A method and system for distinguishing the fault type of power transmission line based on capsule network

Abstract

The present invention proposes a method and system for distinguishing the fault type of a power transmission line based on a capsule network. By acquiring a large amount of PMU (Phasor Measurement Unit) data, the PMU data is graphically generated into a radar chart; a power transmission line fault type distinction model based on a capsule network is constructed, a training set and a test set are divided, the training set is input into the model for training, and the change characteristics of the graphical PMU data are extracted; the test set is input into the trained model for testing, and the test results are classified and output in the form of a confusion matrix; the fault type of the transmission line is distinguished. The present invention does not require logical reasoning, nor does it require the introduction of a large amount of knowledge describing the behavior of the protection system. It only needs to extract the features corresponding to different fault types from the graphical PMU data. The modeling and model modification process is simple and easy to operate, the PMU data is graphically generated, computing resources are saved, the model training and testing speed is fast, and the diagnostic accuracy is relatively high.

Description

Technical Field

[0001] The present invention relates to the field of power grid fault diagnosis, and more specifically, to a method and system for distinguishing power transmission line fault types based on capsule networks. Background Art

[0002] With the large-scale access of new energy sources such as wind and solar energy, as well as flexible loads, controllable loads, and distributed power sources to the power grid, the complexity and uncertainty of the power grid are increasing. While the development of complex power grids and power markets has brought obvious economic benefits, it has also posed severe challenges to the safety of the power grid. Power grid fault diagnosis is the basis for accident analysis and accident handling, and is an important application for realizing the self-healing function of smart grids. It is of great significance to improve the stability of the power grid.

[0003] At present, power grid fault diagnosis is mainly divided into rule-based fault diagnosis and data-driven fault diagnosis. Among them, rule-based fault diagnosis methods include expert systems, information fusion, Petri nets, analytical models, etc., and data-driven fault diagnosis methods are mainly based on neural networks, but the neural networks used are mostly 3-4 layers in structure, and the ability to extract deep knowledge is insufficient.

[0004] For the diagnosis of online power grid faults, the existing fault diagnosis technology needs to introduce a large number of protection and equipment action rules in the early stage of modeling, and the modeling process and model maintenance process are difficult and have low versatility. With the rapid development of computer technology and artificial intelligence technologies such as deep learning, due to the excellent data feature extraction and learning capabilities of deep learning, the effective use of data for power grid fault diagnosis has become an increasing research hotspot. Summary of the invention

[0005] In order to effectively utilize the existing fault information data and improve the efficiency and accuracy of fault type identification. The present invention proposes a method for distinguishing the fault type of a power transmission line based on a capsule network, which takes the graphical PMU (phasor measurement unit) data of the transmission line as the analysis object, and can fully explore the change characteristics of the graphical PMU data before, after and during the fault to determine the fault type. The present invention does not take the PMU data itself as the analysis object, but converts the PMU data into a radar chart by graphical conversion, extracts the change characteristics of the PMU data radar chart before, after and during the fault based on the excellent feature extraction ability of the capsule network, and finally classifies the fault to determine the fault type.

[0006] Specifically, a method for distinguishing the type of power transmission line fault based on capsule network of the present invention comprises the following steps:

[0007] S1: Build a power grid simulation model, set model parameters and operating conditions, and obtain the PMU data of the phasor measurement device when the transmission line fails;

[0008] S2: selecting characteristic quantities of the PMU data, performing normalization processing, and converting the PMU data into a radar chart;

[0009] S3: Build a transmission line fault type discrimination model based on capsule network and randomly divide the training set and test set;

[0010] S4: Train the model through training set samples, extract and learn features of PMU data radar chart, and obtain a trained transmission line fault type discrimination model;

[0011] S5: Input the test set into the trained transmission line fault type discrimination model for testing, and use the tested model to discriminate the transmission line fault type.

[0012] Preferably, the step S1 comprises:

[0013] A simulation model is built based on the power system simulation software DIgSILENT, and the model flow is adjusted according to the actual measured power grid flow to improve the model accuracy.

