Power transmission and distribution switch state detection method and system

Abstract

A method and system for detecting the state of a power transmission and distribution switch is disclosed. First, the operating parameter information at each predetermined time point is passed through a parameter context encoder to obtain multiple operating parameter context semantic understanding feature vectors. Then, the multiple operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and input into a dual-branch network structure to obtain a classification feature map. Finally, the classification feature map is passed through a classifier to obtain a classification result indicating whether the state of the power transmission and distribution switch to be detected is normal, and the classification result is displayed on a screen. In this way, accurate detection of the state of the power transmission and distribution switch can be achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a method and system for detecting the status of a power transmission and distribution switch. Background Art

[0002] In the power transmission and distribution system, the power transmission and distribution isolating switches are used to switch the lines under no-load conditions, and to electrically isolate or connect the busbars, circuit breakers and other electrical equipment under maintenance with the lines. Therefore, the reliability and stability requirements for the power transmission and distribution isolating switches are becoming increasingly higher.

[0003] There are many ways to detect the status of power transmission and distribution disconnectors, but automatic detection technology can effectively improve the safety of DC disconnectors and make the system highly reliable. The comprehensive application of sensor technology, microelectronics technology, automatic control technology, computer technology, signal analysis and processing technology, data communication technology, etc. is of great significance for timely, objective and accurate detection of mechanical properties such as jamming, rust and pressure of power transmission and distribution disconnectors.

[0004] Therefore, a state detection solution for power transmission and distribution switches is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and system for detecting the state of a power transmission and distribution switch. First, the operating parameter information of each predetermined time point is respectively passed through a parameter context encoder to obtain a plurality of operating parameter context semantic understanding feature vectors, then, the plurality of operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and then input into a dual-branch network structure to obtain a classification feature map, and finally, the classification feature map is passed through a classifier to obtain a classification result indicating whether the state of the power transmission and distribution switch to be detected is normal, and the classification result is displayed on the screen. In this way, accurate detection of the state of the power transmission and distribution switch can be achieved.

[0006] According to one aspect of the present application, a method for detecting a state of a power transmission and distribution switch is provided, comprising:

[0007] Obtaining operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor;

[0008] Passing the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors;

[0009] Arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and inputting the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use dilated convolution kernels with different dilated rates;

[0010] The classification feature graph is passed through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and the classification result is displayed on a screen.

[0011] In the above-mentioned power transmission and distribution switch state detection method, the operating parameter information of each predetermined time point is respectively passed through a parameter context encoder including an embedding layer to obtain multiple operating parameter context semantic understanding feature vectors, including:

[0012] Performing word segmentation processing on the operation parameter information at each predetermined time point to convert the operation parameter information at each predetermined time point into a first word sequence consisting of a plurality of words;

[0013] Mapping each word in the first word sequence to a word vector using the embedding layer of the parameter context encoder including the embedding layer to obtain a sequence of first word vectors;

[0014] And, using the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the multiple running parameter context semantic understanding feature vectors.

[0015] In the above-mentioned power transmission and distribution switch state detection method, using the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the plurality of operating parameter context semantic understanding feature vectors, including:

[0016] Arrange the sequence of the first word vectors in one dimension to obtain a first global word sequence feature vector;

[0017] Calculating the product of the first global word sequence feature vector and the transposed vector of each word vector in the sequence of the first word vectors to obtain a plurality of first self-attention association matrices;

[0018] Normalizing each of the plurality of first self-attention association matrices to obtain a plurality of first standardized self-attention association matrices;

[0019] Each of the multiple first standardized self-attention association matrices is passed through a Softmax classification function to obtain multiple first probability values; and each of the first word vectors in the sequence of the first word vectors is weighted using each of the multiple first probability values ​​as a weight to obtain the multiple operating parameter context semantic understanding feature vectors.

[0020] In the above-mentioned power transmission and distribution switch state detection method, the plurality of operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, including:

[0021] Using the first convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a first classification feature matrix, wherein the first convolutional neural network model has a first dilated convolution kernel with a first dilated rate;

[0022] Performing two-dimensional convolution encoding on the two-dimensional feature matrix using the second convolutional neural network model of the dual-branch network structure to obtain a second classification feature matrix, wherein the second convolutional neural network model has a second dilated convolution kernel with a second dilated rate, and the first dilated rate is different from the second dilated convolution kernel;

[0023] And, the first classification feature matrix and the second classification feature matrix are fused to obtain the classification feature map.

[0024] In the above-mentioned power transmission and distribution switch state detection method, it also includes training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier;

[0025] The step of training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier comprises:

[0026] Acquire training data, wherein the training data includes training operation parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, and a real value of whether the state of the power transmission and distribution switch to be detected is normal;

[0027] Passing the training operation parameter information at each predetermined time point through the parameter context encoder including the embedding layer to obtain a plurality of training operation parameter context semantic understanding feature vectors;

[0028] Arranging the plurality of training operation parameter context semantic understanding feature vectors into a training two-dimensional feature matrix and inputting the matrix into the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model to obtain a training classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use dilated convolution kernels with different dilated rates;

[0029] Performing feature distribution optimization on the training classification feature graph to obtain an optimized training classification feature graph;

[0030] Passing the optimized training classification feature map through the classifier to obtain a classification loss function value;

[0031] And, based on the classification loss function value and through back propagation of gradient descent, the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier are trained.

