A feature selection method for power transmission line parameter prediction and related devices

Abstract

The application provides a feature selection method for power transmission line parameter prediction, and a corresponding feature evaluation neural network model is designed, which comprises an input layer, a pooling layer, a full connection layer, an attention layer and a prediction model layer, so that the feature evaluation and the prediction model are combined, the relative importance of each feature is determined by obtaining the attention weight corresponding to the features and labels of each power transmission line operation condition, feature screening based on the attention mechanism is realized, the most suitable power transmission line condition feature combination can be selected according to different power transmission line data, prediction tasks and prediction models, interference feature factors are excluded, the stability of feature selection is strengthened, and the operation speed and accuracy of power transmission line parameter prediction are effectively improved.

Description

Technical Field

[0001] This application pertains to a feature selection method, specifically a feature selection method and related apparatus for predicting transmission line parameters. Background Technology

[0002] The design of existing transmission line parameters is mostly based on the calculation of electrical parameters at both ends of the line after it has been erected but not yet put into operation, by artificially applying a certain voltage using electrical equipment and measuring the electrical parameters using precision measuring instruments. However, once the transmission line is energized and in operation, factors such as the operation mode of the transmission network, the line environment, and seasonal changes cause the actual line parameters to differ from the initially calculated values. This reduces the accuracy of state estimation calculations in local power grid areas and the accuracy of calculation results for other advanced applications.

[0003] With the rapid development of big data and artificial intelligence, deep learning models have received widespread attention in the field of transmission line parameter prediction. Deep learning models can handle the nonlinear and multivariate relationships involved in transmission line parameter prediction; however, the accuracy of deep learning models largely depends on the selection of feature combinations. Appropriate feature combinations can significantly improve the accuracy of deep learning models in predicting transmission line parameters. Therefore, appropriate feature combinations are a prerequisite for deep learning models to achieve accurate results. However, existing transmission line operating condition feature data is becoming increasingly complex. While this provides more data support for using deep learning models to predict transmission line parameters, it also presents a challenge in selecting appropriate feature combinations from the numerous data features affecting transmission line parameters.

[0004] Currently, the most common method for selecting combinations of transmission line operating condition features is correlation analysis, such as Pearson correlation analysis and Spearman correlation analysis. These methods calculate the correlation between each feature and transmission line parameters, and then select and combine features based on the correlation. While Pearson and Spearman correlation analyses are good at analyzing linear relationships between data, factors such as weather, transmission line operating data, and transmission line parameters themselves exhibit significant nonlinearity, making them difficult to handle. In contrast, the Maximum Information Coefficient (MIC) method based on mutual information theory can evaluate different types of correlations, identify a wide range of relationship types, and detect the strength of nonlinear correlations between data. Therefore, MIC can be used to measure the nonlinear relationships between transmission line-related data, serving as a basis for selecting feature combinations for deep learning prediction models.

[0005] The above feature selection methods are all data-driven analysis methods, which select feature combinations only based on the data analysis results. This may lead to instability when the selected feature combinations are applied to the prediction model, mainly in three aspects: (1) The selected feature combinations may only be suitable for specific data types. When the data changes, it may not be possible to select appropriate feature combinations; (2) The selected feature combinations may be suitable for a certain type of prediction model, but when the prediction model changes, it will lead to inaccurate prediction; (3) The selected feature combinations may be suitable for a specific type of prediction task, but when the prediction task changes, the selected feature combinations may no longer be suitable for the prediction model. Summary of the Invention

[0006] The purpose of this application is to solve the problems in the prior art and provide a feature selection method and related apparatus for predicting transmission line parameters.

