Power transmission line uneven icing and deicing accident risk assessment method and system

Abstract

The invention provides a power transmission line uneven icing and deicing accident risk assessment method and system, and relates to the technical field of power transmission line icing disaster prevention and control and machine learning, and the method comprises the steps: obtaining the related data of uneven icing and different periods of deicing of a power transmission line, and carrying out the preprocessing of the related data, and constructing a data set; constructing an improved Transform-CNN-based dynamic and static approaching distance calculation model between wires or between the wires and the ground wires, and training the model by using a data set to obtain a trained model; collecting line parameters and icing and deicing working condition data of a to-be-evaluated power transmission line, inputting the processed data into the trained model, and calculating to obtain a static and dynamic minimum distance between an upper wire and a lower wire or between a ground wire; and according to the line voltage grade, the calculated value and the line and working condition parameters, carrying out uneven icing and different-period deicing accident risk assessment to obtain a risk grade. According to the invention, accident risk early warning can be effectively realized, and prevention and treatment of uneven icing and different-period deicing accidents of the power transmission line can be guided.

Description

Technical Field

[0001] The present invention relates to the field of transmission line icing disaster prevention and control and machine learning technology, and in particular to a method and system for assessing the risk of uneven icing and deicing accidents of transmission lines. Background Art

[0002] In recent years, global warming and extreme weather events have occurred frequently, and icing has become one of the main threats to the safe operation of power transmission lines. Uneven icing and deicing of transmission lines refers to the uneven icing and deicing at different periods of time that occur on transmission lines. Affected by micro-topography, micro-meteorology, lines, operating parameters, etc., uneven icing and deicing at different periods are prone to occur between conductors, ground wires and gears in actual lines, which in turn leads to insufficient insulation gaps between conductors and ground wires, resulting in discharge accidents. According to statistics, in the past three years, icing faults of 220 kV and above voltage level transmission lines in a northern province accounted for more than 50% of the total faults, of which 70% were caused by uneven icing and deicing at different periods. In addition, the winter load is high and the pressure on power supply is high, so it is imperative to assess and warn the risks of uneven icing and deicing at different periods of transmission lines.

[0003] Existing methods mainly focus on the calculation of deicing jump height in the dynamic process of uneven ice coverage and different deicing periods, mainly including experimental, theoretical and numerical simulation methods. Numerical simulation has become the main means to study the dynamic response of deicing, with the characteristics of accurate calculation results, but its modeling process is relatively complicated, the calculation efficiency is not high, and it is very inconvenient in engineering applications. Based on the results of numerical simulation, relevant scholars have given a simplified engineering formula for calculating the deicing jump height. At the same time, the design regulations also give the calculation formula for the deicing jump height, but both formulas require nonlinear iteration to first determine the sag difference before and after the conductor deicing, which is inconvenient to use. Patent CN116070490A proposes to use numerical simulation and MATLAB to fit the calculation formula for the deicing jump height of the conductor to carry out deicing jump risk warning. First, each evaluation calculation requires numerical simulation, which is complex to model and inefficient to calculate; second, the fitting regression formula has low calculation accuracy for high-dimensional nonlinear problems such as deicing jump. With the popularization of machine learning technology, a calculation method based on machine learning has been proposed. Patent CN116842808A proposes a method for assessing the risk of deicing of continuous conductors with non-uniform icing. It uses a random forest algorithm to calculate the deicing height of the conductor and conduct risk assessment. It does not consider the risk of discharge accidents that may occur in the static condition of non-uniform icing. At the same time, the random forest algorithm belongs to the category of traditional machine learning. It has high-dimensional nonlinearity and feature extraction problems in non-uniform icing and deicing processes at different stages, and its learning efficiency and performance need to be verified. The document "Conductor Deicing Jump Height Prediction Model Based on BP Neural Network" uses a BP neural network algorithm to predict the deicing jump height. The efficiency has been significantly improved compared to finite element calculations, but no accident risk assessment method is given. At the same time, the BP neural network is highly dependent on data, usually requiring large-scale data sets to obtain good performance, and has an overfitting problem.

[0004] Therefore, in view of the high-dimensional nonlinearity and difficulty in extracting characteristic quantities of uneven icing and deicing processes in different periods, as well as the current situation that relevant data cannot be obtained on a large scale, it is urgent to propose an accurate and efficient gap distance calculation and accident risk assessment method. Summary of the invention

[0005] To this end, an embodiment of the present invention provides a method and system for assessing the risk of uneven icing and de-icing accidents of transmission lines, which are used to solve the problems in the prior art of high-dimensional nonlinearity and difficulty in extracting characteristic quantities of uneven icing and de-icing processes at different stages, as well as the inability to obtain relevant data on a large scale.

[0006] In order to solve the above problems, an embodiment of the present invention provides a method for assessing the risk of uneven icing and deicing accidents of power transmission lines, the method comprising:

[0007] S1: Obtain data related to uneven icing and deicing at different periods of transmission lines, pre-process the relevant data according to time series changes to construct a data set, and divide it into training set, validation set and test set in proportion, and normalize the data at the same time;

[0008] S2: Based on the network structure of the CNN model and the Transformer model, a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN;

[0009] The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes:

[0010] First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ...