[0032] In the above-mentioned power transmission and distribution switch state detection method, the feature distribution of the training classification feature map is optimized to obtain an optimized training classification feature map, including:

[0033] Expanding the training classification feature map into multiple classification feature matrices along the channel dimension;

[0034] Converting each of the classification feature matrices into a classification diagonal matrix to obtain a plurality of classification diagonal matrices;

[0035] The plurality of classification diagonal matrices are optimized by bitwise shift association matching of intrinsic unitization to obtain a plurality of optimized classification diagonal matrices; and the plurality of optimized classification diagonal matrices are dimensionally reconstructed to obtain the optimized training classification feature map.

[0036] In the above-mentioned power transmission and distribution switch state detection method, the plurality of classification diagonal matrices are subjected to intrinsic unitization bit-by-bit shift association matching optimization to obtain a plurality of optimized classification diagonal matrices, including:

[0037] Performing bit-by-bit shift association matching optimization of the plurality of classified diagonal matrices with intrinsic unitization according to the following enhanced formula to obtain the plurality of optimized classified diagonal matrices;

[0038] Wherein, the strengthening formula is:

[0039]

[0040]

[0041] in, is the diagonal matrix of each classification, arrive is obtained by performing eigendecomposition on the diagonal matrices of each classification eigenvalues, For the The eigenvalues ​​are arranged along the diagonal to obtain the eigennormal matrix, and and are all diagonal matrices, is the distance between the intrinsic unit matrix and each of the classification diagonal matrices, represents matrix multiplication, represents matrix addition, It means point multiplication by position. A diagonal matrix is ​​classified for each of the optimizations.

[0042] In the above-mentioned power transmission and distribution switch state detection method, the optimized training classification feature map is passed through the classifier to obtain the classification loss function value, including:

[0043] The classifier processes the optimized training classification feature map using the following classification formula to generate a classification result, wherein the classification formula is: ,in represents projecting the optimized training classification feature map into a vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the classification result and the true value as the classification loss function value.

[0044] According to another aspect of the present application, a power transmission and distribution switch state detection system is provided, comprising:

[0045] A data acquisition module, used to acquire operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor;

[0046] A parameter context encoding module, configured to pass the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors;

[0047] A dual-branch convolution module, used for arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and then inputting the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use a dilated convolution kernel with different dilated rates;

[0048] A classification module is used to pass the classification feature map through a classifier to obtain a classification result, and the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and a display module is used to display the classification result on a screen.

[0049] In the above-mentioned power transmission and distribution switch state detection system, the parameter context encoding module includes:

[0050] A word segmentation unit, configured to perform word segmentation processing on the operation parameter information at each predetermined time point to convert the operation parameter information at each predetermined time point into a first word sequence consisting of a plurality of words;

[0051] A word embedding unit, used to use the embedding layer of the parameter context encoder including the embedding layer to map each word in the first word sequence to a word vector to obtain a sequence of first word vectors; and a context semantic encoding unit, used to use the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the multiple running parameter context semantic understanding feature vectors.

[0052] Compared with the prior art, the power transmission and distribution switch state detection method and system provided by the present application firstly passes the operating parameter information of each predetermined time point through the parameter context encoder to obtain multiple operating parameter context semantic understanding feature vectors, then arranges the multiple operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and inputs them into a dual-branch network structure to obtain a classification feature map, and finally passes the classification feature map through a classifier to obtain a classification result indicating whether the state of the power transmission and distribution switch to be detected is normal, and displays the classification result on the screen. In this way, accurate power transmission and distribution switch state detection can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The following drawings are not deliberately scaled according to the actual size, and the focus is on illustrating the main purpose of the present application.

[0054] Figure 1 This is an application scenario diagram of the power transmission and distribution switch status detection method according to an embodiment of the present application.

[0055] Figure 2 Flow chart of a method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0056] Figure 3 Schematic diagram of the architecture of a method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0057] Figure 4 This is a flowchart of sub-step S120 of the method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0058] Figure 5 This is a flowchart of sub-step S123 of the method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0059] Figure 6 This is a flowchart of sub-step S130 of the method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0060] Figure 7 This is a flowchart of sub-step S200 in the method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application.

[0061] Figure 8 This is a flowchart of sub-step S240 in the power transmission and distribution switch status detection method according to an embodiment of the present application.

[0062] Fig. 9 4 is a block diagram of a power transmission and distribution switch status detection system according to an embodiment of the present application.

[0063] Fig.10 It is a block diagram of the parameter context encoding module in the power transmission and distribution switch state detection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0065] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0066] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0067] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0068] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0069] As mentioned above, there are many means of status detection of power transmission and distribution isolating switches, but automatic detection technology can effectively improve the safety of DC isolating switches and make the system highly reliable. The comprehensive application of sensor technology, microelectronics technology, automatic control technology, computer technology, signal analysis and processing technology, data communication technology, etc., is of great significance for timely, objective and accurate detection of mechanical properties such as jamming, corrosion and pressure of power transmission and distribution isolating switches. Therefore, a power transmission and distribution switch status detection solution is desired.