[0007] To achieve the above objectives, this application adopts the following technical solution:

[0008] Firstly, this application proposes a feature selection method for predicting transmission line parameters, including:

[0009] Time-series data of transmission line operating conditions are acquired to form a time-series data set of operating conditions; the transmission line operating conditions include features characterizing the transmission line operating data and labels characterizing the transmission line parameters;

[0010] The time-series data set of the operating conditions is input into a trained feature evaluation neural network model; the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected transmission line operating conditions X from the time-series data set of operating conditions. T×(m+1) The pooling layer is used to monitor the operating conditions X of the transmission line. T×(m+1) Perform global average pooling to reduce the dimensionality of the data Y to one dimension. 1×(m+1) The fully connected layer is used to obtain two feature vectors Y1 through the activation function f1. 1×(m+1) and Y2 1×(m+1) The attention layer is used to process the feature vector Y1. 1×(m+1) and Y2 1×(m+1) The summation is performed, and attention weights corresponding to the features and labels of each transmission line's operating conditions are generated through the activation function f2. These weights are then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data.

[0011] The attention weights corresponding to the features and labels of each transmission line's operating conditions are obtained from the attention layer of the feature evaluation neural network model, and the importance of each feature relative to the label in the operating conditions of the transmission line is calculated.

[0012] Sort the features from highest to lowest importance to obtain the sorted feature set;

[0013] From the sorted feature set, p features are selected p times to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations;

[0014] The time series data to be measured is input into the prediction model layer to obtain p prediction results. The feature combination corresponding to the prediction result with the smallest prediction error is selected as the feature for predicting transmission line parameters.

[0015] Furthermore, the fully connected layer comprises two parallel and structurally identical fully connected neural networks;

[0016] The two feature vectors Y1 are obtained through the activation function f1. 1×(m+1) and Y2 1×(m+1) ,include:

[0017] Y1 1×(m+1) =f1(W1) dense *Y 1×(m+1) +b1 dense )

[0018] Y2 1×(m+1) =f1(W2) dense *Y 1×(m+1) +b2 dense )

[0019] Among them, W1 dense and W2 dense Let b1 and b2 represent the weight parameters of the two fully connected neural networks, respectively. dense b2 dense These represent the offset parameters of the two fully connected neural networks.

[0020] Furthermore, the selection of the feature combination corresponding to the prediction result with the smallest prediction error includes:

[0021] Calculate the MAE, RMSE, or R2 of the prediction results respectively, and use the values ​​of MAE, RMSE, or R2 as the standard for the magnitude of the prediction error.

[0022] Furthermore, the activation function f1 is the tanh activation function, and the activation function f2 is the sigmoid activation function.

[0023] Furthermore, the label includes the reactance, resistance, and / or susceptance values ​​of the transmission line;

[0024] The features include the active power, reactive power, voltage, and current at the beginning of the transmission line, the active power and reactive power at the end of the line, the voltage at the end of the line, the current at the end of the line, the air temperature, and the wind speed.

[0025] Furthermore, the prediction model layer is a CNN model.

[0026] Furthermore, the rule for feature selection is as follows: during the q-th selection, the first q features are taken; where p is an integer greater than or equal to 2, and p≥q≥1;

[0027] Before performing p feature selections from the sorted feature set, the process also includes: deleting features whose importance is below a threshold from the sorted feature set.

[0028] Secondly, this application proposes a feature selection system for transmission line parameter prediction, comprising:

[0029] The data module is used to acquire time-series data of the operating conditions of transmission lines and form a time-series data set of operating conditions; the operating conditions of transmission lines include features that characterize the operating data of transmission lines and labels that characterize the parameters of transmission lines.

[0030] The model module is used to input the time-series data set of the operating conditions into the trained feature evaluation neural network model; the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected time-series data sets of transmission line operating conditions X from the time-series data set of operating conditions. T×(m+1) The pooling layer is used to monitor the operating conditions X of the transmission line. T×(m+1) Perform global average pooling to reduce the dimensionality of the data Y to one dimension. 1×(m+1) The fully connected layer is used to obtain two feature vectors Y1 through the activation function f1. 1×(m+1) and Y2 1×(m+1) The attention layer is used to process the feature vector Y1. 1×(m+1) and Y2 1×(m+1) The summation is performed, and attention weights corresponding to the features and labels of each transmission line's operating conditions are generated through the activation function f2. These weights are then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data.