[0011] S3: Collect the line parameters and ice-covering and ice-shedding condition data of the transmission line to be evaluated, and process them in the data form required by the model. Input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and ground wires;

[0012] S4: Risk assessment of uneven icing and deicing accidents at different stages is conducted based on the line voltage level, the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the line and operating parameters to determine the risk level.

[0013] Preferably, in step S1, the method for obtaining data related to uneven icing and deicing of power transmission lines at different stages is: obtaining data related to uneven icing and deicing of power transmission lines at different stages through physical experiments and finite element simulations.

[0014] Preferably, in step S1, the data related to uneven icing and deicing in different periods of the power transmission line include line parameters and icing and deicing operating condition parameters;

[0015] The line parameters include: the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of adjacent upper and lower conductors or ground wires;

[0016] The ice-covering and de-icing working condition parameters include: ice-covering thickness of adjacent upper and lower conductors or ground conductors, de-icing level, de-icing rate, wind speed and wind direction perpendicular to the line direction.

[0017] Preferably, in step S1, the data is normalized, and the normalization formula is:

[0018]

[0019] Where, i∈{1,2,3,4,5,6,7,8,9,10,11,12}, different i represent the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between the adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of the adjacent upper and lower conductors or ground wires, the ice thickness, deicing section number, deicing rate, wind speed and wind direction perpendicular to the line direction of the adjacent upper and lower conductors or ground wires; and Represents the maximum and minimum values ​​of the corresponding variables; x i ′ represents the normalized variable.

[0020] Preferably, in step S2, the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is composed of a CNN model and a Transformer model, the data processed by the CNN model is input into the Transformer model, and the processed data is input into X′ n After the linear change of the fully connected layer, we get Then with the corresponding position code Pos(X′ n ) are added and input into the encoder E; the mask input is X′ mask After the linear change of the fully connected layer, we get Then with the corresponding position decoder Pos(X′ mask ) is added together with the encoder output and input to the decoder Di The decoder output is then multiplied by The linear mapping is used to obtain the final result. The following loss function is used in the whole process:

[0021]

[0022] In the formula, n represents the number of samples, y i is the actual value, is the calculated value, All are weights.

[0023] Preferably, the CNN model consists of 3 convolutional layers. First, the input data dimension is [60, 12, 1], the first convolutional layer has 32 convolution kernels, and the output data dimension is [60, 3, 32]; the second convolutional layer has 64 convolution kernels, and the output data dimension is [60, 2, 64]; the third convolutional layer has 1 convolutional layer, and the output data dimension is [60, 1, 1]

[0024] Preferably, the Transformer model consists of an encoder and a decoder;

[0025] The encoder is composed of an input layer, a positional encoding, and three identical encoding layers. In the input layer design, the data processed by the CNN module is input and mapped into a vector through a fully connected network. Then, the data is added to the vector obtained by the positional encoding, and the result is input into the encoding layer. Each encoding layer includes a multi-head attention module and a feedforward neural network, and the outputs of the multi-head attention module and the feedforward neural network are normalized. The feedforward neural network is composed of two linear transformations, and the middle is a linear rectification activation function relu(x)=max(0,x). The feedforward neural network formula is expressed as follows:

[0026] FN(x)=relu(x)W2+b2=max(0,xW1+b1)W2+b2

[0027] Among them, W1, W2, b1, and b2 are the transformation matrices and bias items of the two linear transformations respectively. The outputs of the first two encoding layers are used as the inputs of the next encoding layer. Finally, the encoding layer outputs other dimensional vectors and inputs them into the decoding layer. To avoid the multi-head attention mechanism not considering time, a sinusoidal curve function is added to encode the input data position information:

[0028]

[0029] Among them, Pos is the position of the input data, i is the dimension of the input data, and the position encoding obtains a d-dimensional position encoding vector;

[0030] The decoder includes an input layer, a position encoding and three identical decoding layers; in the decoding layer, a masked multi-head attention mechanism is inserted before the multi-head attention mechanism to limit the model to only process data at the current time without referring to future sequences, so as to better meet the causal relationship requirements in sequence tasks.

[0031] Preferably, the structure of the multi-head attention module specifically includes:

[0032] First, multiply each input by the transformation matrix to obtain the corresponding vectors q, k, and v to calculate the self-attention mechanism. The dimensions of vectors q and k are d v , q is dot-multiplied by k and divided by the scale Then use the following softmax equivalent formula to calculate the numerical weight

[0033]

[0034] Then, the calculated weight is multiplied by the vector v, and all inputs are transformed to obtain q i ,k i ,v i The matrices are represented as Q, K, V, and the weights are expressed as the following formula:

[0035]

[0036] In the formula, softmax() is the activation function;

[0037] Finally, the input is mapped into three spaces, input A i Multiply by h transformation matrices to form multiple (Q i ,K i ,V i ), merge all the results, and then input the matrix into the fully connected layer for linear transformation and output.

