Electric power communication network fault diagnosis method and device based on convolutional neural network

Through the power communication network fault diagnosis method based on convolutional neural network, faults in the power communication network are identified and diagnosed, and the problem of the existing technology being unable to identify potential faults of the main communication line is solved, ensuring the stability and efficiency of power communication transmission.

CN118972262APending Publication Date: 2024-11-15YUNCHENG POWER SUPPLY COMPANY OF STATE GRID SHANXI ELECTRIC POWER

Patent Information

Application Number
CN202410968360.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing monitoring methods cannot identify and diagnose potential communication faults of the main communication line when the real-time communication line is the main communication line, resulting in the spread of communication line faults and affecting the transmission of power communication.

Method used

The fault diagnosis method of power communication network based on convolutional neural network is adopted. By performing principal component analysis of the power communication network structure, the transmission network monitoring points are deployed, the visual monitoring model is generated, and fault diagnosis based on transfer learning mechanism and convolutional neural network is pre-constructed. Model, iteratively trained and validated, identify and diagnose faults.

Benefits of technology

It realizes the analysis and diagnosis of all monitoring points and potential faults in the power communication network, avoids the spread of communication line failures, and ensures the transmission efficiency of power communication.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118972262A_ABST
    Figure CN118972262A_ABST
Patent Text Reader

Abstract

The invention discloses an electric power communication network fault diagnosis method and device based on a convolutional neural network, belongs to the technical field of electric power communication, and solves the problem that a potential communication fault of a main communication line cannot be identified and diagnosed when a real-time communication line is the main communication line in an existing monitoring method. The method comprises the steps of deploying transmission network monitoring points based on a communication transmission facility weight; collecting monitoring data of a transmission network monitoring point in real time; pre-constructing a fault diagnosis model based on the combination of a transfer learning mechanism and a convolutional neural network, and executing the fault diagnosis model to obtain a fault diagnosis result of a transmission network monitoring point in the visual monitoring model; according to the method, the fault diagnosis model based on the combination of the transfer learning mechanism and the convolutional neural network is pre-constructed, so that monitoring data identification and expansion convolution are realized, the faults are classified by using the fault type database, and further analysis and diagnosis of all monitoring point faults and potential faults in the power communication network are realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of power communication, and in particular relates to a method and device for diagnosing power communication network faults based on a convolutional neural network. Background Art

[0002] In power communication systems, faults often lead to large-scale power outages and have a wide impact, so it is crucial to diagnose and handle faults quickly and accurately. With the continuous advancement of technology, traditional fault handling methods have been gradually replaced by intelligent technologies, which not only improve the efficiency and accuracy of fault handling, but also greatly reduce operation and maintenance costs.

[0003] Online monitoring technology is also an indispensable part of fault diagnosis. Modern power systems are equipped with a large number of sensors and data acquisition devices, which can monitor the operating status of power equipment in real time and transmit data to the background center for analysis and processing through wireless communication technology. Among them, Chinese patent CN117792957A discloses a monitoring method and system for power dispatching communication faults, including: obtaining a real-time communication line for power dispatching communication; when the acquired real-time communication line is the main communication line, there is no power dispatching communication fault; when the acquired real-time communication line is a backup communication line, there is a power dispatching communication fault; when there is a power dispatching communication fault, the main communication line triggers communication fault processing, interrupts the communication data transmission on the main communication circuit, and the backup communication line receives network element data in real time, performs security identity authentication of the communication line and updates the network element data according to the received network element data, and completes the communication data transmission when the power dispatching communication fault occurs; however, the existing monitoring method cannot identify and diagnose the potential communication fault of the main communication line when the real-time communication line is the main communication line, thereby causing the communication line fault to spread and affect the power communication transmission. In response to the above problems, we propose a power communication network fault diagnosis method and device based on a convolutional neural network. Summary of the invention

[0004] The purpose of the present invention is to address the shortcomings of the prior art and provide a power communication network fault diagnosis method and device based on a convolutional neural network, which solves the problem that the existing monitoring method is unable to identify and diagnose potential communication faults of the main communication line when the real-time communication line is the main communication line, thereby causing the communication line fault to spread and affect power communication transmission.

[0005] The present invention is implemented in this way: a power communication network fault diagnosis method based on a convolutional neural network, the power communication network fault diagnosis method based on a convolutional neural network comprises:

[0006] Traverse the power communication network, perform principal component analysis on the power communication network structure, obtain the weights of communication transmission facilities, and deploy transmission network monitoring points based on the weights of communication transmission facilities;

[0007] Collect monitoring data from transmission network monitoring points in real time, and generate a visual monitoring model based on the transmission network monitoring points and monitoring data;

[0008] Building a fault type database based on the expert database, traversing the fault type database, indexing the fault data set in the fault type database, and dividing the fault data set into a training set and a test set;

[0009] Pre-build a fault diagnosis model based on the transfer learning mechanism combined with a convolutional neural network, obtain training sets and test sets, iteratively train the fault diagnosis model based on the training sets and test sets, complete model training and verification, and output a converged fault diagnosis model;

[0010] The monitoring data is loaded, and the fault diagnosis model is executed with the monitoring data as input to obtain the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

[0011] The method for deploying transmission network monitoring points based on communication transmission facility weights specifically includes:

[0012] Identify the transmission lines, data lines, power supply lines, and transmission facilities in the electric power communication network, and collect the operation logs and operation flow data of the transmission lines, data lines, power supply lines, and transmission facilities;

[0013] Load the operation log and operation flow data, and extract three feature vectors, namely, attribute feature, transmission feature, and line feature, from the operation log and operation flow data based on attribute feature, transmission frequency, and line association;

[0014] The mean shift clustering algorithm is used to cluster the three feature vectors, and the feature vectors with high feature similarity are divided into the same fault indicator factor cluster. The feature similarity uses Manhattan distance as the similarity measurement standard.