[0070] Accordingly, considering that in the actual process of detecting the state of the power transmission and distribution switch, it is necessary to analyze and detect the various operating parameter information of the power transmission and distribution switch to determine whether the state of the power transmission and distribution switch is normal, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor. However, since the various operating parameters of the power transmission and distribution switch not only have temporal dynamic change characteristic information in the time dimension, the temporal change characteristic information of these parameters reflects the temporal change of the state of the power transmission and distribution switch, and the temporal dynamic change characteristics of the various operating parameters of the power transmission and distribution switch also have a correlation relationship in the parameter sample dimension, which brings difficulties to the accurate detection of the state of the power transmission and distribution switch. Therefore, in order to improve the accuracy of the state detection of the power transmission and distribution switch, in the technical solution of the present application, the difficulty lies in how to fully mine the dynamic correlation characteristic distribution information of the various operating parameter information of the power transmission and distribution switch in time and sample, so as to improve the expression ability of the state temporal change characteristics of the power transmission and distribution switch.

[0071] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.

[0072] The development of deep learning and neural networks provides new solutions and plans for mining the dynamic correlation characteristic distribution information of various operating parameter information of the transmission and distribution switch in time and samples.

[0073] Specifically, in the technical solution of the present application, first, the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period is obtained, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor. Next, considering that at each predetermined time point, there is a collaborative correlation between the various operating parameters of the power transmission and distribution switch to be detected, therefore, in order to improve the expression ability of the collaborative correlation features of the various operating parameters of the power transmission and distribution switch to be detected at each predetermined time point, in the technical solution of the present application, the operating parameter information of each predetermined time point is further encoded by a parameter context encoder including an embedding layer, so that after the various operating parameter information is embedded and encoded to be converted into data information of the same dimension, the global context-based collaborative correlation features of the various operating parameters of the power transmission and distribution switch to be detected at each predetermined time point are extracted, thereby obtaining multiple operating parameter context semantic understanding feature vectors.

[0074] Then, considering that the global-based collaborative correlation features of each operating parameter at each predetermined time point have a dynamic change law of time series in the time dimension, and this dynamic change law of time series has different change characteristic information in different operating parameter sample dimensions and time dimensions. Therefore, in order to enhance the adequacy of the expression of the time series collaborative correlation features of each operating parameter, so as to accurately detect the state of the power transmission and distribution switch to be detected, in the technical solution of the present application, the multiple operating parameter context semantic understanding feature vectors are further arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model for processing to obtain a classification feature map. In particular, here, the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates to extract the multi-scale operating parameter time series collaborative correlation feature distribution information under different parameter sample dimensions and different time spans.

[0075] Furthermore, the classification feature graph is passed through a classifier to obtain a classification result indicating whether the state of the power transmission and distribution switch to be detected is normal. That is, in the technical solution of the present application, the label of the classifier includes the state of the power transmission and distribution switch to be detected is normal (first label), and the state of the power transmission and distribution switch to be detected is abnormal (second label), wherein the classifier determines which classification label the classification feature graph belongs to through a soft maximum function. If the classification label is the first label, it is determined that the state of the power transmission and distribution switch to be detected is normal. If the classification label is the second label, it is determined that the state of the power transmission and distribution switch to be detected is abnormal. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether the state of the power transmission and distribution switch to be detected is normal". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the state of the power transmission and distribution switch is normal is actually converted into a binary classification class probability distribution that conforms to the law of nature through the classification label. In essence, the physical meaning of the natural probability distribution of the label is used, rather than the linguistic text meaning of "whether the state of the power transmission and distribution switch to be detected is normal". It should be understood that in the technical solution of the present application, the classification label of the classifier is the detection judgment label of whether the state of the power transmission and distribution switch to be detected is normal. Therefore, after obtaining the classification result, the state of the power transmission and distribution switch can be accurately detected based on the classification result, and the result information of whether the state of the power transmission and distribution switch is normal can be displayed on the screen.

[0076] In particular, in the technical solution of the present application, here, for each classification feature matrix arranged along the channel dimension of the classification feature map, since it is obtained by fusing the feature matrices obtained by the first convolutional neural network model and the second convolutional neural network model using the dilated convolution kernels with different dilation rates, and the dilated convolution kernels with different dilation rates of the first convolutional neural network model and the second convolutional neural network model will make the feature matrices obtained therefrom misaligned in the feature space distribution, which results in the existence of outliers that deviate from the overall feature distribution in the classification feature matrix obtained after the fusion, thereby affecting the training effect of the model when classification is performed through the classifier.

[0077] Therefore, in the technical solution of the present application, it is preferred to first convert each classification feature matrix into a classification diagonal matrix, for example, , and then perform bit-by-bit shift correlation matching optimization of intrinsic unitization, expressed as:

[0078]

[0079]

[0080] arrive is the classification diagonal matrix After eigendecomposition, we get eigenvalues, For the The eigenvalues ​​are arranged along the diagonal to obtain the eigennormalized matrix, which is also a diagonal matrix. is the intrinsic unit matrix With the classification diagonal matrix The distance between.

[0081] That is, by classifying the diagonal matrix based on The eigenvalue matrix obtained by the eigendecomposition of To classify the diagonal matrix Perform bitwise shift association and sort the diagonal matrix Matching the feature association relationship relative to the projection distance in the intrinsic unit space can solve the problem of model parameters being lost during back propagation due to the classification diagonal matrix The mismatch problem of optimization direction caused by the weak correlation distribution of local abnormal features is avoided, and the eigenvalues ​​at the edge of the classification target domain of the classification feature map composed of the classification feature matrix are prevented from being mismatched and constrained in the opposite optimization direction, resulting in poor training effect. In this way, the state detection of the power transmission and distribution switch can be accurately performed, so as to timely and accurately perform the mechanical performance detection and evaluation of the jamming, rusting, pressure, etc. of the power transmission and distribution disconnector.