[0014] Preferably, step S2 comprises:

[0015] The three-phase voltage amplitude, three-phase current and zero-sequence current amplitude of the transmission line are selected as feature quantities. The selected 7 electrical feature quantities are normalized and mapped to the range [0, 1]. Then the graphic interval is set to [-0.5, 1.5]. After hiding the corresponding horizontal and vertical coordinates, the PMU data is converted into a 7-dimensional radar chart.

[0016] Preferably, in step S3,

[0017] The capsule network model includes a network of a convolution layer, a main capsule layer, a digital capsule layer and three fully connected layers. The model inputs a PMU data radar chart, and outputs a feature map after 32 convolution operations with 32 convolution kernels in the convolution layer. After 8 convolution operations with 8 convolution kernels in the main capsule layer, 9 groups of feature maps are obtained. The feature maps are then flattened and converted into 4 18-dimensional capsules through digital capsules. The capsules are transmitted to the fully connected layer and flattened by the flatten function to reconstruct an image of the same size as the input image, and the fault type is classified by building a Softmax classifier.

[0018] A large number of PMU data radar charts generated by transformation are randomly divided into training set and test set in a ratio of 8:2. The samples in the training set and the test set are both labeled with corresponding labels, including: single-phase ground short circuit, two-phase short circuit, two-phase ground short circuit, and three-phase short circuit.

[0019] Preferably, step S4 comprises:

[0020] The training set is input into the model to train it, and the radar chart features of PMU data are extracted and learned. After repeated model parameter optimization and hyperparameter adjustment, the trained transmission line fault type discrimination model is obtained.

[0021] Preferably, the step S5 comprises:

[0022] The test set samples are input into the training set for training and repeated parameter adjustment to obtain the discrimination model, and finally the confusion matrix is ​​output, including the accuracy, precision, recall rate and F1 value evaluation index of the transmission line fault type discrimination and the fault type discrimination result.

[0023] The present invention also proposes a power transmission line fault type identification system based on capsule network, the system comprising a fault data acquisition module, a data graph conversion module, a model construction module, a model training module, a model testing module and an output module;

[0024] The fault data acquisition module is used to build a power grid simulation model, set model parameters and operating conditions, and obtain PMU data when a transmission line fails;

[0025] The data graph conversion module is used to convert the acquired PMU data into a 7-dimensional radar graph, and randomly divide the PMU data radar graph into a training set and a test set in a ratio of 8:2;

[0026] The model building module is used to build a power transmission line fault type discrimination model based on a capsule network;

[0027] The model training module uses a training set to train the capsule network;

[0028] The model testing module uses a test set to test the trained capsule network, and the capsule network extracts and learns the features of the PMU data radar map and classifies the fault types to achieve the discrimination of the transmission line fault type;

[0029] The output module outputs a confusion matrix, including the accuracy, precision, recall rate and F1 value evaluation index of the transmission line fault type discrimination and the fault type discrimination result.

[0030] The output module presents the test results in the form of a confusion matrix, making the test results clearer and more intuitive.

[0031] The present invention has the following beneficial technical effects:

[0032] 1. PMU data has the advantages of high sampling frequency and good real-time performance. Dense sampling points are conducive to more detailed reflection of fault information. When a large amount of PMU data is input into the neural network, the huge amount of calculation requires a huge amount of computer memory, which has extremely high requirements for computer equipment and takes a long time to calculate. Therefore, PMU data is converted into radar charts to improve the speed of model training and testing and save computing resources.

[0033] 2. The present invention proposes a method for distinguishing the type of transmission line fault based on a capsule network model. The capsule network initializes the weights of the entire neural network during training, and extracts the changing characteristics of the PMU data radar chart before, during and after the fault through excellent feature extraction capabilities. The expression of the confusion matrix is ​​intuitive and detailed, and it is easy to show the distinction of various fault types.