[0015] The method for deploying transmission network monitoring points based on communication transmission facility weights specifically includes:

[0016] Load multiple groups of fault indicator factor clusters, calculate the factor cluster characteristic values ​​of the fault indicator factor clusters based on principal component analysis, and generate factor cluster characteristic queues;

[0017] Select the largest attribute feature factor cluster, transmission feature factor cluster, and line feature factor cluster in the factor cluster feature queue respectively;

[0018] The variance contribution rates of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster are used to determine the principal component analysis weight coefficient, and the weight of the communication transmission facility is set based on the principal component analysis weight coefficient;

[0019] Deploy transmission network monitoring points using weighted discriminant functions and weights of communication transmission facilities;

[0020] Among them, when deploying the transmission network monitoring point, the importance of the communication transmission facility is calculated based on the weighted discriminant function, and an important threshold is preset to determine whether the importance of the communication transmission facility exceeds the preset important threshold. If it exceeds the preset important threshold, the transmission network monitoring point is deployed for the communication transmission facility;

[0021]

[0022] Among them, C x is the importance of communication transmission facilities, q(x) is the weight of communication transmission facilities, n is the number of communication transmission facilities, and g is the coupling coefficient of the weighted discriminant function;

[0023]

[0024] Among them, q(x) is the weight of the communication transmission facility, n is the number of communication transmission facilities, and x n is the factor cluster eigenvalue, and λ is the data dimension of the factor cluster eigenvalue.

[0025] The method for generating a visual monitoring model based on transmission network monitoring points and monitoring data specifically includes:

[0026] Obtain the coordinate data and monitoring data of the transmission network monitoring points, and perform missing value processing, outlier processing and normalization processing on the coordinate data and monitoring data of the monitoring points;

[0027] Use 3D modeling tools to identify the coordinate data of monitoring points and build an initial visualization model;

[0028] Load the processed monitoring data, identify the monitoring data, classify the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points, and obtain a monitoring data set with classification labels;

[0029] The initial visualization model is edited and adjusted based on the monitoring data set containing the partition labels to generate a visualization monitoring model.

[0030] The method for pre-building a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network specifically includes:

[0031] The VGG19 model is pre-trained using the training set to obtain a pre-trained VGG19 model, and a residual network structure is introduced to replace the fully connected layer of the VGG19 model;

[0032] Improve the pooling layer of the VGG19 model based on the transfer learning mechanism, where the pooling layer is improved through the fine-tuning layer of the transfer learning mechanism model;

[0033] Customize and improve the convolutional neural network of the VGG19 model, and place the convolutional neural network after the convolutional layer of the VGG19 model to strengthen the convolution results of the VGG19 model;

[0034] Select Huber loss as the loss function of the VGG19 model, select Adma as the optimizer of the VGG19 model, define the learning rate and hyperparameters of the VGG19 model, and complete the pre-construction of the fault diagnosis model.

[0035] The method for iteratively training the fault diagnosis model based on the training set and the test set specifically includes:

[0036] Obtain a pre-built fault diagnosis model, a training set, and a test set, wherein the fault diagnosis model includes an input layer, a convolutional layer, a convolutional neural network, an improved pooling layer, a residual network structure, and an output layer;

[0037] Using the training set as input, the fault diagnosis model is executed, and a small amount of test set data in the target domain is used to migrate to the source domain for fine-tuning training;

[0038] The training set is input into the convolutional neural network, and the convolutional neural network performs dilated convolution on the training set data to obtain the dilated convolution result;

[0039] Load the dilated convolution result, input the dilated convolution result into the improved pooling layer, and the improved pooling layer performs dimensionality reduction processing on the dilated convolution result to obtain the dimensionality reduction output result;

[0040] Input the dimension reduction output result into the residual network structure to obtain the training diagnosis result;

[0041] Determine whether the training diagnosis result meets the preset diagnosis threshold. If it meets the preset diagnosis threshold, output the current fault diagnosis model;

[0042] Load the test set and use the test set to verify whether the accuracy of the fault diagnosis model meets the preset accuracy;

[0043] If the preset accuracy is met, the verified fault diagnosis model is output.

[0044] The method of executing the fault diagnosis model to obtain the fault diagnosis result of the transmission network monitoring point in the visual monitoring model specifically includes:

[0045] The monitoring data of the transmission network monitoring points in the visual monitoring model are obtained, and the monitoring data are input into the convolutional neural network. The convolutional neural network performs dilation convolution on the monitoring data to obtain the fault state matrix.

[0046] Load the fault state matrix, improve the pooling layer to reduce the dimension of the fault state matrix, and obtain the reduced-dimensional fault feature set;

[0047] Identify the reduced-dimensionality fault feature set, input the reduced-dimensionality fault feature set into the residual network structure, and obtain the fault abnormality degree corresponding to the reduced-dimensionality fault feature set of the monitoring point;

[0048] Among them, the fault abnormality is calculated by formula (3):

[0049]

[0050] Among them, Y(x) is the fault abnormality, m is the data volume of the reduced-dimensional fault feature set, and x i is the input value of the dimension-reduced fault feature set, k is the number of convolution kernels in the convolutional neural network, α is the convolution expansion factor of the convolutional neural network, and w is the gradient value of the residual network structure;

[0051] w=sign(x i ) β *max(x i |-σ,0)(4)

[0052] Among them, w is the gradient value of the residual network structure, β is the gradient coefficient of the activation function of the residual network structure, and σ is the preset threshold parameter of the residual network structure.

[0053] Based on the fault abnormality degree, it is judged whether there is an abnormality at the current monitoring point. If there is an abnormality, the features of the reduced-dimensional fault feature set are identified, and the fault is classified using the fault type database to achieve fault diagnosis.

