result of the target sample determined by the kernel function, which can be used to characterize the characteristics and self-similarity of the target sample in the kernel space. ―1 Characterize the penalty term and control the impact of new samples on the model.

[0151] Based on the above formula, the target matrix can be adapted to the data distribution of new samples without retraining the target matrix. In this way, a prediction model trained by incremental learning can be obtained.

[0152] It should be noted that the above process of updating the kernel matrix by incremental learning, when the prediction model only includes one classifier with an extreme learning machine structure, can actually be a process of updating the kernel matrix of the classifier. When the prediction model includes multiple classifiers with extreme learning machine structures, it can be a process of updating the kernel matrix of each classifier. The server can update the kernel matrix of each classifier separately for each classifier, or it can simultaneously update the kernel matrices of all classifiers based on the sample prediction results determined by the classification results output by each classifier and the gap between the labels of the target samples.

[0153] The above is an icing condition prediction method provided by one or more embodiments of this specification. Based on the same idea, this specification also provides an icing condition prediction device, such as Figure 2 shown.

[0154] Figure 2 A schematic diagram of a model training device provided in this specification, wherein:

[0155] An acquisition module 200 is used to determine ice coverage data of a target line at a historical moment and environmental data at a current moment as input data;

[0156] The prediction module 202 is used to input the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data;

[0157] The determination module 204 is used to determine the ice coverage of the target line according to the ice coverage data at the current moment.

[0158] The apparatus further comprises: a training module 206;

[0159] The training module 206 is trained to obtain the prediction model in the following manner: determining the confidence of the prediction model for the input data, and when the confidence is lower than a preset confidence threshold, determining that the input data is a target sample, and determining the label of the target sample, obtaining the prediction model pre-trained based on the training samples and their labels, and adjusting the parameters of the prediction model based on the target samples and their labels, and adjusting the parameters of the prediction model based on the target samples and their labels.

[0160] Optionally, the training module 206 adopts the following method to pre-train the prediction model based on the training samples and their annotations: for each historical moment, the environmental data and ice data of the sample line at the historical moment are determined according to the historical ice data of the sample line and the environmental data of the historical ice data, and each training sample and its annotation are determined according to the environmental data and ice data of the sample line at the historical moment, and the prediction model to be trained is trained according to the training samples and their annotations to obtain a trained prediction model.

[0161] Optionally, the training module 206 is used to determine the kernel function corresponding to the prediction model and the coefficients to be adjusted corresponding to the kernel function, determine the kernel matrix to be adjusted according to the coefficients to be adjusted and the kernel function, input the training samples into the prediction model to obtain the prediction results of the samples output by the prediction model, and take minimizing the gap between the prediction results of the samples and the labels of the corresponding training samples as the optimization goal, adjust the matrix parameters in the kernel matrix to be adjusted to obtain the target matrix; the target matrix is ​​used to predict the ice cover data according to the input data of the prediction model, wherein the structure of the prediction model is an extreme learning machine structure.

[0162] Optionally, the training module 206 is used to, for each iterative process, use the matrix parameters of the kernel matrix adjusted in the previous iterative process as the initial matrix parameters, adjust the initial matrix parameters through the Gray Wolf Optimization Algorithm to obtain an adjusted kernel matrix, determine the sample prediction results of each training sample in the iterative process according to the adjusted kernel matrix, and judge whether the iteration termination condition is met according to the sample prediction results of each training sample in the iterative process; if so, determine the target matrix according to the matrix parameters of the kernel matrix adjusted in the iterative process, wherein the prediction model is obtained through training of multiple iterative processes.

[0163] Optionally, the training module 206 is used to input the training samples into the classifiers included in the prediction model to obtain classification results output by the classifiers.

[0164] Determine the weight of each classifier in the current iteration process, and determine the sample prediction results corresponding to each training sample according to the classification results output by each classifier and the weights corresponding to each classifier;

[0165] The prediction model to be trained is trained according to the sample prediction results corresponding to each training sample and the annotations of each training sample, wherein the prediction model includes multiple classifiers, the weights of the multiple classifiers are not completely the same, and the structure of the classifier is an extreme learning machine structure.

[0166] Optionally, the training module 206 is used to determine, in each iteration process, for each training sample, the weight of the training sample in the iteration process, determine the iteration difference of each training sample in the iteration process according to the weight of each training sample in the iteration process, the sample prediction results corresponding to each training sample, and the annotation of each training sample, and train the prediction model to be trained with minimizing the difference between the iteration differences corresponding to each iteration process as the optimization goal.

[0167] Optionally, the training module 206 is also used to determine the number of target classification results for each training sample, the target classification result being a classification result whose difference with the annotation of the training sample is greater than a preset difference threshold, and the weight of the training sample in the next iteration process is determined based on the number of target classification results.

[0168] The specification also provides a computer-readable storage medium, which stores a computer program, and the computer program can be used to execute the above-mentioned ice condition prediction method.