[0034] 3. The method proposed in the present invention has a certain anti-noise ability. It can still accurately identify the type of transmission line fault when extracting multiple electrical quantity characteristics. It has strong robustness and is not easily restricted by environmental, topographical, climatic and other conditions. It has universal applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0036] Figure 1 The flowchart of the method of the present invention is shown in FIG.

[0037] Figure 2 This is an example of a PMU data radar chart of the present invention.

[0038] Figure 3 Structural diagram of the capsule network model designed for the present invention.

[0039] Figure 4 The coding structure diagram of the transmission line fault type discrimination model designed for the present invention.

[0040] Figure 5 Decoding structure diagram of the transmission line fault type discrimination model designed for the present invention.

[0041] Figure 6 Schematic diagram of the Softmax classification principle designed for the present invention.

[0042] Figure 7 Confusion matrix diagram of test results designed for the present invention.

[0043] Figure 8 It is a system principle block diagram of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments, and is only used to explain the present invention, and is not intended to limit the present invention. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0045] Example 1

[0046] In order to improve the efficiency and accuracy of transmission line fault type identification, the present invention proposes a method for transmission line fault type identification based on capsule network. This embodiment uses DIgSILENT to build a simulation model based on the Sichuan power grid network architecture, calls DIgSILENT through the python program interface to obtain a large amount of PMU data, selects seven electrical quantities, namely the three-phase current amplitude and zero-sequence current amplitude and three-phase voltage amplitude at a certain end of the transmission line as feature quantities, and transforms the PMU data of the above seven electrical quantities into a 7-dimensional radar chart after normalization. Then, the PMU data radar chart is randomly divided into a training set and a test set at a ratio of 8:2, and the training set samples are input into the capsule network for model training. After the model training is completed, the test set samples are input for testing, thereby realizing the identification of the transmission line fault type.

[0047] Specific as Figure 1 As shown, the method of this embodiment includes the following steps:

[0048] S1. Obtain a large amount of PMU data through power grid simulation model simulation experiments.

[0049] This embodiment uses the python program interface to call the power grid simulation model built by DIgSILENT to obtain a large amount of PMU data. The simulation experiment sets the data sampling time to 3s, the electromechanical transient step to 0.01s, the data sampling starts from -0.1s, the fault is set to occur at 0s, and the corresponding circuit breaker is actuated at 0.04s. The electrical quantities obtained include the three-phase current amplitude and phase angle, zero-sequence current amplitude and phase angle, three-phase voltage amplitude and phase angle, active power and reactive power at both ends of the transmission line.

[0050] S2. Graph the PMU data.

[0051] When converting PMU data into graphics, seven electrical quantities, including the three-phase current amplitude, zero-sequence current amplitude, and three-phase voltage amplitude at one end of the transmission line, are selected as feature quantities. The PMU data of the selected feature quantities are normalized and mapped to the range [0, 1]. Then, the graphic interval is set to [-0.5, 1.5]. After hiding the corresponding horizontal and vertical axes, the PMU data is converted into a radar chart. Examples of radar charts for single-phase ground short circuit, two-phase short circuit, two-phase ground short circuit, and three-phase short circuit in the transmission line are shown below. Figure 2 As shown in (a), (b), (c) and (d).

[0052] S3. Build a transmission line fault type discrimination model based on capsule network and randomly divide the training set and test set.

[0053] The capsule network model is mainly divided into three parts:

[0054] The first part is a convolution layer. The convolution kernel is the most basic and important component in convolution. In the capsule network, the convolution layer is used as the base layer to extract feature information from the input image. Multiple convolution kernels are selected according to the actual situation. Each element of the convolution kernel corresponds to a weight coefficient and a bias, which is similar to a neuron in a feedforward neural network. Each convolution kernel is convolved with the input image of the upper layer to obtain the corresponding feature map, which is used as the input of the next layer.

[0055] The second part is the main capsule layer, which replaces the scalar output in the convolution operation with a vector output, thereby retaining the image content and the location information of the graphic content.

[0056] The third part is the digital capsule layer, in which the capsule calculates the relationship between the child capsule and the parent capsule by multiplying it with the transformation matrix, and then generates the final capsule and its probability for each class according to the routing protocol.