[0054] On the other hand, the present invention also provides a power communication network fault diagnosis device based on a convolutional neural network, and the power communication network fault diagnosis device based on a convolutional neural network comprises:

[0055] Monitoring point deployment module, which is used to traverse the power communication network, perform principal component analysis on the power communication network structure, obtain the weight of communication transmission facilities, and deploy the transmission network monitoring points based on the weight of communication transmission facilities;

[0056] A visualization presentation module is used to collect monitoring data of the transmission network monitoring points in real time and generate a visualization monitoring model based on the transmission network monitoring points and monitoring data;

[0057] Database construction module: The database construction module constructs a fault type database based on the expert database, traverses the fault type database, indexes the fault data set in the fault type database, and divides the fault data set into a training set and a test set;

[0058] The diagnostic model generation module is used to pre-build a fault diagnosis model based on the combination of transfer learning mechanism and convolutional neural network, obtain training sets and test sets, iteratively train the fault diagnosis model based on the training sets and test sets, complete model training and verification, and output a converged fault diagnosis model;

[0059] The result output module is used to load monitoring data, take the monitoring data as input, execute the fault diagnosis model, and obtain the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

[0060] The monitoring point deployment module includes:

[0061] The data acquisition unit is used to identify the transmission lines, data lines, power supply lines and transmission facilities in the electric power communication network, and collect the operation logs and operation flow data of the transmission lines, data lines, power supply lines and transmission facilities, and extract three feature vectors of attribute features, transmission features and line features from the operation logs and operation flow data based on attribute features, transmission frequency and line association;

[0062] The feature queue generation unit uses the mean shift clustering algorithm to perform clustering operations on the three feature vectors, divides the feature vectors with high feature similarity into the same fault indicator factor cluster, calculates the factor cluster characteristic value of the fault indicator factor cluster based on principal component analysis, and generates a factor cluster feature queue;

[0063] A weight setting unit selects the largest attribute characteristic factor cluster, transmission characteristic factor cluster, and line characteristic factor cluster in the factor cluster characteristic queue respectively, determines the principal component analysis weight coefficient by using the variance contribution rate of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster, sets the weight of the communication transmission facility based on the principal component analysis weight coefficient, and deploys the transmission network monitoring points by using the weight discriminant function and the weight of the communication transmission facility.

[0064] The visual presentation module includes:

[0065] A data processing unit, the data processing unit is used to obtain the coordinate data of the monitoring points of the transmission network and the monitoring data, and perform missing value processing, outlier processing and normalization processing on the coordinate data of the monitoring points and the monitoring data;

[0066] A visualization construction unit, which uses a three-dimensional modeling tool to identify the coordinate data of the monitoring points, constructs an initial visualization model, loads the processed monitoring data, identifies the monitoring data, and classifies the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points to obtain a monitoring data set containing classification labels;

[0067] Model optimization unit,The model optimization unit edits and adjusts the initial visualization model based on the monitoring data set containing the partition labels to generate a visualization monitoring model.

[0068] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0069] In the embodiment of the present invention, a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network is pre-constructed to realize monitoring data recognition and dilated convolution, obtain a fault state matrix, and classify faults using a fault type database, thereby realizing analysis and diagnosis of all monitoring point faults and potential faults in the power communication network, avoiding the spread of communication line faults, and ensuring the efficiency of power communication transmission.

[0070] In the embodiment of the present invention, three characteristic vectors, namely, attribute characteristics, transmission characteristics, and line characteristics, are extracted from operation logs and operation flow data, and then a mean shift clustering algorithm is used to perform a clustering operation on the three characteristic vectors. The factor cluster characteristic values ​​of the fault indicator factor cluster are calculated through principal component analysis, and the principal component analysis weight coefficient is determined based on the variance contribution rate of the indicator factor, thereby ensuring the accuracy of the weighting of communication transmission facilities and facilitating the deployment of transmission network monitoring points.

[0071] In an embodiment of the present invention, the fault diagnosis model is a VGG19 model based on a transfer learning mechanism and a convolutional neural network. At the same time, the pooling layer of the VGG19 model is improved based on the transfer learning mechanism, thereby reducing the data dimension and retaining feature information. At the same time, the convolutional neural network of the VGG19 model is customized and improved to strengthen the convolution result of the VGG19 model, thereby avoiding the overfitting phenomenon of the VGG19 model when training on a small data set, thereby ensuring the convergence of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic diagram of the implementation flow of the power communication network fault diagnosis method based on convolutional neural network provided by the present invention.

[0073] Figure 2 It is a schematic diagram of the implementation flow of the method for deploying transmission network monitoring points based on communication transmission facility weights provided by the present invention.

[0074] Figure 3It is a schematic diagram of the implementation process of the method for generating a visual monitoring model based on transmission network monitoring points and monitoring data provided by the present invention.

[0075] Figure 4 It is a schematic diagram of the implementation process of the method for pre-building a fault diagnosis model based on the combination of a transfer learning mechanism and a convolutional neural network provided by the present invention.

[0076] Figure 5 It is a schematic diagram of the implementation flow of the method for iteratively training a fault diagnosis model based on a training set and a test set provided by the present invention.

[0077] Figure 6 It is a schematic diagram of the implementation flow of the method for executing the fault diagnosis model provided by the present invention to obtain the fault diagnosis results of the transmission network monitoring point in the visual monitoring model.

[0078] Figure 7 It is a schematic diagram of the framework of the power communication network fault diagnosis device based on convolutional neural network provided by the present invention.

[0079] Figure 8 It is a schematic diagram of the framework of the monitoring point deployment module provided by the present invention.