[0038] Preferably, the accident risk assessment includes static risk assessment and dynamic risk assessment;

[0039] The static risk assessment method is: the static minimum distance d between the upper and lower conductors or between the conductors and the ground wires j Minimum gap d between the ground wire or conductor of the corresponding voltage level under switching overvoltage c Compare:

[0040] When j >1.2d c When , the static risk level is 0;

[0041] When c <d j ≤1.2d cWhen , the static risk level is 1;

[0042] When j ≤d c When , the static risk level is 2;

[0043] The dynamic risk assessment method is: the dynamic minimum distance d between the upper and lower conductors or between the conductors and the ground wires d The minimum gap d between the ground wire or conductor of the corresponding voltage level under power frequency voltage g Compare:

[0044] When d >1.2d g When , the dynamic risk level is 0;

[0045] When g <d d ≤1.2d g When , the dynamic risk level is 1;

[0046] When d ≤d g The dynamic risk level is 2.

[0047] Preferably, the risk level is calculated by combining the static risk level and the dynamic risk level, specifically including:

[0048] When the static risk level is 0, and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 0, 1, and 1, respectively;

[0049] When the static risk level is 1 and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 1, 1, and 2 respectively;

[0050] When the static risk level is 2, and the dynamic risk levels are 0, 1, and 2 respectively, the corresponding risk levels are 2, 2, and 2 respectively.

[0051] Preferably, when the deicing rate is 0, only static risk assessment is performed, and the dynamic risk level is calculated as 0.

[0052] Preferably, the minimum clearance value between the ground wire or conductor of the corresponding voltage level under the power frequency voltage and the operational overvoltage is the value specified in the latest national or power industry electrical design regulations plus the conductor splitting circle radius or the conductor splitting circle diameter.

[0053] The embodiment of the present invention further provides a transmission line uneven icing and deicing accident risk assessment system, the system is used to implement the above-mentioned transmission line uneven icing and deicing accident risk assessment method, specifically comprising:

[0054] The data acquisition and preprocessing module is used to obtain data related to uneven ice coverage and de-icing at different periods of the transmission line, preprocess the relevant data according to the time series changes to construct a data set, and divide it into training set, verification set and test set according to proportion, and normalize the data at the same time;

[0055] A training module is used to construct a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the network structure of the CNN model and the Transformer model, and train the model using a data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN;

[0056] The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes:

[0057] First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ...

[0058] The calculation module is used to collect the line parameters and ice-covering and de-icing condition data of the transmission line to be evaluated, and process them in the data form required by the model, and input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires;

[0059] The accident risk assessment module is used to conduct uneven icing and deicing accident risk assessment at different stages according to the line voltage level, the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the line and operating parameters to derive the risk level.

[0060] An embodiment of the present invention also provides an electronic device, which includes a processor, a memory and a bus system, wherein the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned method for assessing the risk of uneven icing and de-icing accidents of transmission lines.

[0061] An embodiment of the present invention further provides a computer storage medium storing a computer software product. The computer software product includes a number of instructions for enabling a computer device to execute the above-mentioned method for assessing the risk of uneven icing and deicing accidents of transmission lines.

[0062] It can be seen from the above technical solutions that the present invention has the following advantages:

[0063] In the embodiment of the present invention, a method and system for assessing the risk of uneven icing and deicing accidents of transmission lines are proposed. By obtaining data related to uneven icing and deicing in different periods of transmission lines, the relevant data are preprocessed to construct a data set; a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using the data set to obtain a trained model; the line parameters and icing and deicing operating condition data of the transmission line to be evaluated are collected, and the processed data are input into the trained model to calculate the static and dynamic minimum distances between the upper and lower conductors or between the conductors and ground wires; the risk of uneven icing and deicing accidents in different periods is assessed according to the line voltage level, the calculated value, and the line and operating condition parameters to obtain the risk level. The present invention is based on the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires of the improved Transformer-CNN, and calculates the dynamic and static minimum distances between conductors or between conductors and ground wires with uneven icing and deicing in different periods, which solves the problem of mining and feature extraction of high-dimensional complex data in this process; compared with the traditional machine learning and shallow model solutions such as random forests and BP neural networks, the model calculation efficiency and accuracy are higher. At the same time, this method clearly gives the assessment method of the risk of uneven icing and deicing accidents in different periods by evaluating static and dynamic risks respectively. It has strong operability and can effectively realize accident risk warning and guide the prevention and control of uneven icing and deicing accidents in different periods of transmission lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0065] Figure 1 A flow chart of a method for assessing the risk of uneven icing and deicing accidents of power transmission lines provided in an embodiment;

[0066] Figure 2 Schematic diagram of the Transformer model structure in the embodiment;

[0067] Figure 3 This is a schematic diagram of the structure of the multi-head attention mechanism in the embodiment;

[0068] Figure 4 It is a block diagram of a transmission line uneven icing and deicing accident risk assessment system provided in an embodiment. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0070] Embodiment 1

[0071] like Figure 1 As shown, an embodiment of the present invention provides a method for assessing the risk of uneven icing and deicing accidents of power transmission lines, the method comprising:

[0072] S1: Obtain data related to uneven icing and deicing at different periods of transmission lines, pre-process the relevant data according to time series changes to construct a data set, and divide it into training set, validation set and test set in proportion, and normalize the data at the same time;

[0073] S2: Based on the network structure of the CNN model and the Transformer model, a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN;

[0074] The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes:

[0075] First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ...