[0082] Figure 1 FIG. 1 is an application scenario diagram of the power transmission and distribution switch state detection method according to an embodiment of the present application. Figure 1 As shown, in this application scenario, first, the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period is obtained (for example, Figure 1 D), wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor, and then the operating parameter information at each predetermined time point is input into a server deployed with a power transmission and distribution switch state detection algorithm (for example, Figure 1 S), wherein the server is capable of using the power transmission and distribution switch status detection algorithm to process the operating parameter information of each predetermined time point to obtain a classification result indicating whether the status of the power transmission and distribution switch to be detected is normal.

[0083] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0084] Figure 2 FIG. 1 is a flow chart of a method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application. Figure 2 As shown, according to the embodiment of the present application, the power transmission and distribution switch state detection method includes the steps of: S110, obtaining the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor; S120, respectively passing the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors; S130, arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and then inputting the two-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates; S140, passing the classification feature map through a classifier to obtain a classification result, and the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and, S150, displaying the classification result on the screen.

[0085] Figure 3 FIG. 1 is a schematic diagram of the architecture of a method for detecting the state of a power transmission and distribution switch according to an embodiment of the present application. Figure 3As shown, in the network architecture, first, the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period is obtained, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor; then, the operating parameter information of each predetermined time point is respectively passed through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors; then, the plurality of operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates; then, the classification feature map is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; finally, the classification result is displayed on the screen.

[0086] More specifically, in step S110, the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period is obtained, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor, and the voltage of the motor. In the actual process of detecting the state of the power transmission and distribution switch, it is necessary to analyze and detect the various operating parameter information of the power transmission and distribution switch to determine whether the state of the power transmission and distribution switch is normal, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor, and the voltage of the motor. However, since the various operating parameters of the power transmission and distribution switch not only have time-series dynamic change characteristic information in the time dimension, the time-series change characteristic information of these parameters reflects the time-series change of the state of the power transmission and distribution switch, and the time-series dynamic change characteristics of the various operating parameters of the power transmission and distribution switch also have a correlation relationship in the parameter sample dimension, which brings difficulties to the accurate detection of the state of the power transmission and distribution switch. Therefore, in the technical solution of the present application, the dynamic correlation feature distribution information of each operating parameter information of the power transmission and distribution switch in time and sample is mined to improve the expression ability of the state timing change characteristics of the power transmission and distribution switch, thereby improving the accuracy of state detection of the power transmission and distribution switch.

[0087] More specifically, in step S120, the operating parameter information at each predetermined time point is respectively passed through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors. Considering that at each predetermined time point, there is a collaborative correlation between the various operating parameters of the power transmission and distribution switch to be detected, therefore, in order to improve the expression ability of the collaborative correlation features of the various operating parameters of the power transmission and distribution switch to be detected at each predetermined time point, in the technical solution of the present application, the operating parameter information at each predetermined time point is further encoded through a parameter context encoder including an embedding layer, so that after the various operating parameter information is embedded and converted into data information of the same dimension, the global context collaborative correlation features of the various operating parameters of the power transmission and distribution switch to be detected at each predetermined time point are extracted, thereby obtaining a plurality of operating parameter context semantic understanding feature vectors.

[0088] It should be understood that through the context encoder, the relationship between a certain word segmentation and other word segmentations in the vector representation sequence can be analyzed to obtain the corresponding feature information. The context encoder aims to mine the hidden patterns between the contexts in the word sequence. Optionally, the encoder includes: CNN (Convolutional Neural Network), Recursive NN (Recursive Neural Network), Language Model, etc. The CNN-based method has a better extraction effect on local features, but it is not effective for the long-term dependency problem in sentences. Therefore, the encoder based on Bi-LSTM (Long Short-Term Memory) is widely used. Recursive NN treats sentences as tree structures rather than sequences. In theory, it has stronger representation capabilities, but it has weaknesses such as difficulty in sample annotation, easy gradient disappearance in deep layers, and difficulty in parallel calculation. Therefore, it is rarely used in practical applications. Transformer is a widely used network structure that has the characteristics of both CNN and RNN. It has a good extraction effect on global features and has certain advantages in parallel computing compared to RNN (Recurrent Neural Network).

[0089] Accordingly, in a specific example, Figure 4As shown, the operating parameter information of each predetermined time point is respectively passed through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors, including: S121, performing word segmentation processing on the operating parameter information of each predetermined time point to convert the operating parameter information of each predetermined time point into a first word sequence composed of a plurality of words; S122, using the embedding layer of the parameter context encoder including the embedding layer to map each word in the first word sequence to a word vector to obtain a sequence of first word vectors; and, S123, using the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the plurality of operating parameter context semantic understanding feature vectors.