[0031] The weighting module is used to evaluate the attention layer of the neural network model from the features, obtain the attention weights corresponding to the features and labels of each transmission line operating condition, and calculate the importance of each feature relative to the label in the transmission line operating condition.

[0032] The sorting module is used to sort the features from highest to lowest importance, resulting in a sorted feature set.

[0033] The selection module is used to select features p times from the sorted feature set to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations.

[0034] The selection module is used to input the time series data to be tested into the prediction model layer to obtain p prediction results, and select the feature combination corresponding to the prediction result with the smallest prediction error as the feature for predicting transmission line parameters.

[0035] Thirdly, this application proposes an electronic device, comprising:

[0036] Memory, used to store computer programs;

[0037] A processor is configured to implement the steps of the feature selection method for transmission line parameter prediction described above when executing the computer program.

[0038] Fourthly, this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the feature selection method for transmission line parameter prediction described above.

[0039] Compared with the prior art, this application has the following beneficial effects:

[0040] This application proposes a feature selection method for transmission line parameter prediction. A corresponding feature evaluation neural network model is designed, including an input layer, pooling layer, fully connected layer, attention layer, and prediction model layer. This combines feature evaluation with the prediction model. By obtaining the attention weights corresponding to the features and labels of each transmission line's operating conditions, the relative importance of each feature is determined, achieving feature selection based on an attention mechanism. This method can select the most suitable combination of transmission line operating condition features according to different transmission line data, prediction tasks, and prediction models, eliminating interfering feature factors, enhancing the stability of feature selection, and effectively improving the computational speed and accuracy of transmission line parameter prediction. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1This is a schematic flowchart of an embodiment of the feature selection method for predicting transmission line parameters in this application;

[0043] Figure 2 This is a schematic diagram of the network structure of the feature evaluation neural network model in the embodiments of this application;

[0044] Figure 3 This is a schematic diagram of a connection of an embodiment of the feature selection system for transmission line parameter prediction in this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0046] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0047] To provide better combinations of input features for the prediction model, it is necessary to closely integrate feature selection with the prediction model itself, thereby providing the required input features. When faced with different transmission line data, prediction tasks, and prediction models, the feature selection method of this application for transmission line parameter prediction is more suitable for scenarios where deep learning prediction models are used to predict transmission line parameters.

[0048] In predicting transmission line parameters, deep learning models can combine various feature combinations for prediction, such as line geometric parameters (line length, conductor radius, tower height, tower spacing, etc.), meteorological conditions (wind speed, temperature, humidity, rainfall, etc.), and geographical location information (longitude, latitude, altitude). In practical applications, appropriate feature combinations and processing methods can be selected according to specific circumstances to improve the model's prediction accuracy and generalization ability.

[0049] like Figure 1 The diagram shown is a flowchart illustrating a feature selection method for transmission line parameter prediction according to this application, which may include:

[0050] S101, acquire time-series data of transmission line operating conditions to form a time-series data set of operating conditions. Transmission line operating conditions include features characterizing transmission line operating data and labels characterizing transmission line parameters.

[0051] In practical applications, the labels for the operating conditions of transmission lines can include data such as line reactance, line resistance, or line susceptance. Characteristics of the operating conditions of transmission lines can include data such as active power at the beginning of the line, reactive power at the beginning of the line, voltage at the beginning of the line, current at the beginning of the line, active power at the end of the line, reactive power at the end of the line, voltage at the end of the line, current at the end of the line, temperature, and wind speed.

[0052] S102, input the time-series data set of operating conditions into the trained feature evaluation neural network model. For example, Figure 2 As shown, the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected transmission line operating conditions X from the time-series data set of operating conditions. T×(m+1) Pooling layer, used to monitor the operating conditions X of transmission lines. T×(m+1) Perform global average pooling to reduce the dimensionality of the data Y to one dimension. 1×(m+1) A fully connected layer is used to obtain two feature vectors Y1 through the activation function f1. 1×(m+1) and Y2 1×(m+1) Attention layer, used to process feature vector Y1 1×(m+1) and Y2 1×(m+1) The summation is performed, and attention weights corresponding to the features and labels of each transmission line's operating conditions are generated through the activation function f2. These weights are then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data.