[0080] Fig. 9 It is a schematic diagram of the framework of the visualization presentation module provided by the present invention. DETAILED DESCRIPTION

[0081] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0082] When the real-time communication line is the main communication line, the existing monitoring method cannot identify and diagnose the potential communication fault of the main communication line, thereby causing the communication line fault to spread and affect the power communication transmission. In view of the above problems, we propose a power communication network fault diagnosis method and device based on convolutional neural network. In short, the method includes deploying transmission network monitoring points based on the weight of communication transmission facilities, generating a visual monitoring model based on the transmission network monitoring points and monitoring data, indexing the fault data set in the fault type database, dividing the fault data set into a training set and a test set, iteratively training the fault diagnosis model based on the training set and the test set, completing model training and verification, outputting a converged fault diagnosis model, executing the fault diagnosis model, and obtaining the fault diagnosis result of the transmission network monitoring point in the visual monitoring model. In the embodiment of the present invention, by pre-constructing a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network, the monitoring data is identified and dilated, and a fault state matrix is ​​obtained. The fault is classified using the fault type database, and then the faults and potential faults of all monitoring points in the power communication network are analyzed and diagnosed, the spread of communication line faults is avoided, and the efficiency of power communication transmission is guaranteed.

[0083] The embodiment of the present invention provides a method for diagnosing faults in a power communication network based on a convolutional neural network. Figure 1 The schematic diagram of the implementation process of the power communication network fault diagnosis method based on convolutional neural network is shown. The power communication network fault diagnosis method based on convolutional neural network specifically includes:

[0084] Step S10, traversing the power communication network, performing principal component analysis on the power communication network structure, obtaining communication transmission facility weights, and deploying transmission network monitoring points based on the communication transmission facility weights;

[0085] It should be noted that the power communication network includes a backbone communication network, an access communication network, communication facilities, transmission facilities, and distribution facilities. Among them, the backbone communication network includes a transmission network, a data network, and a support network, and the access communication network includes a power transmission and transformation communication network and a distribution communication network.

[0086] Step S20, collecting monitoring data of the transmission network monitoring points in real time, and generating a visual monitoring model based on the transmission network monitoring points and the monitoring data;

[0087] In this embodiment, the monitoring data includes but is not limited to grid operation parameters, equipment operation parameters, and meteorological environment data, and the grid operation parameters include but are not limited to main / branch line temperature and load, insulator leakage current, main / branch line transmission frequency, and the equipment operation parameters include but are not limited to power / communication facility operating parameters, specifications, and video monitoring parameters.

[0088] Step S30, building a fault type database based on the expert database, traversing the fault type database, indexing the fault data set in the fault type database, and dividing the fault data set into a training set and a test set, wherein the ratio of the training set to the test set is 7:2;

[0089] In this embodiment, the fault type database is constructed based on the expert database, and the expert database is used to collect fault reports and processing records, and the processing records include fault type, occurrence time, occurrence node, impact range, fault cause and solution. The expert database divides fault types into hardware faults, software defects, configuration errors, network attacks, etc. based on expert scoring, and the fault type database is a structured database that stores fault cases and expert knowledge. The database should support efficient data retrieval and update to ensure the real-time and accuracy of information.

[0090] Step S40, pre-constructing a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network, obtaining a training set and a test set, iteratively training the fault diagnosis model based on the training set and the test set, completing model training and verification, and outputting a converged fault diagnosis model;

[0091] Step S50, loading monitoring data, taking the monitoring data as input, executing the fault diagnosis model, and obtaining the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

[0092] In the embodiment of the present invention, a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network is pre-constructed to realize monitoring data recognition and dilated convolution, obtain a fault state matrix, and classify faults using a fault type database, thereby realizing analysis and diagnosis of all monitoring point faults and potential faults in the power communication network, avoiding the spread of communication line faults, and ensuring the efficiency of power communication transmission.

[0093] The embodiment of the present invention provides a method for deploying transmission network monitoring points based on communication transmission facility weights. Figure 2 The schematic diagram of the implementation process of the method for deploying transmission network monitoring points based on the weight of communication transmission facilities is shown. The method for deploying transmission network monitoring points based on the weight of communication transmission facilities specifically includes:

[0094] Step S101, identifying transmission lines, data lines, power supply lines, and transmission facilities in the electric power communication network, and collecting operation logs and operation flow data of the transmission lines, data lines, power supply lines, and transmission facilities;

[0095] Step S102, loading the operation log and the operation flow data, extracting three feature vectors of attribute feature, transmission feature and line feature from the operation log and the operation flow data based on the attribute feature, transmission frequency and line association;

[0096] Step S103, clustering the three feature vectors using a mean shift clustering algorithm, and classifying feature vectors with high feature similarity into the same fault indicator factor cluster, wherein the feature similarity uses Manhattan distance as a similarity measurement standard.

[0097] It should be noted that, in the embodiment of the present invention, three feature vectors, namely, attribute features, transmission features, and line features, are extracted from operation logs and operation traffic data, and then the mean shift clustering algorithm is used to perform clustering operations on the three feature vectors. The factor cluster eigenvalues ​​of the fault indicator factor cluster are calculated through principal component analysis, and the principal component analysis weight coefficient is determined based on the variance contribution rate of the indicator factor, thereby ensuring the accuracy of the weighting of communication transmission facilities and facilitating the deployment of transmission network monitoring points.

[0098] Step S104, loading multiple groups of fault indicator factor clusters, calculating factor cluster characteristic values ​​of the fault indicator factor clusters based on principal component analysis, and generating factor cluster characteristic queues;

[0099] Step S105, selecting the largest attribute feature factor cluster, transmission feature factor cluster, and line feature factor cluster in the factor cluster feature queue respectively;

[0100] Step S106, determining the principal component analysis weight coefficient using the variance contribution rates of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster, and setting the weight of the communication transmission facility based on the principal component analysis weight coefficient;

[0101] Step S107, deploying transmission network monitoring points using a weighted discriminant function and weights of communication transmission facilities;

[0102] Among them, when deploying the transmission network monitoring point, the importance of the communication transmission facility is calculated based on the weighted discriminant function, and an important threshold is preset to determine whether the importance of the communication transmission facility exceeds the preset important threshold. If it exceeds the preset important threshold, the transmission network monitoring point is deployed for the communication transmission facility;

[0103]

[0104] Among them, C x is the importance of the communication transmission facility, q(x) is the weight of the communication transmission facility, n is the number of the communication transmission facility, in this embodiment, the number of the communication transmission facility is 10-2000, g is the coupling coefficient of the weighted discriminant function, the weighted discriminant function may be a TOPSIS function or an entropy weighted TOPSIS function, and the coupling coefficient of the weighted discriminant function may be 0.1-0.4;

[0105]

[0106] Among them, q(x) is the weight of the communication transmission facility, n is the number of communication transmission facilities, and x n is the factor cluster eigenvalue, λ is the data dimension of the factor cluster eigenvalue. In this embodiment, the data dimension can be 1-5.