[0090] Accordingly, in a specific example, Figure 5 As shown, the converter of the parameter context encoder including the embedding layer is used to perform global context semantic encoding on the sequence of the first word vectors to obtain the multiple running parameter context semantic understanding feature vectors, including: S1231, arranging the sequence of the first word vectors in one dimension to obtain a first global word sequence feature vector; S1232, calculating the product between the first global word sequence feature vector and the transposed vector of each word vector in the sequence of the first word vectors to obtain multiple first self-attention association matrices; S1233, respectively normalizing each first self-attention association matrix in the multiple first self-attention association matrices to obtain multiple first standardized self-attention association matrices; S1234, passing each first standardized self-attention association matrix in the multiple first standardized self-attention association matrices through a Softmax classification function to obtain multiple first probability values; and, S1235, respectively weighting each first word vector in the sequence of the first word vectors using each first probability value in the multiple first probability values ​​as a weight to obtain the multiple running parameter context semantic understanding feature vectors.

[0091] More specifically, in step S130, the plurality of operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use dilated convolution kernels with different dilation rates. Considering that the global-based collaborative association features of the various operating parameters at the predetermined time points have a dynamic change law of time series in the time dimension, and this dynamic change law of time series has different change feature information in different operating parameter sample dimensions and time dimensions. Therefore, in order to enhance the expression adequacy of the time series collaborative association features of the various operating parameters so as to accurately detect the state of the power transmission and distribution switch to be detected, in the technical solution of the present application, the plurality of operating parameter context semantic understanding feature vectors are further arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model for processing to obtain a classification feature map. In particular, here, the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates to extract multi-scale operating parameter temporal collaborative correlation feature distribution information under different parameter sample dimensions and different time spans.

[0092] It should be understood that Convolutional Neural Network (CNN) is an artificial neural network that is widely used in image recognition and other fields. Convolutional neural network may include input layer, hidden layer and output layer, wherein the hidden layer may include convolution layer, pooling layer, activation layer and fully connected layer, etc. The previous layer performs corresponding operations according to the input data and outputs the operation results to the next layer. The input initial data is processed through multiple layers of operations to obtain a final result.

[0093] Accordingly, in a specific example, Figure 6As shown, the multiple operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and then input into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, including: S131, using the first convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a first classification feature matrix, wherein the first convolutional neural network model has a first dilated convolution kernel with a first dilated rate; S132, using the second convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a second classification feature matrix, wherein the second convolutional neural network model has a second dilated convolution kernel with a second dilated rate, and the first dilated rate is different from the second dilated convolution kernel; and, S133, fusing the first classification feature matrix and the second classification feature matrix to obtain the classification feature map.

[0094] More specifically, in step S140, the classification feature graph is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal. That is, in the technical solution of the present application, the label of the classifier includes the state of the power transmission and distribution switch to be detected is normal (first label), and the state of the power transmission and distribution switch to be detected is abnormal (second label), wherein the classifier determines to which classification label the classification feature graph belongs through a soft maximum function. It should be understood that in the technical solution of the present application, the classification label of the classifier is a detection judgment label of whether the state of the power transmission and distribution switch to be detected is normal. Therefore, after obtaining the classification result, the state detection of the power transmission and distribution switch can be accurately performed based on the classification result, and the result information of whether the state of the power transmission and distribution switch is normal can be displayed on the screen.

[0095] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

[0096] More specifically, in step S150, the classification result is displayed on the screen.

[0097] Accordingly, in a specific example, the power transmission and distribution switch state detection method further includes training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier; wherein, Figure 7 As shown, the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier are trained, including: S210, obtaining training data, the training data including training operation parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, and the true value of whether the state of the power transmission and distribution switch to be detected is normal; S220, respectively passing the training operation parameter information of each predetermined time point through the parameter context encoder including the embedding layer to obtain multiple training operation parameter context semantic understanding feature vectors; S230, arranging the multiple training operation parameter context semantic understanding feature vectors into a training two-dimensional feature matrix and then inputting the matrix into the The dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model is used to obtain a training classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates; S240, feature distribution optimization is performed on the training classification feature map to obtain an optimized training classification feature map; S250, the optimized training classification feature map is passed through the classifier to obtain a classification loss function value; and, S260, based on the classification loss function value and through back propagation of gradient descent, the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier are trained.

[0098] Accordingly, in a specific example, Figure 8 As shown, the feature distribution of the training classification feature map is optimized to obtain an optimized training classification feature map, including: S241, expanding the training classification feature map along the channel dimension into multiple classification feature matrices; S242, converting each classification feature matrix into a classification diagonal matrix to obtain multiple classification diagonal matrices; S243, performing bit-by-bit shift association matching optimization of the multiple classification diagonal matrices with intrinsic unitization to obtain multiple optimized classification diagonal matrices; and, S244, reconstructing the dimensions of the multiple optimized classification diagonal matrices to obtain the optimized training classification feature map.

[0099] In particular, in the technical solution of the present application, here, for each classification feature matrix arranged along the channel dimension of the classification feature map, since it is obtained by fusing the feature matrices obtained by the first convolutional neural network model and the second convolutional neural network model respectively using the dilated convolution kernels with different dilation rates, and the dilated convolution kernels with different dilation rates of the first convolutional neural network model and the second convolutional neural network model will cause the feature matrices obtained therefrom to be misaligned in the feature space distribution, which results in the existence of outliers that deviate from the overall feature distribution in the classification feature matrix obtained after fusion, thereby affecting the training effect of the model when classification is performed through the classifier. Therefore, in the technical solution of the present application, it is preferred to first convert each classification feature matrix into a classification diagonal matrix, and then perform bit-by-bit shift association matching optimization of intrinsic unitization.