[0053] S103, through the attention layer of the feature evaluation neural network model, obtains the attention weights corresponding to the features and labels of each transmission line operating condition, and calculates the importance of each feature relative to the label in the transmission line operating condition.

[0054] In practical applications, by using a feature evaluation neural network model, attention layers can be used to output the features and corresponding attention weights of each transmission line's operating conditions during feature selection. Furthermore, the feature evaluation neural network model can be trained based on a complete feature evaluation neural network model structure.

[0055] S104. Sort the features according to their importance from high to low to obtain the sorted feature set.

[0056] S105, select features p times from the sorted feature set to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations.

[0057] S106, input the time series data to be measured into the prediction model layer respectively to obtain p prediction results, and select the feature combination corresponding to the prediction result with the smallest prediction error as the feature for predicting transmission line parameters.

[0058] As a preferred embodiment of the feature selection method for transmission line parameter prediction in this application, it may include the following steps:

[0059] S201, collect time-series data of transmission line operating conditions to form a time-series data set of operating conditions, in the form of {C1,C2,...,C...} i ,…,C n}, where C i Let C represent the operating condition of the transmission line at time i, where n is the total number of time points in the time series data set, and both i and n are positive integers greater than or equal to 1. i It contains m features and 1 label, in the form of Where A represents a feature, the superscript i represents the time value, the subscripts 1 to m represent the feature number, m is a positive integer greater than or equal to 1, and L represents the label. The features of the transmission line operating conditions refer to the operating data of the transmission line, and the labels of the transmission line operating conditions refer to the transmission line parameters.

[0060] S202, Construct a feature evaluation neural network model, which includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer.

[0061] The input layer is used to input time-series data of the transmission line's operating conditions. This data consists of T sequentially connected time-series data points of the transmission line's operating conditions selected from the time-series data set of operating conditions, denoted as X. T×(m+1) It contains T rows of data, each row containing m+1 elements, in the form {C j C j+1 ,...,C j+T-1}, j∈[1,n-T+1], where j is a positive integer.

[0062] Pooling layers are used to process the input data X from the input layer. T×(m+1) Perform global average pooling to reduce the dimensionality to one-dimensional data Y. 1×(m+1) .

[0063] The fully connected layer uses two parallel and identical fully connected neural networks connected to the pooling layer, and obtains two feature vectors Y1 through the activation function f1. 1×(m+1) With Y2 1×(m+1) ;

[0064] Y1 1×(m+1) =f1(W1) dense *Y 1×(m+1) +b1dense (1)

[0065] Y2 1×(m+1) =f1(W2) dense *Y 1×(m+1) +b2 dense (2)

[0066] Among them, W1 dense and W2 dense Let b1 and b2 represent the weight parameters of the two fully connected neural networks, respectively. dense and b2 dense These represent the offset parameters of the two fully connected neural networks. The activation function f1 can be the tanh activation function.

[0067] The attention layer is used to process the output data Y1 of the fully connected layer. 1×(m+1) With Y2 1×(m+1) The summation is performed, and the attention weights corresponding to each feature and label of the operating condition are generated through the activation function f2, in the form of {I A1 ,I A2 ,…,I Am ,I L The input layer data is then weighted based on attention weights, meaning that each type of working condition data is multiplied by its corresponding attention weight to generate weighted time series data, in the form of... The activation function f2 can be selected from the sigmoid activation function.

[0068] The prediction model layer is used as input for weighted time series data, and outputs the prediction results of transmission line parameters.

[0069] In practical applications, the prediction model layer can be selected from traditional time series models such as ARMA and ARIMA, or machine learning time series models such as LSTM and CNN, or other types of time series models. For the problem of transmission line parameter prediction, a CNN model can be used. A CNN model is generally composed of an input layer, convolutional layers, pooling layers, and fully connected layers (activation layers).