[0057] In addition, the dynamic routing algorithm is the core of the capsule network and plays a decisive role in the application of the entire capsule network. The algorithm is based on the coupling coefficient and is adjusted and determined through multiple iterative routing. The child capsule is calculated by the parent capsule, and the probability of each parent capsule j connecting to the child capsule i is

[0058]

[0059] In the formula, c ij Also known as coupling coefficient; b ij is the prior probability that child capsule i is connected to parent capsule j, and the initial value is set to 0. According to the coupling coefficient c ij Calculate the weighted sum to get the input vector s j .

[0060]

[0061]

[0062] In the formula, u i is the subcapsule vector, is the parent capsule prediction vector, W ij is the weight conversion matrix, s j is the input vector. Nonlinear compression is achieved using discriminative learning, and the activation function of the capsule layer is realized through it. The output vector v j It can be expressed as

[0063]

[0064] In the formula, v j is the output vector, s j is the input vector. The length of the capsule's output vector represents the probability of the entity, so a nonlinear compression function is used to limit the vector's modulus to between (0,1). The longer the output vector, the greater the probability that the image belongs to that class.

[0065] In the routing update phase, the output vector v j With the parent capsule prediction vector Multiply to update b ij . b ij The update calculation formula is:

[0066]

[0067] The capsule network model designed by the present invention includes a network of a convolutional layer (Conv2D), a primary capsule layer (PrimaryCaps), a digital capsule layer (DigitCaps) and three fully connected layers (FC). The model structure is as follows: Figure 3 As shown. Among them, BN / Relu means batch normalization before the Relu activation function; 16*16 convolution kernel (32, step length = 1, S) means the convolution kernel size is 16*16, the number is 32, the step length is 1, and S is SAME padding; 9*9 main capsule (8, step length = 2, V) means the convolution kernel size in the main capsule layer is 9*9, the number is 8, the step length is 2, and V is VALID padding. The input is the graphical PMU data, the size is 64*64*3. The fully connected layer uses the Softmax function, and the output is the probability value of the corresponding fault type.

[0068] The encoding structure of the power transmission line fault type discrimination model based on the capsule network in this embodiment is as follows: Figure 4 The calculation formula of the convolution size when SAME padding is performed during the convolution of the convolution layer is:

[0069]

[0070] In the formula, N is the output size after convolution, S is the step size, and W is the input size. Therefore, the size of the input PMU data radar map is 64*64, and the output size is still 64*64 after the convolution operation of the convolution layer with a step size of 1. The input image is an RGB color image, so the number of channels is 3, and the final input image size is 64*64*3. After 32 convolution operations of the convolution layer Conv2D with 32 convolution kernels, the output feature map of size 64*64*32 is obtained.

[0071] The calculation formula for the convolution size when VALID padding is performed during the convolution of the main capsule layer is:

[0072]

[0073] In the formula, N is the output size after convolution, F is the convolution kernel size, S is the step size, W is the input size, and the calculation result is rounded down. Therefore, the input feature map size is 64*64, and the output size is 28*28 after the main capsule layer convolution operation with a step size of 2 and a convolution kernel size of 9*9. After 8 convolution operations with 8 convolution kernels in the main capsule layer, 9 groups of 28*28*8 feature maps are obtained.

[0074] The feature map is then flattened, and the final vector neuron size is 6272*9, that is, 6272 (28*28*8) capsules, each capsule consists of a 9-dimensional vector. When the capsule of the main capsule layer enters the digital capsule layer, it passes through a dynamic routing weight conversion matrix W with a shape of 18*9 ij The 9-dimensional capsule is converted into an 18-dimensional capsule, where each capsule corresponds to a classification category, where the conversion matrix W ij The size is set in the network parameters. There are 4 types of faults in the transmission line fault type discrimination of the present invention, so there are 4 classification categories, and 4 18-dimensional capsules are output.