[0076] S3: Collect the line parameters and ice-covering and ice-shedding condition data of the transmission line to be evaluated, and process them in the data form required by the model. Input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and ground wires;

[0077] S4: Risk assessment of uneven icing and deicing accidents at different stages is conducted based on the line voltage level, the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the line and operating parameters to determine the risk level.

[0078] It can be seen from the above technical scheme that the present invention provides a method for assessing the risk of uneven icing and deicing accidents of transmission lines. By obtaining data related to uneven icing and deicing in different periods of transmission lines, the relevant data are preprocessed to construct a data set; a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using the data set to obtain a trained model; the line parameters and icing and deicing operating condition data of the transmission line to be evaluated are collected, and the processed data are input into the trained model to calculate the static and dynamic minimum distances between the upper and lower conductors or between the conductors and ground wires; the risk of uneven icing and deicing accidents in different periods is assessed according to the line voltage level, calculated value, and line and operating condition parameters to obtain the risk level. The present invention is based on the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires of the improved Transformer-CNN, and calculates the dynamic and static minimum distances between conductors or between conductors and ground wires with uneven icing and deicing in different periods, which solves the problem of mining and feature extraction of high-dimensional complex data in this process; compared with the traditional machine learning and shallow model solutions such as random forests and BP neural networks, the model calculation efficiency and accuracy are higher. At the same time, this method clearly gives the assessment method of the risk of uneven icing and deicing accidents in different periods by evaluating static and dynamic risks respectively. It has strong operability and can effectively realize accident risk warning and guide the prevention and control of uneven icing and deicing accidents in different periods of transmission lines.

[0079] In this embodiment, in step S1, data related to uneven icing and deicing in different periods of the transmission line are obtained through physical experiments and finite element simulation. The relevant data are preprocessed according to the time series changes to construct a data set, and are divided into a training set, a validation set, and a test set in proportion, and the data are normalized at the same time.

[0080] Specifically, the data related to uneven icing and deicing of transmission lines include line parameters and icing and deicing operating parameters. Line parameters include: the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of adjacent upper and lower conductors or ground wires. Icing and deicing operating parameters include: ice thickness of adjacent upper and lower conductors or ground wires, deicing section number, deicing rate, wind speed and direction perpendicular to the line direction.

[0081] Furthermore, the data is normalized, and the normalization formula is:

[0082]

[0083] Where, i∈{1,2,3,4,5,6,7,8,9,10,11,12}, different i represent the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between the adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of the adjacent upper and lower conductors or ground wires, the ice thickness, deicing section number, deicing rate, wind speed and wind direction perpendicular to the line direction of the adjacent upper and lower conductors or ground wires; and Represents the maximum and minimum values ​​of the corresponding variables; x i ′ represents the normalized variable.

[0084] Furthermore, the training batch size is determined to be 60, and a total of 12 attributes of line parameters and icing and de-icing operating parameters are used as the network input data form to construct data with a structure like [60,12].

[0085] In this embodiment, in step S2, based on the network structure of the CNN model and the Transformer model, a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using a data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN.

[0086] Among them, the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is composed of a CNN model and a Transformer model.

[0087] Specifically, the CNN model consists of three convolutional layers. The input data dimension is [60, 12, 1]. The first convolutional layer has 32 convolution kernels and the output data dimension is [60, 3, 32]. The second convolutional layer has 64 convolution kernels and the output data dimension is [60, 2, 64]. The third convolutional layer has 1 convolutional layer and the output data dimension is [60, 1, 1].

[0088] This implementation is set up through a convolutional neural network (CNN). The data with a data dimension of [60, 12] is multiplied rightward and downward by the corresponding matrix elements and summed. After a linear rectification activation function, [60, 3] is obtained. The entire convolution calculation process can be expressed by the following formula.

[0089]

[0090] Where a is the output of the convolution operation, x is the input line parameters and ice-covering and de-icing condition parameters, w is the convolution kernel parameter, and b is the bias parameter.

[0091] Furthermore, the Transformer model consists of an encoder and a decoder. Figure 2 As shown, the encoder consists of an input layer, a positional encoding, and three identical encoding layers. In the input layer design, the data processed by the CNN module is input and mapped into a vector through a fully connected network; then it is added to the vector obtained by the positional encoding, and the result is input to the encoding layer. Each encoding layer includes a multi-head attention module and a feedforward neural network, and the outputs of the multi-head attention module and the feedforward neural network are normalized; the feedforward neural network consists of two linear transformations with a linear rectification activation function relu(x)=max(0,x) in the middle. The feedforward neural network formula is expressed as follows:

[0092] FN(x)=relu(x)W2+b2=max(0,xW1+b1)W2+b2

[0093] Among them, W1, W2, b1, b2 are the transformation matrices and bias terms of the two linear transformations respectively. The outputs of the first two encoding layers are used as the inputs of the next encoding layer. Finally, the encoding layer outputs other dimensional vectors and inputs them into the decoding layer. In order to avoid the multi-head attention mechanism not considering time, a sinusoidal function is added to encode the input data position information:

[0094]

[0095] Among them, Pos is the position of the input data, i is the dimension of the input data, and the position encoding obtains a d-dimensional position encoding vector.