[0107] The embodiment of the present invention provides a method for generating a visual monitoring model based on transmission network monitoring points and monitoring data. Figure 3 The schematic diagram of the implementation process of the method for generating a visual monitoring model based on the monitoring points and monitoring data of the transmission network is shown. The method for generating a visual monitoring model based on the monitoring points and monitoring data of the transmission network specifically includes:

[0108] Step S201, obtaining the coordinate data of the monitoring points and the monitoring data of the transmission network, and performing missing value processing, outlier processing and normalization processing on the coordinate data of the monitoring points and the monitoring data;

[0109] It should be noted that the transmission network monitoring point coordinate data may be two-dimensional or three-dimensional coordinate data of the transmission network monitoring point, and the methods for processing the monitoring point coordinate data and monitoring data include but are not limited to missing value processing, outlier processing and normalization processing.

[0110] Step S202, using a three-dimensional modeling tool to identify the coordinate data of the monitoring points and construct an initial visualization model;

[0111] In this embodiment, the 3D modeling tool may be PhotoCatch, KIRI Engine, COLMAP, and Meshroom.

[0112] Step S203, loading the processed monitoring data, identifying the monitoring data, and classifying the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points to obtain a monitoring data set containing classification labels;

[0113] Step S204: edit and adjust the initial visualization model based on the monitoring data set containing the division labels to generate a visualization monitoring model.

[0114] In an embodiment of the present invention, by performing missing value processing, outlier processing and normalization processing on the monitoring point coordinate data and the monitoring data, it is convenient to identify the monitoring point coordinate data. At the same time, editing and adjusting the initial visualization model based on the monitoring data set containing the partitioning labels can present the visualization monitoring model in real time, which facilitates the correlation tracing and analysis of monitoring point anomalies.

[0115] The embodiment of the present invention provides a method for pre-building a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network. Figure 4A schematic diagram of the implementation process of a method for pre-building a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network is shown. The method for pre-building a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network specifically includes:

[0116] Step S301, pre-training the VGG19 model using the training set to obtain the pre-trained VGG19 model, and introducing a residual network structure to replace the fully connected layer of the VGG19 model;

[0117] Step S302, improving the pooling layer of the VGG19 model based on the transfer learning mechanism, wherein the pooling layer is improved by the fine-tuning layer of the transfer learning mechanism model;

[0118] Step S303, customizing a convolutional neural network of the improved VGG19 model, placing the convolutional neural network after the convolutional layer of the VGG19 model, to strengthen the convolution result of the VGG19 model;

[0119] Step S304, select Huber loss as the loss function of the VGG19 model, select Adma as the optimizer of the VGG19 model, define the learning rate and hyperparameters of the VGG19 model, and complete the pre-construction of the fault diagnosis model.

[0120] In an embodiment of the present invention, the fault diagnosis model is a VGG19 model based on a transfer learning mechanism and a convolutional neural network. At the same time, the pooling layer of the VGG19 model is improved based on the transfer learning mechanism, thereby reducing the data dimension and retaining feature information. At the same time, the convolutional neural network of the VGG19 model is customized and improved to strengthen the convolution result of the VGG19 model, thereby avoiding the overfitting phenomenon of the VGG19 model when training on a small data set, thereby ensuring the convergence of the model.

[0121] The embodiment of the present invention provides a method for iteratively training a fault diagnosis model based on a training set and a test set. Figure 5 The schematic diagram of the implementation flow of the method for iteratively training the fault diagnosis model based on the training set and the test set is shown. The method for iteratively training the fault diagnosis model based on the training set and the test set specifically includes:

[0122] Step S401, obtaining a pre-built fault diagnosis model, a training set, and a test set, wherein the fault diagnosis model includes an input layer, a convolutional layer, a convolutional neural network, an improved pooling layer, a residual network structure, and an output layer;

[0123] Step S402, using the training set as input, executing the fault diagnosis model, and using a small amount of test set data in the target domain to migrate to the source domain for fine-tuning training;

[0124] Step S403, inputting the training set into the convolutional neural network, and the convolutional neural network performs dilated convolution on the training set data to obtain a dilated convolution result;

[0125] Step S404, loading the dilated convolution result, inputting the dilated convolution result into the improved pooling layer, and the improved pooling layer performs dimensionality reduction processing on the dilated convolution result to obtain a dimensionality reduction output result;

[0126] Step S405, inputting the dimension reduction output result into the residual network structure to obtain a training diagnosis result;

[0127] Step S406, determining whether the training diagnosis result meets the preset diagnosis threshold;

[0128] Step S407, if it meets the preset diagnosis threshold, output the current fault diagnosis model, if it does not meet the preset diagnosis threshold, return to step S402;

[0129] In this embodiment, the preset diagnostic threshold may be 0.3-0.35.

[0130] Step S408, loading the test set, and using the test set to verify whether the accuracy of the fault diagnosis model meets the preset accuracy;

[0131] Step S409: if the preset accuracy is met, the verified fault diagnosis model is output; if the preset accuracy is not met, the process returns to step S402.

[0132] In this embodiment, the preset accuracy can be set to 0.9-0.95.