[0169] This manual also provides Figure 3 The schematic structure diagram of the electronic device shown in FIG. Figure 3 As described above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and of course may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the above-mentioned ice cover prediction method. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0170] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0171] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0172] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0173] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0174] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0175] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0176] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0177] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0178] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0179] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0180] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0181] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0182] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0183] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing nodes connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage nodes.

[0184] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0185] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A method for predicting ice conditions, characterized in that: include: Determine the ice coverage data of the target line at the historical moment and the environmental data at the current moment as input data; Input the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data; The icing condition of the target line is determined according to the icing data at the current moment.

2. The ice cover prediction method according to claim 1, characterized in that: The prediction model is trained in the following way: Determining the confidence of the prediction model for the input data, and when the confidence is lower than a preset confidence threshold, determining that the input data is a target sample, and determining a label for the target sample; A prediction model pre-trained based on training samples and their annotations is obtained, and parameters of the prediction model are adjusted based on the target samples and their annotations to obtain a prediction model trained using incremental learning.

3. The ice cover prediction method according to claim 2, characterized in that: The prediction model pre-trained based on the training samples and their annotations is trained in the following way: For each historical moment, according to the historical ice data of the sample line and the environmental data of the historical ice data, determine the environmental data and ice data of the sample line at the historical moment, and determine the training sample and its annotation according to the environmental data and ice data of the sample line at the historical moment; According to each training sample and its annotation, the prediction model to be trained is trained to obtain a trained prediction model.

4. The ice cover prediction method according to claim 3, characterized in that: The structure of the prediction model is an extreme learning machine structure; According to the training samples and their annotations, the prediction model to be trained is trained, specifically including: Determine a kernel function corresponding to the prediction model and a coefficient to be adjusted corresponding to the kernel function; Determining a kernel matrix to be adjusted according to the coefficient to be adjusted and the kernel function; Inputting each training sample into the prediction model to obtain prediction results of each sample output by the prediction model; Taking minimizing the gap between the prediction results of each sample and the annotations of its corresponding training samples as the optimization goal, the matrix parameters in the kernel matrix to be adjusted are adjusted to obtain a target matrix; the target matrix is ​​used to predict ice cover data based on the input data of the prediction model.

5. The ice cover prediction method according to claim 4, characterized in that: The prediction model is trained through multiple iterative processes; Taking minimizing the gap between the prediction results of each sample and the annotations of the corresponding training samples as the optimization goal, the matrix parameters in the kernel matrix to be adjusted are adjusted to obtain the target matrix, which specifically includes: For each iteration process, the matrix parameters of the kernel matrix adjusted in the previous iteration process are determined as the initial matrix parameters; The initial matrix parameters are adjusted by using the Grey Wolf Optimization Algorithm to obtain an adjusted kernel matrix; Determining the sample prediction results of each training sample in the iterative process according to the adjusted kernel matrix; Whether the iteration termination condition is met is determined according to the sample prediction results of each training sample in the iteration process. If so, the target matrix is ​​determined according to the matrix parameters of the kernel matrix adjusted in the iteration process.

6. The ice cover prediction method according to claim 3, characterized in that: The prediction model includes multiple classifiers, the weights of the multiple classifiers are not completely the same, and the structure of the classifier is an extreme learning machine structure; According to each training sample and its annotation, the prediction model to be trained is trained, including: Input each training sample into each classifier included in the prediction model to obtain the classification results output by each classifier; Determine the weight of each classifier in the current iteration process, and determine the sample prediction results corresponding to each training sample according to the classification results output by each classifier and the weights corresponding to each classifier; The prediction model to be trained is trained according to the sample prediction results corresponding to the training samples and the labels of the training samples.

7. The ice cover prediction method according to claim 6, characterized in that: According to the sample prediction results corresponding to the training samples and the labels of the training samples, the prediction model to be trained is trained, specifically including: In each iteration process, for each training sample, determine the weight of the training sample in the iteration process; Determining the iteration difference of each training sample in the iteration process according to the weight of each training sample in the iteration process, the sample prediction results corresponding to each training sample, and the label of each training sample; The prediction model to be trained is trained by minimizing the difference between the iteration differences corresponding to each iteration process as the optimization goal; The method further comprises: For each training sample, determine the number of target classification results, where the target classification results are classification results whose difference from the annotation of the training sample is greater than a preset difference threshold; The weight of the training sample in the next iteration process is determined according to the number of the target classification results.

8. An icing condition prediction device, characterized in that: include: An acquisition module is used to determine ice coverage data of a target line at a historical moment and environmental data at a current moment as input data; A prediction module, used for inputting the input data into a pre-trained prediction model to obtain a prediction result output by the prediction model as ice cover data at the current moment; wherein the prediction model is pre-trained by incremental learning based on historical ice cover data and historical environmental data; The determination module is used to determine the ice coverage of the target line according to the ice coverage data at the current moment.

9. A computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

Images