[0096] The decoder includes an input layer, a positional encoder, and three identical decoding layers. The structure of the decoder is similar to that of the encoder, where a masked multi-head attention mechanism is inserted before the multi-head attention mechanism in the decoding layer to limit the model to only process data at the current time, without referring to future sequences, so as to better meet the causal relationship requirements in sequence tasks. During the de-icing process, the attention mechanism can lock the data at the current moment and use the data at the current moment for training. Finally, the decoder outputs the vector, which is directly mapped through a fully connected neural network to obtain the output.

[0097] Furthermore, the structure of the multi-head attention module used in this embodiment is as follows Figure 3 As shown, specifically including:

[0098] First, multiply each input by the transformation matrix to obtain the corresponding vectors q, k, and v to calculate the self-attention mechanism. The dimensions of vectors q and k are d v , q is dot-multiplied by k and divided by the scale Then use the following softmax equivalent formula to calculate the numerical weight

[0099]

[0100] Then, the calculated weight is multiplied by the vector v, and all inputs are transformed to obtain q i ,k i ,v i The matrices are represented as Q, K, V, and the weights are expressed as the following formula:

[0101]

[0102] In the formula, softmax() is the activation function;

[0103] Finally, the input is mapped into three spaces, input A i Multiply by h transformation matrices to form multiple (Q i ,K i ,V i ), merge all the results, and then input the matrix into the fully connected layer for linear transformation and output.

[0104] Considering that the conductor deicing jump process is related to many aspects, this embodiment adopts a multi-head attention mechanism, so that the attention mechanism can pay more attention to the information in the line parameters and ice covering and deicing condition data, and then extract richer features. Based on the prior experimental results, this embodiment uses a three-head attention mechanism. By referencing multiple attention heads, each head learns a set of weights, allowing the model to simultaneously focus on different features in different parts of the input sequence. In response to the conductor deicing problem, combined with prior experimental data, this model is innovatively adjusted and uses an embedded weight adjustment module to enable the attention mechanism to pay more attention to the three factors of span, ice thickness, and deicing rate during network training.

[0105] Furthermore, the data set is used to train a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, and a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained, which specifically includes:

[0106] First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ...

[0107] Specifically, during the training process, the data processed by the CNN model is input into the Transformer model, and the processed data is input into X′ n After the linear change of the fully connected layer, we get Then with the corresponding position code Pos(X′ n ) are added and input into the encoder E; the mask input is X′ mask After the linear change of the fully connected layer, we get Then with the corresponding position decoder Pos(X′ mask ) is added together with the encoder output and input to the decoder D i The decoder output is then multiplied by The linear mapping is used to obtain the final result. The following loss function is used in the whole process:

[0108]

[0109] In the formula, n represents the number of samples, y i is the actual value, is the calculated value, The model of the present invention adopts Adam optimizer, and the learning rate is set to 0.005.

[0110] In this embodiment, in step S3, the line parameters and ice coating and de-icing condition data of the transmission line to be evaluated are collected, and processed in the data form required by the model, and the processed data are input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires.

[0111] In this embodiment, in step S4, the risk assessment of uneven icing and deicing accidents at different stages is performed based on the line voltage level, the static minimum distance between the upper and lower conductors or between the ground wires, the dynamic minimum distance between the upper and lower conductors or between the ground wires, and the line and operating conditions parameters to obtain the risk level.

[0112] Specifically, accident risk assessment includes static risk assessment and dynamic risk assessment. The static risk assessment method is: the static minimum distance d between the upper and lower conductors or between the conductor and the ground wire is j Minimum gap d between the ground wire or conductor of the corresponding voltage level under switching overvoltage c Compare:

[0113] When j >1.2d c When , the static risk level is 0;

[0114] When c <d j ≤1.2d c When , the static risk level is 1;

[0115] When j ≤d c The static risk level is 2.

[0116] The dynamic risk assessment method is: the dynamic minimum distance d between the upper and lower conductors or between the conductors and the ground wires d The minimum gap d between the ground wire or conductor of the corresponding voltage level under power frequency voltage g Compare:

[0117] When d >1.2d g When , the dynamic risk level is 0;

[0118] When g <d d ≤1.2d g When , the dynamic risk level is 1;

[0119] When d ≤d g The dynamic risk level is 2.

[0120] The final risk level is calculated by combining the static risk level and the dynamic risk level according to Table 1 (Risk Level Assessment Table), including:

[0121] When the static risk level is 0, and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 0, 1, and 1, respectively;

[0122] When the static risk level is 1 and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 1, 1, and 2 respectively;

[0123] When the static risk level is 2, and the dynamic risk levels are 0, 1, and 2 respectively, the corresponding risk levels are 2, 2, and 2 respectively.