[0070] S203 divides the time-series data set of transmission line operating conditions into a training set and a test set. The training set is used to input the model for training, and the test set is used to evaluate the training effect of the model.

[0071] The time-series data set of the operating conditions can be divided into a certain proportion according to the chronological order. The specific proportion can be selected according to actual needs, and this application does not impose any restrictions. For example, the first 70% of the dataset can be used as the training set, and the last 30% as the test set.

[0072] S204, Input the time-series data of the transmission line operating conditions in the test set into the trained feature evaluation neural network model, and obtain the attention weights {I} in the attention layer. A1 ,I A2 ,…,I Am ,I L The importance of the characteristics and labels of transmission line operating conditions is determined, and the importance RI of each characteristic relative to the label is calculated. Ak =I Ak / I L Among them, RI Ak Indicates feature A k The relative importance of , where k∈[1,m], and k is a positive integer.

[0073] S205, sort the features according to their relative importance from high to low, and the reordered feature set is {A}. r1 A r2 ,…,A rl ,...,A rm}, where A rl The feature is ranked lth in relative importance. P features are selected, and the selection rule can be set as follows: in the q-th selection, the first q features are taken; where p is an integer greater than or equal to 2, and p ≥ q ≥ 1. In other embodiments of this application, other feature selection rules can also be used, which can be set according to the required selection requirements, and this application does not impose any restrictions. In the q-th selection, the first q features are sequentially taken from the reordered feature set to form a feature combination, where p ≤ m, q ∈ [1, p], and p and q are both positive integers. Finally, p feature combinations are formed, in the form {A... r1},{A r1 A r2},...,{A r1 A r2 ,...,A rp}

[0074] In practical applications, before performing p feature selections, features whose relative importance is below a threshold can be deleted. This threshold is either manually defined or based on experience and can be set as needed.

[0075] S206, generate time series data corresponding to p feature combinations, input them into the prediction model layer respectively, and select the feature combination with the smallest prediction error as the input feature based on the output of the prediction model.

[0076] In practical applications, prediction error can be characterized by the following methods: MAE (Mean Absolute Error), RMSE (Root Mean Square Error), or R² (Coefficient of Determination) can be used to evaluate the accuracy of model predictions. The specific calculation methods are as follows:

[0077]

[0078]

[0079]

[0080] Among them, P act,i This represents the actual values ​​of the transmission line parameters. P represents the average value of actual transmission line parameters. pre,i This represents the predicted values ​​of the transmission line parameters, where n represents the number of samples.

[0081] It should be noted that only a few typical judgment indicators are shown here. Other judgment indicators may also be used in other embodiments of this application.

[0082] The following is an example of the feature selection method used in this application for transmission line parameter prediction:

[0083] S301 collects time-series data of the operating conditions of transmission lines to form a time-series data set of operating conditions.

[0084] The time series data set for the operating conditions contains 5000 data sets, in the form of {C1, C2, ..., C...} 5000 As an example, Table 1 shows only the time-series data of six sets of transmission line operating conditions. The features include active power (Pa), reactive power (Qa), voltage (Ua), current (Ia) at the head-end, active power (Pb), reactive power (Qb), voltage (Ub), current (Ib) at the tail-end, temperature (T), and wind speed (S). The label is the line reactance value (X). The operating condition C of the transmission line at time i is... i It can be represented as {Pa i Pb i Qa i ,Qb i Ua i Ub i ,Ia i ,Ib i ,T i ,S i ,X i}

[0085] Table 1. Example table of time-series datasets for transmission lines.

[0086]

[0087] S302, Construct a feature evaluation neural network model, which includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer.

[0088] The input layer takes into account the time-series data of the transmission line's operating conditions. This data consists of two consecutively selected time points from the time-series data set of operating conditions, denoted as X. 2×11 It contains two rows of data, each row containing 11 elements, in the form of {C j C j+1}, j∈[1,4999], where j is a positive integer.