[0133] The embodiment of the present invention provides a method for executing the fault diagnosis model to obtain the fault diagnosis result of the transmission network monitoring point in the visual monitoring model. Figure 6 The schematic diagram of the implementation flow of the method for executing the fault diagnosis model to obtain the fault diagnosis results of the transmission network monitoring point in the visual monitoring model is shown. The method for executing the fault diagnosis model to obtain the fault diagnosis results of the transmission network monitoring point in the visual monitoring model specifically includes:

[0134] Step S501, obtaining monitoring data of the transmission network monitoring point in the visual monitoring model, inputting the monitoring data into a convolutional neural network, and performing dilation convolution on the monitoring data by the convolutional neural network to obtain a fault state matrix;

[0135] Step S502, loading the fault state matrix, and performing dimension reduction processing on the fault state matrix by improving the pooling layer to obtain a dimension reduction fault feature set;

[0136] Step S503, identifying a reduced-dimensionality fault feature set, inputting the reduced-dimensionality fault feature set into a residual network structure, and obtaining a fault abnormality degree corresponding to the reduced-dimensionality fault feature set of the monitoring point;

[0137] Among them, the fault abnormality is calculated by formula (3):

[0138]

[0139] Among them, Y(x) is the fault abnormality, m is the data volume of the reduced-dimensional fault feature set, and x i is the input value of the dimension reduction fault feature set, k is the number of convolution kernels in the convolutional neural network. In this embodiment, the number of convolution kernels can be 3-9, α is the convolution expansion factor of the convolutional neural network. In this embodiment, the convolution expansion factor can be 0.1-0.6, and w is the gradient value of the residual network structure;

[0140] w=sign(x i ) β *max(|x i |-σ,0) (4)

[0141] Among them, w is the gradient value of the residual network structure, β is the gradient coefficient of the activation function of the residual network structure. In this embodiment, the gradient coefficient can be 2-5, σ is the preset threshold parameter of the residual network structure, and the preset parameter threshold can be 0.02-0.06.

[0142] Step S504, judging whether there is an abnormality at the current monitoring point based on the fault abnormality degree;

[0143] Among them, when judging whether there is an abnormality at the current monitoring point based on the fault abnormality, the abnormality threshold is used for judgment, and the abnormality threshold can be set to 0.15-0.25.

[0144] Step S505, if there is an abnormality, identify the features of the dimension-reduced fault feature set, classify the fault using the fault type database, and then implement fault diagnosis;

[0145] Step S506: If there is no abnormality, the identification of the reduced-dimensional fault feature set is terminated, and the fault abnormality degree of the current monitoring point is output.

[0146] The embodiment of the present invention also provides a power communication network fault diagnosis device based on convolutional neural network, Figure 7 A schematic diagram of the framework of the power communication network fault diagnosis device based on a convolutional neural network is shown. The power communication network fault diagnosis device based on a convolutional neural network specifically includes:

[0147] The monitoring point deployment module 100 is used to traverse the power communication network, perform principal component analysis on the power communication network structure, obtain the communication transmission facility weights, and deploy the transmission network monitoring points based on the communication transmission facility weights;

[0148] A visualization module 200 is used to collect monitoring data of the transmission network monitoring points in real time, and generate a visualization monitoring model based on the transmission network monitoring points and monitoring data;

[0149] A database construction module 300, the database construction module 300 constructs a fault type database based on the expert database, traverses the fault type database, indexes the fault data set in the fault type database, and divides the fault data set into a training set and a test set;

[0150] The diagnostic model generation module 400 is used to pre-build a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network, obtain a training set and a test set, iteratively train the fault diagnosis model based on the training set and the test set, complete model training and verification, and output a converged fault diagnosis model;

[0151] The result output module 500 is used to load monitoring data, take the monitoring data as input, execute the fault diagnosis model, and obtain the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

[0152] In this embodiment, the monitoring point deployment module 100, the visualization presentation module 200, the database construction module 300, the diagnosis model generation module 400, and the result output module 500 can be connected by 5G or DTU communication.

[0153] The embodiment of the present invention provides a monitoring point deployment module 100. Figure 8 The schematic diagram of the framework of the monitoring point deployment module 100 is shown. The monitoring point deployment module 100 specifically includes:

[0154] The data acquisition unit 110 is used to identify the transmission lines, data lines, power supply lines and transmission facilities in the electric power communication network, and collect the operation logs and operation flow data of the transmission lines, data lines, power supply lines and transmission facilities, and extract three feature vectors of attribute features, transmission features and line features from the operation logs and operation flow data based on attribute features, transmission frequency and line association;

[0155] The feature queue generation unit 120 uses a mean shift clustering algorithm to perform clustering operations on the three feature vectors, divides feature vectors with high feature similarity into the same fault indicator factor cluster, calculates factor cluster feature values ​​of the fault indicator factor cluster based on principal component analysis, and generates a factor cluster feature queue;

[0156] The weight setting unit 130 selects the largest attribute characteristic factor cluster, transmission characteristic factor cluster, and line characteristic factor cluster in the factor cluster characteristic queue respectively, determines the principal component analysis weight coefficient by using the variance contribution rate of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster, sets the weight of the communication transmission facility based on the principal component analysis weight, and deploys the transmission network monitoring points by using the weight discriminant function and the weight of the communication transmission facility.

[0157] The embodiment of the present invention provides a visualization presentation module 200. Fig. 9 The schematic diagram of the framework of the visualization presentation module 200 is shown. The visualization presentation module 200 specifically includes:

[0158] The data processing unit 210 is used to obtain the coordinate data of the monitoring points of the transmission network and the monitoring data, and perform missing value processing, outlier processing and normalization processing on the coordinate data of the monitoring points and the monitoring data;

[0159] The visualization construction unit 220 uses a three-dimensional modeling tool to identify the coordinate data of the monitoring points, construct an initial visualization model, load the processed monitoring data, identify the monitoring data, and classify the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points to obtain a monitoring data set containing classification labels;

[0160] The model optimization unit 230 edits and adjusts the initial visualization model based on the monitoring data set containing the partition labels to generate a visualization monitoring model.