[0124] Table 1

[0125]

[0126] Furthermore, when the de-icing rate is 0, only static risk assessment is performed and the dynamic risk level is calculated as 0.

[0127] Furthermore, the minimum clearance value under the power frequency voltage and operational overvoltage between the ground wire or conductor of the corresponding voltage level is the value specified in the latest national or power industry electrical design regulations plus the conductor splitting circle radius (in the case of ground wire) or the conductor splitting circle diameter (in the case of between conductors).

[0128] In order to illustrate the advantages of the method of the present invention, it is described below in conjunction with a specific power transmission line.

[0129] Taking a ±660kV line as an example, firstly, the line parameters and ice-covering and ice-shedding parameters are collected, including the tension section data, tower type data, conductor data, ground wire data and ice-covering thickness, as shown in Tables 2 to 6.

[0130] Table 2

[0131]

[0132] Table 3

[0133] Tower Type Horizontal distance of ground wire (m) Vertical distance of ground wire (m) Z 2.71 6.63 J 5.1 11.62

[0134] Table 4

[0135] project Data Value <![CDATA[Cross-sectional area (mm 2 )]]> 1045.38 Initial tension (MPa) 54.08

[0136] Table 5

[0137] project Data Value <![CDATA[Cross-sectional area (mm 2 )]]> 148.07 Initial tension (MPa) 301.5

[0138] Table 6

[0139] File No. Conductor ice thickness (mm) Ground wire ice thickness (mm) 1 15 15 2 10 15 3 5 20 4 15 20 5 15 10 6 10 10 7 15 20

[0140] The de-icing level is 4; the de-icing rate is 80%; the wind speed in the vertical direction is 5m / s; the wind direction is towards the inside of the tower.

[0141] Arrange the actual data collected according to the initial data when training the model to form real working condition data that conforms to the model data format. First, collect the 12 types of attribute data required by the model, arrange the collected information in order, and generate a set of roughly 12 types of data according to the attribute arrangement. The specific number of each type of data in this set varies and needs to be processed separately. For example, there are seven data in the gear spacing data class. Before entering the model, each data in this single attribute data class needs to be processed, and after processing, it is input into the model.

[0142] Then, this part of the data is normalized; among them, the span, height difference, horizontal offset distance and vertical offset distance between adjacent ground wires of each tower, cross-sectional area of ​​adjacent ground wires, initial stress and other data are normalized. After normalization, a set of values ​​within 0-1 can be obtained; the values ​​of the number of gears, deicing gear number, wind direction and other values ​​are fixed and cannot be normalized. In this method, they are set to fixed values ​​between 0-1. For example, the wind direction of the vertical line is divided into inside and outside the tower, and the inside of the tower is recorded as 0.3 and the outside of the tower is recorded as 0.7; the deicing gear number is divided into seven equal parts, and finally a data scale of 12 is formed and sent to the model.

[0143] Input the processed data into the model and run it to obtain the static minimum distance d between the ground wires under the actual working condition. j =3.6m, upper and lower conductor ground wire or dynamic minimum distance d d =1.71m.

[0144] In this embodiment, according to the latest power industry regulations, the minimum clearance values ​​for the power frequency voltage and switching overvoltage between the ±660kV ground wires at an altitude of 1800m are 2.05m and 4.7m respectively. The conductor is four-split, and the radius of the split circle is 0.25m. Therefore, the power frequency voltage d corresponding to the ±660kV ground wire in this example is g and switching overvoltage d c The minimum clearance values ​​are 2.3m and 4.95m respectively.

[0145] Among them, d j >1.2d c , the static approach risk level is 0;

[0146] And d d ≤d g , the dynamic approach risk level is 2.

[0147] According to the risk level assessment table (Table 1), the risk level of uneven icing and deicing accidents on the transmission line in this case is level 1.

[0148] Embodiment 2

[0149] like Figure 4 As shown, the present invention provides a transmission line uneven icing and deicing accident risk assessment method system, the system is used to implement the transmission line uneven icing and deicing accident risk assessment method described in Example 1, specifically comprising:

[0150] The data acquisition and preprocessing module 10 is used to obtain data related to uneven ice coverage and deicing in different periods of the transmission line, preprocess the relevant data according to the time series changes to construct a data set, and divide it into a training set, a validation set and a test set in proportion, and normalize the data at the same time;

[0151] The training module 20 is used to construct a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the network structure of the CNN model and the Transformer model, and train the model using a data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN;

[0152] The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes:

[0153] First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ...

[0154] The calculation module 30 is used to collect the line parameters and ice-covering and de-icing condition data of the transmission line to be evaluated, and process them in the data form required by the model, input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, and calculate the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires;

[0155] The accident risk assessment module 40 is used to evaluate the risk of uneven icing and deicing accidents in different periods according to the line voltage level, the static minimum distance between the upper and lower conductors or between the ground wires, the dynamic minimum distance between the upper and lower conductors or between the ground wires, and the line and operating parameters to obtain the risk level.