[0089] The pooling layer is used to process the data X from the input layer. 2×11 Perform global average pooling to reduce the dimensionality to one-dimensional data Y. 1×11 .

[0090] The fully connected layer uses two parallel and identical fully connected neural networks connected to the pooling layer, and obtains two feature vectors Y1 through the tanh activation function. 1×11 With Y2 1×11 .

[0091] The attention layer is used to process the output data Y1 of the fully connected layer. 1×11 With Y2 1×11 The summation is performed, and the attention weights corresponding to each feature and label of the running condition are generated through the sigmoid activation function, in the form of {I Pa ,I Qa ,...,I S ,I X The input layer data is then weighted based on attention weights, meaning that each type of working condition data is multiplied by its corresponding attention weight to generate weighted time series data, in the form of {{Pa}. j ×I Pa Qa j ×I Qa ,...,S j ×I S ,X j ×I X},{Pa j+1 ×I Pa Qa j+1 ×I Qa ,...,S j+1 ×I S ,X j+1 ×I X}}.

[0092] The prediction model layer is used to input weighted time series data. After passing through the CNN time series prediction model, the predicted results of transmission line parameters are output. The CNN model selects one-dimensional convolution, which contains 3 convolutional layers, 3 pooling layers and 1 single fully connected layer. The prior stride is set to 32, the convolutional kernel is set to 3, and the output sequence is 1.

[0093] S303 divides the time-series data set of transmission line operating conditions into a training set and a test set. The training set is used to input the model for training, and the test set is used to evaluate the training effect of the model.

[0094] The training set consists of the first 70% of the time-series data set on transmission line operating conditions, containing 3,500 sets of data. The test set consists of the last 30% of the time-series data set on transmission line operating conditions, containing 1,500 sets of data.

[0095] S304, input the time series data of the transmission line operating conditions in the test set into the trained feature evaluation neural network model, obtain the attention weights in the attention layer as the importance of the transmission line operating condition features and labels, and calculate the importance of each feature relative to the label.

[0096] The relative importance of Pa, Qa, Ua, Ia, Pb, Qb, Ub, Ib, T, and S are 1.62, 0.09, 0.13, 0.65, 1.45, 0.98, 0.19, 0.47, 1.22, and 0.02, respectively.

[0097] S305, delete features whose relative importance is below the threshold of 0.1, sort the features according to their relative importance, and then take out different numbers of features in sequence to form multiple feature combinations.

[0098] The reordered feature set is {Pa,Pb,T,Qb,Ia,Ib,Ub,Ua}. Eight feature selections are performed. In the q-th selection, the first q features are sequentially taken from the reordered feature set to form a feature combination, where q is a positive integer. Finally, eight feature combinations are formed, namely {Pa}, {Pa,Pb},...,{Pa,Pb,T,Qb,Ia,Ib,Ub,Ua}.

[0099] S306. Input the time series data corresponding to multiple feature combinations into the prediction model respectively, and select the feature combination with the smallest prediction error as the input feature based on the output of the prediction model.

[0100] The accuracy of the model predictions was evaluated using MAE, RMSE, and R2 metrics. As shown in Table 2, the feature combination {Pa,Pb,T,Qb,Ia,Ib} had the smallest error value, so the final selected input feature combination was {Pa,Pb,T,Qb,Ia,Ib}.

[0101] Table 2. Feature Combination Error Index Values

[0102]

[0103] like Figure 3The diagram shown is a connection schematic of a feature selection system for transmission line parameter prediction provided in an embodiment of this application, including:

[0104] The data module is used to acquire time-series data of the operating conditions of transmission lines and form a time-series data set of operating conditions; the operating conditions of transmission lines include features that characterize the operating data of transmission lines and labels that characterize the parameters of transmission lines.