[0100] Accordingly, in a specific example, performing bit-by-bit shift association matching optimization of intrinsic unitization on the multiple classified diagonal matrices to obtain multiple optimized classified diagonal matrices includes: performing bit-by-bit shift association matching optimization of intrinsic unitization on the multiple classified diagonal matrices according to the following enhanced formula to obtain the multiple optimized classified diagonal matrices; wherein the enhanced formula is:

[0101]

[0102]

[0103] in, is the diagonal matrix of each classification, arrive is obtained by performing eigendecomposition on the diagonal matrices of each classification eigenvalues, For the The eigenvalues ​​are arranged along the diagonal to obtain the eigennormal matrix, and and are all diagonal matrices, is the distance between the intrinsic unit matrix and each of the classification diagonal matrices, represents matrix multiplication, represents matrix addition, It means point multiplication by position. A diagonal matrix is ​​classified for each of the optimizations.

[0104] That is, by performing bit-by-bit shift association on the classification diagonal matrix based on the intrinsic unitization matrix obtained by the eigendecomposition of the classification diagonal matrix, and matching the feature association relationship with the projection distance of the classification diagonal matrix relative to the intrinsic unitization space, the optimization direction mismatch problem caused by the weak correlation distribution of the local abnormal features of the classification diagonal matrix during back propagation of the model parameters can be solved, and the eigenvalues ​​of the classification feature graph composed of the classification feature matrix at the edge of the classification target domain are prevented from being mismatched and constrained in the opposite optimization direction, resulting in poor training effect. In this way, the state detection of the power transmission and distribution switch can be accurately performed, so as to timely and accurately perform the mechanical performance detection and evaluation of the power transmission and distribution disconnector, such as jamming, rusting, and pressure.

[0105] Accordingly, in a specific example, the optimized training classification feature map is passed through the classifier to obtain a classification loss function value, including: the classifier processes the optimized training classification feature map with the following classification formula to generate a classification result, wherein the classification formula is: ,in represents projecting the optimized training classification feature map into a vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the classification result and the true value as the classification loss function value.

[0106] In summary, based on the power transmission and distribution switch state detection method of the embodiment of the present application, the operating parameter information of each predetermined time point is first passed through the parameter context encoder to obtain multiple operating parameter context semantic understanding feature vectors, then the multiple operating parameter context semantic understanding feature vectors are arranged into a two-dimensional feature matrix and input into a dual-branch network structure to obtain a classification feature map, and finally, the classification feature map is passed through a classifier to obtain a classification result indicating whether the state of the power transmission and distribution switch to be detected is normal, and the classification result is displayed on the screen. In this way, accurate power transmission and distribution switch state detection can be achieved.

[0107] Fig. 9 FIG. 1 is a block diagram of a power transmission and distribution switch state detection system 100 according to an embodiment of the present application. Fig. 9As shown, according to the embodiment of the present application, the power transmission and distribution switch state detection system 100 includes: a data acquisition module 110, which is used to obtain the operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor; a parameter context encoding module 120, which is used to respectively pass the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain multiple operating parameter context semantic understanding feature vectors; a dual-division A convolution module 130 is used to arrange the multiple operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and then input the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates; a classification module 140 is used to pass the classification feature map through a classifier to obtain a classification result, and the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and a display module 150 is used to display the classification result on a screen.

[0108] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, as Fig.10 As shown, the parameter context encoding module 120 includes: a word segmentation unit 121, used to perform word segmentation processing on the operating parameter information of each predetermined time point to convert the operating parameter information of each predetermined time point into a first word sequence composed of multiple words; a word embedding unit 122, used to use the embedding layer of the parameter context encoder including the embedding layer to map each word in the first word sequence to a word vector to obtain a sequence of first word vectors; and a context semantic encoding unit 123, used to use the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the multiple operating parameter context semantic understanding feature vectors.

[0109] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, the context semantic encoding unit 123 is used to: arrange the sequence of the first word vectors in one dimension to obtain a first global word sequence feature vector; calculate the product between the first global word sequence feature vector and the transposed vector of each word vector in the sequence of the first word vector to obtain multiple first self-attention association matrices; standardize each of the multiple first self-attention association matrices to obtain multiple first standardized self-attention association matrices; pass each of the multiple first standardized self-attention association matrices through the Softmax classification function to obtain multiple first probability values; and, weight each first word vector in the sequence of the first word vector using each of the multiple first probability values ​​as a weight to obtain the multiple operating parameter context semantic understanding feature vectors.

[0110] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, the dual-branch convolution module is used to: use the first convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a first classification feature matrix, wherein the first convolutional neural network model has a first void convolution kernel with a first void rate; use the second convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a second classification feature matrix, wherein the second convolutional neural network model has a second void convolution kernel with a second void rate, and the first void rate is different from the second void convolution kernel; and, fuse the first classification feature matrix and the second classification feature matrix to obtain the classification feature map.