[0075] The decoding structure is as follows Figure 5 As shown. The decoding process is used to reconstruct the image. There are three fully connected layers in the model of the present invention. The four 18-dimensional capsules output by the digital capsule layer are all transmitted to the 512 neurons of the first fully connected layer, and then transmitted to the 1024 neurons of the second fully connected layer, and finally transmitted to the third fully connected layer. After flattening by the flatten function, a 1*12288 (64*64*3) image with the same size as the input size of 64*64*3 is reconstructed, and the fault type is classified by constructing a Softmax classifier. The Softmax function formula is

[0076]

[0077] In the formula, the numerator is the e of a certain original output value zi, the denominator is e of all original output values ​​from 1 to k zj , Softmax maps all original output values ​​to values ​​between (0,1), and the sum of all output mappings is 1. The Softmax classification principle diagram is as follows Figure 6 As shown in the figure, the white box is the original output value, the black box is the probability value of each category calculated by Softmax, and the category with the largest probability value is the final prediction result.

[0078] In addition, the fault sample composition of this embodiment is shown in Table 1, and the PMU data radar chart is randomly divided into a training set and a test set at a ratio of 8:2.

[0079] Table 1

[0080] Data Source Single-phase ground short circuit Two-phase short circuit Two-phase ground short circuit Three-phase short circuit DIgSILENT 1746 1752 1673 1845

[0081] S4. Train the model using training set samples.

[0082] The training set samples are input into the model to train it, and the changing characteristics of the PMU data radar chart before, during and after the fault are extracted. After repeated model parameter optimization and hyperparameter adjustment, a transmission line fault type discrimination model with excellent performance is obtained.

[0083] S5. Input the test set into the trained model for testing to identify the type of transmission line fault.

[0084] This embodiment uses the model trained by the above method to test fault samples, and outputs four evaluation indicators of fault type discrimination accuracy, precision, recall and F1 value, as well as the confusion matrix of the discrimination results. The calculation formulas of the four evaluation indicators are:

[0085]

[0086]

[0087]

[0088]

[0089] In the formula, TP indicates the number of events where the true value is positive and the predicted value is positive; FN indicates the number of events where the true value is positive and the predicted value is negative; FP indicates the number of events where the true value is negative and the predicted value is positive; TN indicates the number of events where the true value is negative and the predicted value is negative. The relationship among TP, FN, FP, and TN of the four types of events is shown in Table 2.

[0090] Table 2

[0091]

[0092] The confusion matrix of the discrimination result in this embodiment is as follows Figure 7 As shown in FIG. 3 , the confusion matrix contains the number of samples and percentages of various types of discrimination results. The larger the number, the darker the color of the matrix block, and the smaller the number, the lighter the color of the matrix block. The fault type discrimination test results of this embodiment are shown in Table 3. The results show that the method proposed in this embodiment can more accurately discriminate the type of transmission line faults.

[0093] Table 3

[0094] Fault type Accuracy Accuracy Recall F1 value Single-phase ground short circuit 0.96 0.99 0.93 0.96 Two-phase short circuit 0.96 0.92 0.98 0.95 Two-phase ground short circuit 0.96 0.96 0.98 0.97 Three-phase short circuit 0.96 0.98 0.95 0.96

[0095] Example 2

[0096] Based on the method proposed in the above embodiment, this embodiment further proposes a power transmission line fault type identification system based on capsule network.

[0097] Specific as Figure 8 As shown, the system of this embodiment includes a fault data acquisition module, a data graph conversion module, a model construction module, a model training module, a model testing module and an output module;

[0098] The fault data acquisition module of this embodiment acquires a large amount of PMU data based on the python program interface and the power grid model simulation experiment built by DIgSILENT;

[0099] The data graph conversion module of this embodiment is used to convert the acquired PMU data into a 7-dimensional radar graph, and randomly divide the PMU data radar graph into a training set and a test set in a ratio of 8:2;

[0100] The model training module of this embodiment uses a training set to train the capsule network;

[0101] The model building module is used to build a power transmission line fault type discrimination model based on a capsule network;

[0102] The model testing module of this embodiment uses a test set to test the trained capsule network. The capsule network extracts and learns the features of the PMU data radar chart and classifies the fault types, thereby realizing the discrimination of the transmission line fault type.