[0156] A transmission line uneven icing and de-icing accident risk assessment system in this embodiment is used to implement the aforementioned transmission line uneven icing and de-icing accident risk assessment method. Therefore, the specific implementation method of the transmission line uneven icing and de-icing accident risk assessment system can be seen in the embodiment part of the transmission line uneven icing and de-icing accident risk assessment method in the foregoing text. For example, the data acquisition and preprocessing module 10, the training module 20, the calculation module 30, and the accident risk assessment module 40 are respectively used to implement steps S1, S2, S3, and S4 in the aforementioned transmission line uneven icing and de-icing accident risk assessment method. Therefore, its specific implementation method can refer to the description of the corresponding various parts of the embodiment. In order to avoid redundancy, it will not be repeated here.

[0157] Embodiment 3

[0158] An embodiment of the present invention also provides an electronic device, which includes a processor, a memory and a bus system, wherein the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned method for assessing the risk of uneven icing and de-icing accidents of transmission lines.

[0159] Embodiment 4

[0160] An embodiment of the present invention further provides a computer storage medium storing a computer software product. The computer software product includes a number of instructions for enabling a computer device to execute the above-mentioned method for assessing the risk of uneven icing and deicing accidents of transmission lines.

[0161] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0162] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate 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.

[0163] 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 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A method for assessing the risk of uneven icing and deicing accidents on power transmission lines, characterized in that: include: S1: Obtain data related to uneven icing and deicing at different periods of transmission lines, pre-process the relevant data according to time series changes to construct a data set, and divide it into training set, validation set and test set in proportion, and normalize the data at the same time; S2: Based on the network structure of the CNN model and the Transformer model, a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is constructed, and the model is trained using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN; The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes: First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ... S3: Collect the line parameters and ice-covering and ice-shedding condition data of the transmission line to be evaluated, and process them in the data form required by the model. Input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and ground wires; S4: Risk assessment of uneven icing and deicing accidents at different stages is conducted based on the line voltage level, the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the line and operating parameters to determine the risk level.

2. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 1 is characterized in that: In step S1, the method for obtaining data related to uneven icing and deicing of power transmission lines at different periods is: obtaining data related to uneven icing and deicing of power transmission lines at different periods through physical experiments and finite element simulations.

3. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 1, characterized in that: In step S1, the data related to uneven icing and deicing in different periods of the transmission line include line parameters and icing and deicing operating condition parameters; The line parameters include: the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of adjacent upper and lower conductors or ground wires; The ice-covering and de-icing working condition parameters include: ice-covering thickness of adjacent upper and lower conductors or ground conductors, de-icing level, de-icing rate, wind speed and wind direction perpendicular to the line direction.

4. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 1, characterized in that: In step S1, the data is normalized, and the normalization formula is: Where, i∈{1,2,3,4,5,6,7,8,9,10,11,12}, different i represent the number of line tension sections, the span and height difference of each section, the horizontal offset distance and vertical offset distance between the adjacent upper and lower conductors or ground wires of each tower, the cross-sectional area and initial stress of the adjacent upper and lower conductors or ground wires, the ice thickness, deicing section number, deicing rate, wind speed and wind direction perpendicular to the line direction of the adjacent upper and lower conductors or ground wires; and Represents the maximum and minimum values ​​of the corresponding variables; x i ′ represents the normalized variable.

5. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 1, characterized in that: In step S2, the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is composed of a CNN model and a Transformer model. The data processed by the CNN model is input into the Transformer model, and the processed data is input into X′ n After the linear change of the fully connected layer, we get Then with the corresponding position code Pos(X′ n ) are added and input into the encoder E; the mask input is X′ mask After the linear change of the fully connected layer, we get Then with the corresponding position decoder Pos(X′ mask ) is added together with the encoder output and input to the decoder D i The decoder output is then multiplied by The linear mapping obtains the final result, and the whole process uses the following loss function: In the formula, n represents the number of samples, y i is the actual value, is the calculated value, All are weights.

6. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 5, characterized in that: The CNN model consists of three convolutional layers. The input data dimension is [60, 12, 1]. The first convolutional layer has 32 convolution kernels and the output data dimension is [60, 3, 32]. The second convolutional layer has 64 convolution kernels and the output data dimension is [60, 2, 64]. The third convolutional layer has 1 convolutional layer and the output data dimension is [60, 1, 1].

7. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 5, characterized in that: The Transformer model consists of an encoder and a decoder; The encoder is composed of an input layer, a positional encoding, and three identical encoding layers. In the input layer design, the data processed by the CNN module is input and mapped into a vector through a fully connected network. Then, the data is added to the vector obtained by the positional encoding, and the result is input into the encoding layer. Each encoding layer includes a multi-head attention module and a feedforward neural network, and the outputs of the multi-head attention module and the feedforward neural network are normalized. The feedforward neural network is composed of two linear transformations, and the middle is a linear rectification activation function relu(x)=max(0,x). The feedforward neural network formula is expressed as follows: FN(x)=relu(x)W2+b2=max(0,xW1+b1)W2+b2 Among them, W1, W2, b1, and b2 are the transformation matrices and bias items of the two linear transformations respectively. The outputs of the first two encoding layers are used as the inputs of the next encoding layer. Finally, the encoding layer outputs other dimensional vectors and inputs them into the decoding layer. To avoid the multi-head attention mechanism not considering time, a sinusoidal curve function is added to encode the input data position information: Among them, Pos is the position of the input data, i is the dimension of the input data, and the position encoding obtains a d-dimensional position encoding vector; The decoder includes an input layer, a position encoding and three identical decoding layers; in the decoding layer, a masked multi-head attention mechanism is inserted before the multi-head attention mechanism to limit the model to only process data at the current time without referring to future sequences, so as to better meet the causal relationship requirements in sequence tasks.

8. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 7, characterized in that: The structure of the multi-head attention module specifically includes: First, multiply each input by the transformation matrix to obtain the corresponding vectors q, k, and v to calculate the self-attention mechanism. The dimensions of vectors q and k are d v , q is dot-multiplied by k and divided by the scale Then use the following softmax equivalent formula to calculate the numerical weight Then, the calculated weight is multiplied by the vector v, and all inputs are transformed to obtain q i ,k i ,v i The matrices are represented as Q, K, V, and the weights are expressed as the following formula: In the formula, softmax() is the activation function; Finally, the input is mapped into three spaces, input A i Multiply by h transformation matrices to form multiple (Q i ,K i ,V i ), merge all the results, and then input the matrix into the fully connected layer for linear transformation and output.

9. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 1, characterized in that: The accident risk assessment includes static risk assessment and dynamic risk assessment; The static risk assessment method is: the static minimum distance d between the upper and lower conductors or between the conductors and the ground wires j Minimum gap d between the ground wire or conductor of the corresponding voltage level under switching overvoltage c Compare: When j >1.2d c When , the static risk level is 0; When c <d j ≤1.2d c When , the static risk level is 1; When j ≤d c When , the static risk level is 2; The dynamic risk assessment method is: the dynamic minimum distance d between the upper and lower conductors or between the conductors and the ground wires d The minimum gap d between the ground wire or conductor of the corresponding voltage level under power frequency voltage g Compare: When d >1.2d g When , the dynamic risk level is 0; When g <d d ≤1.2d g When , the dynamic risk level is 1; When d ≤d g The dynamic risk level is 2.

10. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 9, characterized in that: The risk level is calculated by combining the static risk level and the dynamic risk level, including: When the static risk level is 0, and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 0, 1, and 1, respectively; When the static risk level is 1 and the dynamic risk levels are 0, 1, and 2, the corresponding risk levels are 1, 1, and 2 respectively; When the static risk level is 2, and the dynamic risk levels are 0, 1, and 2 respectively, the corresponding risk levels are 2, 2, and 2 respectively.

11. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 10, characterized in that: When the de-icing rate is 0, only static risk assessment is performed and the dynamic risk level is calculated as 0.

12. The method for risk assessment of uneven icing and deicing accidents of power transmission lines according to claim 9, characterized in that: The minimum clearance value between the ground wire or conductor of the corresponding voltage level under power frequency voltage and operating overvoltage is the value specified in the latest national or power industry electrical design regulations plus the conductor splitting circle radius or conductor splitting circle diameter.

13. A transmission line uneven icing and deicing accident risk assessment system, characterized in that: The system is used to implement the method for risk assessment of uneven icing and deicing accidents of power transmission lines according to any one of claims 1 to 12, and specifically comprises: The data acquisition and preprocessing module is used to obtain data related to uneven ice coverage and de-icing at different periods of the transmission line, preprocess the relevant data according to the time series changes to construct a data set, and divide it into training set, verification set and test set according to proportion, and normalize the data at the same time; A training module is used to construct a dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the network structure of the CNN model and the Transformer model, and train the model using a data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN; The model is trained by using the data set to obtain a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN, which specifically includes: First, the divided training set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN for training, and the mode and parameters are established through data training; then, the divided verification set is input into the dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to adjust the model hyperparameters and monitor the model performance to avoid overfitting and underfitting, and finally a trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN is obtained; finally, the divided test set is input into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between conductors and ... The calculation module is used to collect the line parameters and ice-covering and de-icing condition data of the transmission line to be evaluated, and process them in the data form required by the model, and input the processed data into the trained dynamic and static proximity distance calculation model between conductors or between conductors and ground wires based on the improved Transformer-CNN to calculate the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires; The accident risk assessment module is used to conduct uneven icing and deicing accident risk assessment at different stages according to the line voltage level, the static minimum distance between the upper and lower conductors or between the conductors and the ground wires, the dynamic minimum distance between the upper and lower conductors or between the conductors and the ground wires, and the line and operating parameters to derive the risk level.

14. An electronic device, characterized in that: The electronic device includes a processor, a memory and a bus system, the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the method for risk assessment of uneven icing and deicing accidents of transmission lines as described in any one of claims 1 to 12.

15. A computer storage medium, characterized in that The computer storage medium stores a computer software product, and the computer software product includes a number of instructions for enabling a computer device to execute the method for assessing the risk of uneven icing and deicing accidents of transmission lines as described in any one of claims 1 to 12.

Images