[0111] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, it also includes a training module for training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier; wherein the training module is used to: obtain training data, the training data including training operation parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, and the true value of whether the state of the power transmission and distribution switch to be detected is normal; respectively pass the training operation parameter information of each predetermined time point through the parameter context encoder including the embedding layer to obtain multiple training operation parameter context semantic understanding feature vectors; the multiple training operation parameter context semantic understanding feature vectors After understanding that the feature vector is arranged into a training two-dimensional feature matrix, the feature vector is input into the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model to obtain a training classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use void convolution kernels with different void rates; feature distribution optimization is performed on the training classification feature map to obtain an optimized training classification feature map; the optimized training classification feature map is passed through the classifier to obtain a classification loss function value; and, based on the classification loss function value and through back propagation of gradient descent, the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier are trained.

[0112] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, the feature distribution of the training classification feature map is optimized to obtain an optimized training classification feature map, including: expanding the training classification feature map into multiple classification feature matrices along the channel dimension; converting each classification feature matrix into a classification diagonal matrix to obtain a plurality of classification diagonal matrices; performing bit-by-bit shift association matching optimization of the multiple classification diagonal matrices with intrinsic unitization to obtain a plurality of optimized classification diagonal matrices; and, reconstructing the dimensions of the multiple optimized classification diagonal matrices to obtain the optimized training classification feature map.

[0113] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, the plurality of classification diagonal matrices are subjected to intrinsic unitization bit-by-bit shift association matching optimization to obtain a plurality of optimized classification diagonal matrices, including: the plurality of classification diagonal matrices are subjected to intrinsic unitization bit-by-bit shift association matching optimization to obtain the plurality of optimized classification diagonal matrices according to the following enhanced formula; wherein the enhanced formula is:

[0114]

[0115]

[0116] in, is the diagonal matrix of each classification, arrive is obtained by performing eigendecomposition on the diagonal matrices of each classification eigenvalues, For the The eigenvalues ​​are arranged along the diagonal to obtain the eigennormal matrix, and and are all diagonal matrices, is the distance between the intrinsic unit matrix and each of the classification diagonal matrices, represents matrix multiplication, represents matrix addition, It means point multiplication by position. A diagonal matrix is ​​classified for each of the optimizations.

[0117] In one example, in the above-mentioned power transmission and distribution switch state detection system 100, the optimized training classification feature map is passed through the classifier to obtain a classification loss function value, including: the classifier processes the optimized training classification feature map with the following classification formula to generate a classification result, wherein the classification formula is: ,in represents projecting the optimized training classification feature map into a vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the classification result and the true value as the classification loss function value.

[0118] Here, those skilled in the art can understand that the specific functions and operations of the various units and modules in the above-mentioned power transmission and distribution switch state detection system 100 have been described in the above reference. Figures 1 to 8 The description of the power transmission and distribution switch state detection method has been described in detail, and therefore, its repeated description will be omitted.

[0119] As described above, the power transmission and distribution switch state detection system 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a power transmission and distribution switch state detection algorithm. In one example, the power transmission and distribution switch state detection system 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the power transmission and distribution switch state detection system 100 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the power transmission and distribution switch state detection system 100 can also be one of the many hardware modules of the wireless terminal.

[0120] Alternatively, in another example, the power transmission and distribution switch state detection system 100 and the wireless terminal may also be separate devices, and the power transmission and distribution switch state detection system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0121] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

[0122] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0123] All or part of the software may sometimes communicate over a network, such as the Internet or other communications network. Such communications can load software from one computer device or processor to another. For example: from a server or host computer of a video target detection device to a hardware platform of a computer environment, or other computer environment that implements the system, or a system with similar functions related to providing information required for target detection. Therefore, another medium capable of transmitting software elements can also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., transmitted through cables, optical cables or air. Physical media used to carry carriers, such as cables, wireless connections or optical cables and the like, can also be considered as media that carry software. As used herein, unless limited to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media involved in the process of the processor executing any instructions.

[0124] The present application uses specific words to describe the embodiments of the present application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0125] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0126] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0127] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A method for detecting the state of a power transmission and distribution switch, characterized in that: include: Obtaining operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor; Passing the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors; Arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and inputting the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use dilated convolution kernels with different dilated rates; Passing the classification feature graph through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and Displaying the classification results on the screen; The operating parameter information of each predetermined time point is passed through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors, including: Performing word segmentation processing on the operation parameter information at each predetermined time point to convert the operation parameter information at each predetermined time point into a first word sequence consisting of a plurality of words; Mapping each word in the first word sequence to a word vector using the embedding layer of the parameter context encoder including the embedding layer to obtain a sequence of first word vectors; and Using the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the plurality of operating parameter context semantic understanding feature vectors; Using the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the plurality of running parameter context semantic understanding feature vectors, including: Arrange the sequence of the first word vectors in one dimension to obtain a first global word sequence feature vector; Calculating the product of the first global word sequence feature vector and the transposed vector of each word vector in the sequence of the first word vectors to obtain a plurality of first self-attention association matrices; Normalizing each of the plurality of first self-attention association matrices to obtain a plurality of first standardized self-attention association matrices; Passing each of the plurality of first standardized self-attention association matrices through a Softmax classification function to obtain a plurality of first probability values; and Using each first probability value in the plurality of first probability values ​​as a weight, weighting each first word vector in the sequence of the first word vectors to obtain the plurality of operating parameter context semantic understanding feature vectors; Arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and then inputting the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, including: Using the first convolutional neural network model of the dual-branch network structure to perform two-dimensional convolution encoding on the two-dimensional feature matrix to obtain a first classification feature matrix, wherein the first convolutional neural network model has a first dilated convolution kernel with a first dilated rate; Performing two-dimensional convolution encoding on the two-dimensional feature matrix using the second convolutional neural network model of the dual-branch network structure to obtain a second classification feature matrix, wherein the second convolutional neural network model has a second dilated convolution kernel with a second dilated rate, and the first dilated rate is different from the second dilated convolution kernel; and The first classification feature matrix and the second classification feature matrix are fused to obtain the classification feature map.