[0103] The output module of this embodiment presents the test results in the form of a confusion matrix, making the test results clearer and more intuitive.

[0104] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0105] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0106] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0108] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for distinguishing the type of transmission line fault based on capsule network, characterized in that: The method comprises the following steps: S1: Build a power grid simulation model, set model parameters and operating conditions, and obtain the PMU data of the phasor measurement device when the transmission line fails; S2: selecting characteristic quantities of the PMU data, performing normalization processing, and converting the PMU data into a radar chart; S3: Construct a transmission line fault type discrimination model based on capsule network, and randomly divide the training set and test set; the capsule network model includes a network of a convolution layer, a main capsule layer, a digital capsule layer and three fully connected layers. The model inputs the PMU data radar map, and outputs a feature map after 32 convolution operations with 32 convolution kernels in the convolution layer. After 8 convolution operations with 8 convolution kernels in the main capsule layer, 9 groups of feature maps are obtained. The feature maps are then flattened and converted into 4 18-dimensional capsules through digital capsules. The images are transmitted to the fully connected layer and flattened by the flatten function to reconstruct the images of the same size as the input images, and the fault types are classified by constructing a Softmax classifier; The large amount of PMU data radar charts generated by transformation are randomly divided into training set and test set in a ratio of 8:

2. The samples in the training set and the test set are both labeled with corresponding labels, including: single-phase ground short circuit, two-phase short circuit, two-phase ground short circuit, and three-phase short circuit; S4: Train the model through training set samples, extract and learn features of PMU data radar chart, and obtain a trained transmission line fault type discrimination model; S5: Input the test set into the trained transmission line fault type discrimination model for testing, and use the tested model to discriminate the transmission line fault type, including: input the test set samples into the discrimination model obtained after training with the training set and repeated parameter adjustment, and finally output the confusion matrix, including the accuracy, precision, recall rate and F1 value evaluation index of the transmission line fault type discrimination and the fault type discrimination result.

2. The method according to claim 1, characterized in that The step S1 comprises: A simulation model is built based on the power system simulation software DIgSILENT, and the model flow is adjusted according to the actual measured power grid flow to improve the model accuracy.

3. The method according to claim 2, characterized in that The step S2 comprises: The three-phase voltage amplitude, three-phase current and zero-sequence current amplitude of the transmission line are selected as feature quantities. The selected 7 electrical feature quantities are normalized and mapped to the range [0, 1]. Then the graphic interval is set to [-0.5, 1.5]. After hiding the corresponding horizontal and vertical coordinates, the PMU data is converted into a 7-dimensional radar chart.

4. The method according to claim 3, characterized in that The step S4 comprises: The training set is input into the model to train it, and the radar chart features of PMU data are extracted and learned. After repeated model parameter optimization and hyperparameter adjustment, the trained transmission line fault type discrimination model is obtained.

5. A system for distinguishing the fault type of a power transmission line based on a capsule network, used to implement the method according to claims 1-4, characterized in that: The system includes a fault data acquisition module, a data graph conversion module, a model construction module, a model training module, a model testing module and an output module; The fault data acquisition module is used to build a power grid simulation model, set model parameters and operating conditions, and obtain PMU data when a transmission line fails; The data graph conversion module is used to convert the acquired PMU data into a 7-dimensional radar graph, and randomly divide the PMU data radar graph into a training set and a test set in a ratio of 8:2; The model building module is used to build a power transmission line fault type discrimination model based on a capsule network; The model training module trains the capsule network using a training set; The model testing module uses a test set to test the trained capsule network, and the capsule network extracts and learns the features of the PMU data radar map and classifies the fault types to achieve the discrimination of the transmission line fault type; The output module outputs a confusion matrix, including the accuracy, precision, recall rate and F1 value evaluation index of the transmission line fault type discrimination and the fault type discrimination result.

Images