[0105] The model module is used to input the time-series data set of the operating conditions into the trained feature evaluation neural network model; the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected time-series data sets of transmission line operating conditions X from the time-series data set of operating conditions. T×(m+1) The pooling layer is used to monitor the operating conditions X of the transmission line. T×(m+1) Perform global average pooling to reduce the dimensionality of the data Y to one dimension. 1×(m+1) The fully connected layer is used to obtain two feature vectors Y1 through the activation function f1. 1×(m+1) and Y2 1×(m+1) The attention layer is used to process the feature vector Y1. 1×(m+1) and Y2 1×(m+1) The summation is performed, and attention weights corresponding to the features and labels of each transmission line's operating conditions are generated through the activation function f2. These weights are then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data.

[0106] The weighting module is used to evaluate the attention layer of the neural network model from the features, obtain the attention weights corresponding to the features and labels of each transmission line operating condition, and calculate the importance of each feature relative to the label in the transmission line operating condition.

[0107] The sorting module is used to sort the features from highest to lowest importance, resulting in a sorted feature set.

[0108] The selection module is used to select features p times from the sorted feature set to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations.

[0109] The selection module is used to input the time series data to be tested into the prediction model layer to obtain p prediction results, and select the feature combination corresponding to the prediction result with the smallest prediction error as the feature for predicting transmission line parameters.

[0110] It should be noted that in other embodiments of the feature selection system for transmission line parameter prediction in this application, the specific functions of each module can also adopt the aforementioned embodiments of the optimized feature selection method, which will not be repeated here.

[0111] It should be noted that, in the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another device, or some features may be ignored or not executed. The modules described as separate components may or may not be physically separated. The components shown as modules may be one or more physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs.

[0112] Furthermore, in the various embodiments of the present invention, the modules can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The integrated unit described above can be implemented in hardware or as a software functional unit.

[0113] An electronic device provided in this application includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of a feature selection method for predicting transmission line parameters as described in any of the above embodiments.

[0114] Another electronic device provided in this application embodiment may further include: an input port connected to a processor for transmitting multimodal data collected by an external acquisition device to the processor; a display unit connected to the processor for displaying the processor's processing results to the outside world; and a communication module connected to the processor for enabling communication between the electronic device and the outside world. The display unit may be a display panel, a laser scanner, or the like; the communication method used by the communication module includes, but is not limited to, Mobile High Definition Link (HML), Universal Serial Bus (USB), High Definition Multimedia Interface (HDMI), wireless connectivity: Wi-Fi, Bluetooth communication, Bluetooth Low Energy communication, and IEEE 802.11s-based communication technology.

[0115] This application provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of a feature selection method for transmission line parameter prediction as described in any of the above embodiments.

[0116] The computer-readable storage media involved in this application include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage media known in the art.

[0117] For descriptions of relevant parts of the feature selection system, electronic device, and computer-readable storage medium for transmission line parameter prediction provided in this application's embodiments, please refer to the detailed description of the corresponding parts in the feature selection method for transmission line parameter prediction provided in this application's embodiments; they will not be repeated here. Furthermore, parts of the technical solutions provided in this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0118] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A feature selection method for predicting transmission line parameters, characterized in that, include: Acquire time-series data of the operating conditions of transmission lines to form a time-series data set of operating conditions; The transmission line operating conditions include features characterizing the transmission line operating data, and labels characterizing the transmission line parameters; the labels include the transmission line's reactance, resistance, and / or susceptance values. The features include the active power, reactive power, voltage, and current at the beginning of the transmission line, the active power and reactive power at the end of the line, the voltage at the end of the line, the current at the end of the line, the air temperature, and the wind speed. The time-series data set of the operating conditions is input into a trained feature evaluation neural network model; the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected time-series data sets of transmission line operating conditions from the time-series data set of operating conditions. ; The pooling layer is used to monitor the operating conditions of the transmission line. Perform global average pooling to reduce the dimensionality to one-dimensional data. The fully connected layer comprises two parallel and structurally identical fully connected neural networks. The fully connected layer is used to activate a function... Two eigenvectors were obtained. and The attention layer is used to process the feature vector. and Summation is performed, followed by an activation function. Attention weights are generated corresponding to the features and labels of each transmission line's operating conditions, and then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data. The attention weights corresponding to the features and labels of each transmission line's operating conditions are obtained from the attention layer of the feature evaluation neural network model, and the importance of each feature relative to the label in the operating conditions of the transmission line is calculated. Sort the features from highest to lowest importance to obtain the sorted feature set; From the sorted feature set, p features are selected p times to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations; The time series data to be measured is input into the prediction model layer to obtain p prediction results. The feature combination corresponding to the prediction result with the smallest prediction error is selected as the feature for predicting transmission line parameters.