2. The method for detecting the state of a power transmission and distribution switch according to claim 1, characterized in that: It also includes training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier; The step of training the parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier comprises: Acquire training data, wherein the training data includes training operation parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, and a real value of whether the state of the power transmission and distribution switch to be detected is normal; Passing the training operation parameter information at each predetermined time point through the parameter context encoder including the embedding layer to obtain a plurality of training operation parameter context semantic understanding feature vectors; Arranging the plurality of training operation parameter context semantic understanding feature vectors into a training two-dimensional feature matrix and inputting the matrix into the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model to obtain a training classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use dilated convolution kernels with different dilated rates; Performing feature distribution optimization on the training classification feature graph to obtain an optimized training classification feature graph; Passing the optimized training classification feature map through the classifier to obtain a classification loss function value; and The parameter context encoder including the embedding layer, the dual-branch network structure including the first convolutional neural network model and the second convolutional neural network model, and the classifier are trained based on the classification loss function value and through back propagation of gradient descent.

3. The method for detecting the state of a power transmission and distribution switch according to claim 2, characterized in that: Optimizing the feature distribution of the training classification feature graph to obtain an optimized training classification feature graph includes: Expanding the training classification feature map into multiple classification feature matrices along the channel dimension; Converting each of the classification feature matrices into a classification diagonal matrix to obtain a plurality of classification diagonal matrices; Performing bit-by-bit shift correlation matching optimization of intrinsic unitization on the multiple classified diagonal matrices to obtain multiple optimized classified diagonal matrices; and The multiple optimized classification diagonal matrices are reconstructed in dimension to obtain the optimized training classification feature map.

4. The method for detecting the state of a power transmission and distribution switch according to claim 3, characterized in that: Performing bit-by-bit shift association matching optimization of intrinsic unitization on the multiple classified diagonal matrices to obtain multiple optimized classified diagonal matrices, including: Performing bit-by-bit shift association matching optimization of the plurality of classified diagonal matrices with intrinsic unitization according to the following enhanced formula to obtain the plurality of optimized classified diagonal matrices; Wherein, the strengthening formula is: in, are the diagonal matrices of the various classifications, arrive is obtained by performing eigendecomposition on the diagonal matrices of each classification eigenvalues, For the The eigenvalues ​​are arranged along the diagonal to obtain the eigennormal matrix, and and are all diagonal matrices, is the distance between the intrinsic unit matrix and each of the classification diagonal matrices, represents matrix multiplication, represents matrix addition, It means point multiplication by position. A diagonal matrix is ​​classified for each of the optimizations.

5. The method for detecting the state of a power transmission and distribution switch according to claim 4, characterized in that: Passing the optimized training classification feature map through the classifier to obtain a classification loss function value includes: The classifier processes the optimized training classification feature map using the following classification formula to generate a classification result, wherein the classification formula is: ,in represents projecting the optimized training classification feature map into a vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the classification result and the true value as the classification loss function value.

6. A power transmission and distribution switch state detection system for executing the power transmission and distribution switch state detection method according to any one of claims 1 to 5, characterized in that: include: A data acquisition module, used to acquire operating parameter information of the power transmission and distribution switch to be detected at multiple predetermined time points within a predetermined time period, wherein the operating parameter information includes the torque of the main shaft of the transmission mechanism, the input power of the motor, the rotation angle of the motor, the contact pressure of the contact finger of the motor, the current of the motor and the voltage of the motor; A parameter context encoding module, configured to pass the operating parameter information of each predetermined time point through a parameter context encoder including an embedding layer to obtain a plurality of operating parameter context semantic understanding feature vectors; A dual-branch convolution module, used for arranging the plurality of operating parameter context semantic understanding feature vectors into a two-dimensional feature matrix and then inputting the matrix into a dual-branch network structure including a first convolutional neural network model and a second convolutional neural network model to obtain a classification feature map, wherein the first convolutional neural network model and the second convolutional neural network model respectively use a dilated convolution kernel with different dilated rates; a classification module, used for passing the classification feature graph through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the state of the power transmission and distribution switch to be detected is normal; and The display module is used to display the classification result on the screen.

7. The power transmission and distribution switch state detection system according to claim 6, characterized in that: The parameter context encoding module comprises: A word segmentation unit, configured to perform word segmentation processing on the operation parameter information at each predetermined time point to convert the operation parameter information at each predetermined time point into a first word sequence consisting of a plurality of words; A word embedding unit, configured to map each word in the first word sequence to a word vector using the embedding layer of the parameter context encoder including the embedding layer to obtain a sequence of first word vectors; and A context semantic encoding unit is used to use the converter of the parameter context encoder including the embedding layer to perform global context semantic encoding on the sequence of the first word vectors to obtain the multiple running parameter context semantic understanding feature vectors.

Images