[0161] In an embodiment of the present invention, by performing missing value processing, outlier processing and normalization processing on the monitoring point coordinate data and the monitoring data, it is convenient to identify the monitoring point coordinate data. At the same time, editing and adjusting the initial visualization model based on the monitoring data set containing the partitioning labels can present the visualization monitoring model in real time, which facilitates the correlation tracing and analysis of monitoring point anomalies.

[0162] In summary, the present invention provides a method and device for fault diagnosis of an electric power communication network based on a convolutional neural network. In an embodiment of the present invention, a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network is pre-constructed to realize monitoring data recognition and dilated convolution, obtain a fault state matrix, and classify faults using a fault type database, thereby realizing analysis and diagnosis of all monitoring point faults and potential faults in the electric power communication network, avoiding the spread of communication line faults, and ensuring the efficiency of electric power communication transmission.

[0163] It should be noted that, for the above-mentioned embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, some steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0164] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in this field can still combine, add, delete or make other adjustments to the features in the various embodiments of the present invention according to the circumstances without conflict, without making creative work, so as to obtain different other technical solutions that do not deviate from the concept of the present invention in essence, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A power communication network fault diagnosis method based on convolutional neural network, characterized in that: The power communication network fault diagnosis method based on convolutional neural network includes: Traverse the power communication network, perform principal component analysis on the power communication network structure, obtain the weights of communication transmission facilities, and deploy transmission network monitoring points based on the weights of communication transmission facilities; Collect monitoring data from transmission network monitoring points in real time, and generate a visual monitoring model based on the transmission network monitoring points and monitoring data; Building a fault type database based on the expert database, traversing the fault type database, indexing the fault data set in the fault type database, and dividing the fault data set into a training set and a test set; Pre-build a fault diagnosis model based on the transfer learning mechanism combined with a convolutional neural network, obtain training sets and test sets, iteratively train the fault diagnosis model based on the training sets and test sets, complete model training and verification, and output a converged fault diagnosis model; The monitoring data is loaded, and the fault diagnosis model is executed with the monitoring data as input to obtain the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

2. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 1, characterized in that: The method for deploying transmission network monitoring points based on communication transmission facility weights specifically includes: Identify the transmission lines, data lines, power supply lines, and transmission facilities in the electric power communication network, and collect the operation logs and operation flow data of the transmission lines, data lines, power supply lines, and transmission facilities; Load the operation log and operation flow data, and extract three feature vectors, namely, attribute feature, transmission feature, and line feature, from the operation log and operation flow data based on attribute feature, transmission frequency, and line association; The mean shift clustering algorithm is used to cluster the three feature vectors, and the feature vectors with high feature similarity are divided into the same fault indicator factor cluster. The feature similarity uses Manhattan distance as the similarity measurement standard.

3. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 2, characterized in that: The method for deploying transmission network monitoring points based on communication transmission facility weights specifically includes: Load multiple groups of fault indicator factor clusters, calculate the factor cluster characteristic values ​​of the fault indicator factor clusters based on principal component analysis, and generate factor cluster characteristic queues; Select the largest attribute feature factor cluster, transmission feature factor cluster, and line feature factor cluster in the factor cluster feature queue respectively; The variance contribution rates of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster are used to determine the principal component analysis weight coefficient, and the weight of the communication transmission facility is set based on the principal component analysis weight coefficient; Deploy transmission network monitoring points using weighted discriminant functions and weights of communication transmission facilities; Among them, when deploying the transmission network monitoring point, the importance of the communication transmission facility is calculated based on the weighted discriminant function, and an important threshold is preset to determine whether the importance of the communication transmission facility exceeds the preset important threshold. If it exceeds the preset important threshold, the transmission network monitoring point is deployed for the communication transmission facility; Among them, C x is the importance of communication transmission facilities, q(x) is the weight of communication transmission facilities, n is the number of communication transmission facilities, and g is the coupling coefficient of the weighted discriminant function; Where q(x) is the weight of the communication transmission facility, n is the number of communication transmission facilities, and x n is the factor cluster eigenvalue, and λ is the data dimension of the factor cluster eigenvalue.

4. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 1, characterized in that: The method for generating a visual monitoring model based on transmission network monitoring points and monitoring data specifically includes: Obtain the coordinate data and monitoring data of the transmission network monitoring points, and perform missing value processing, outlier processing and normalization processing on the coordinate data and monitoring data of the monitoring points; Use 3D modeling tools to identify the coordinate data of monitoring points and build an initial visualization model; Load the processed monitoring data, identify the monitoring data, classify the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points, and obtain a monitoring data set with classification labels; The initial visualization model is edited and adjusted based on the monitoring data set containing the partition labels to generate a visualization monitoring model.

5. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 4, characterized in that: The method for pre-building a fault diagnosis model based on a transfer learning mechanism combined with a convolutional neural network specifically includes: The VGG19 model is pre-trained using the training set to obtain a pre-trained VGG19 model, and a residual network structure is introduced to replace the fully connected layer of the VGG19 model; Improve the pooling layer of the VGG19 model based on the transfer learning mechanism, where the pooling layer is improved through the fine-tuning layer of the transfer learning mechanism model; Customize and improve the convolutional neural network of the VGG19 model, and place the convolutional neural network after the convolutional layer of the VGG19 model to strengthen the convolution results of the VGG19 model; Select Huber loss as the loss function of the VGG19 model, select Adma as the optimizer of the VGG19 model, define the learning rate and hyperparameters of the VGG19 model, and complete the pre-construction of the fault diagnosis model.

6. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 5, characterized in that: The method for iteratively training the fault diagnosis model based on the training set and the test set specifically includes: Obtain a pre-built fault diagnosis model, a training set, and a test set, wherein the fault diagnosis model includes an input layer, a convolutional layer, a convolutional neural network, an improved pooling layer, a residual network structure, and an output layer; Using the training set as input, the fault diagnosis model is executed, and a small amount of test set data in the target domain is used to migrate to the source domain for fine-tuning training; The training set is input into the convolutional neural network, and the convolutional neural network performs dilated convolution on the training set data to obtain the dilated convolution result; Load the dilated convolution result, input the dilated convolution result into the improved pooling layer, and the improved pooling layer performs dimensionality reduction processing on the dilated convolution result to obtain the dimensionality reduction output result; Input the dimension reduction output result into the residual network structure to obtain the training diagnosis result; Determine whether the training diagnosis result meets the preset diagnosis threshold. If it meets the preset diagnosis threshold, output the current fault diagnosis model; Load the test set and use the test set to verify whether the accuracy of the fault diagnosis model meets the preset accuracy; If the preset accuracy is met, the verified fault diagnosis model is output.

7. The method for diagnosing faults in a power communication network based on a convolutional neural network according to claim 6, characterized in that: The method of executing the fault diagnosis model to obtain the fault diagnosis result of the transmission network monitoring point in the visual monitoring model specifically includes: The monitoring data of the transmission network monitoring points in the visual monitoring model are obtained, and the monitoring data are input into the convolutional neural network. The convolutional neural network performs dilation convolution on the monitoring data to obtain the fault state matrix. Load the fault state matrix, improve the pooling layer to reduce the dimension of the fault state matrix, and obtain the reduced-dimensional fault feature set; Identify the reduced-dimensionality fault feature set, input the reduced-dimensionality fault feature set into the residual network structure, and obtain the fault abnormality degree corresponding to the reduced-dimensionality fault feature set of the monitoring point; Among them, the fault abnormality is calculated by formula (3): Among them, Y(x) is the fault abnormality, m is the data volume of the reduced-dimensional fault feature set, and x i is the input value of the dimension-reduced fault feature set, k is the number of convolution kernels in the convolutional neural network, α is the convolution expansion factor of the convolutional neural network, and w is the gradient value of the residual network structure; w=sign(x i ) β *max(x i -σ,0)(4) Among them, w is the gradient value of the residual network structure, β is the gradient coefficient of the activation function of the residual network structure, and σ is the preset threshold parameter of the residual network structure; Based on the fault abnormality degree, it is judged whether there is an abnormality at the current monitoring point. If there is an abnormality, the features of the reduced-dimensional fault feature set are identified, and the fault is classified using the fault type database to achieve fault diagnosis.

8. A power communication network fault diagnosis device based on a convolutional neural network, implemented by the power communication network fault diagnosis method based on a convolutional neural network according to any one of claims 1 to 7, characterized in that: The power communication network fault diagnosis device based on convolutional neural network includes: Monitoring point deployment module, which is used to traverse the power communication network, perform principal component analysis on the power communication network structure, obtain the weight of communication transmission facilities, and deploy the transmission network monitoring points based on the weight of communication transmission facilities; A visualization presentation module is used to collect monitoring data of the transmission network monitoring points in real time and generate a visualization monitoring model based on the transmission network monitoring points and monitoring data; Database construction module: The database construction module constructs a fault type database based on the expert database, traverses the fault type database, indexes the fault data set in the fault type database, and divides the fault data set into a training set and a test set; The diagnostic model generation module is used to pre-build a fault diagnosis model based on the combination of transfer learning mechanism and convolutional neural network, obtain training sets and test sets, iteratively train the fault diagnosis model based on the training sets and test sets, complete model training and verification, and output a converged fault diagnosis model; The result output module is used to load monitoring data, take the monitoring data as input, execute the fault diagnosis model, and obtain the fault diagnosis results of the transmission network monitoring points in the visual monitoring model.

9. The power communication network fault diagnosis device based on convolutional neural network according to claim 8, characterized in that: The monitoring point deployment module includes: The data acquisition unit is used to identify the transmission lines, data lines, power supply lines and transmission facilities in the electric power communication network, and collect the operation logs and operation flow data of the transmission lines, data lines, power supply lines and transmission facilities, and extract three feature vectors of attribute features, transmission features and line features from the operation logs and operation flow data based on attribute features, transmission frequency and line association; The feature queue generation unit uses the mean shift clustering algorithm to perform clustering operations on the three feature vectors, divides the feature vectors with high feature similarity into the same fault indicator factor cluster, calculates the factor cluster characteristic value of the fault indicator factor cluster based on principal component analysis, and generates a factor cluster feature queue; A weight setting unit selects the largest attribute characteristic factor cluster, transmission characteristic factor cluster, and line characteristic factor cluster in the factor cluster characteristic queue respectively, determines the principal component analysis weight coefficient by using the variance contribution rate of the attribute characteristic factor cluster, the transmission characteristic factor cluster, and the line characteristic factor cluster, sets the weight of the communication transmission facility based on the principal component analysis weight coefficient, and deploys the transmission network monitoring points by using the weight discriminant function and the weight of the communication transmission facility.

10. The power communication network fault diagnosis device based on convolutional neural network according to claim 9, characterized in that: The visual presentation module includes: A data processing unit, the data processing unit is used to obtain the coordinate data of the monitoring points of the transmission network and the monitoring data, and perform missing value processing, outlier processing and normalization processing on the coordinate data of the monitoring points and the monitoring data; A visualization construction unit, which uses a three-dimensional modeling tool to identify the coordinate data of the monitoring points, construct an initial visualization model, load the processed monitoring data, identify the monitoring data, and classify the monitoring points based on the communication service attributes of the monitoring data and the associated attributes of the monitoring points to obtain a monitoring data set containing classification labels; Model optimization unit,The model optimization unit edits and adjusts the initial visualization model based on the monitoring data set containing the partition labels to generate a visualization monitoring model.

Citation Information

Patent Citations

  • Power dispatching communication fault monitoring method and system

    CN117792957A

Cited By

  • Power grid fault diagnosis method and system based on power data analysis

    CN121167496A