2. The feature selection method for transmission line parameter prediction according to claim 1, characterized in that, The activation function Two eigenvectors were obtained. and ,include: in, and These represent the weight parameters of the two fully connected neural networks. , These represent the offset parameters of the two fully connected neural networks.

3. The feature selection method for transmission line parameter prediction according to claim 1 or 2, characterized in that, The selection of the feature combination corresponding to the prediction result with the smallest prediction error includes: Calculate the MAE, RMSE, or R2 of the prediction results respectively, and use the values ​​of MAE, RMSE, or R2 as the standard for the magnitude of the prediction error.

4. The feature selection method for transmission line parameter prediction according to claim 3, characterized in that, The activation function The activation function is tanh. This is the sigmod activation function.

5. The feature selection method for transmission line parameter prediction according to claim 4, characterized in that, The prediction model layer is a CNN model.

6. The feature selection method for transmission line parameter prediction according to claim 5, characterized in that, The rule for feature selection is as follows: for the qth selection, the first q features are taken; where p is an integer greater than or equal to 2, and p≥q≥1; Before performing p feature selections from the sorted feature set, the process also includes: deleting features whose importance is below a threshold from the sorted feature set.

7. A feature selection system for predicting transmission line parameters, characterized in that, include: The data module is used to acquire time-series data of the operating conditions of transmission lines and form a time-series data set of operating conditions. The transmission line operating conditions include features characterizing the transmission line operating data, and labels characterizing the transmission line parameters; the labels include the transmission line's reactance, resistance, and / or susceptance values. The features include the active power, reactive power, voltage, and current at the beginning of the transmission line, the active power and reactive power at the end of the line, the voltage at the end of the line, the current at the end of the line, the air temperature, and the wind speed. The model module is used to input the time-series data set of the operating conditions into the trained feature evaluation neural network model; the feature evaluation neural network model includes an input layer, a pooling layer, a fully connected layer, an attention layer, and a prediction model layer; the input layer is used to select T sequentially connected time-series data sets of transmission line operating conditions from the time-series data set of operating conditions. ; The pooling layer is used to monitor the operating conditions of the transmission line. Perform global average pooling to reduce the dimensionality to one-dimensional data. The fully connected layer comprises two parallel and structurally identical fully connected neural networks. The fully connected layer is used to activate a function... Two eigenvectors were obtained. and The attention layer is used to process the feature vector. and Summation is performed, followed by an activation function. Attention weights are generated corresponding to the features and labels of each transmission line's operating conditions, and then weighted with the corresponding transmission line operating conditions to generate weighted time series data. The prediction model layer is used to output the prediction results of the transmission line parameters based on the weighted time series data. The weighting module is used to evaluate the attention layer of the neural network model from the features, obtain the attention weights corresponding to the features and labels of each transmission line operating condition, and calculate the importance of each feature relative to the label in the transmission line operating condition. The sorting module is used to sort the features from highest to lowest importance, resulting in a sorted feature set. The selection module is used to select features p times from the sorted feature set to obtain p feature combinations, forming the time series data to be tested corresponding to the p feature combinations. The selection module is used to input the time series data to be tested into the prediction model layer to obtain p prediction results, and select the feature combination corresponding to the prediction result with the smallest prediction error as the feature for predicting transmission line parameters.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the feature selection method for transmission line parameter prediction as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the feature selection method for transmission line parameter prediction as described in any one of claims 1 to 6.

Images