Power distribution network topology identification method and system based on anti-noise convolution deep learning
Through the anti-noise convolution deep learning method, Gaussian noise and salt-pepper noise data augmentation and convolutional neural network training, the problem of the impact of traditional topology recognition methods in power systems is solved, and high-precision topology structure recognition is achieved.
Patent Information
- Application Number
- CN202510417160.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-08-01
AI Technical Summary
Traditional topological recognition methods are difficult to adapt to the frequent access and exit of distributed energy in power systems, and the recognition accuracy is susceptible to noise, resulting in a decrease in reliability of topological recognition.
The anti-noise convolution deep learning method is adopted to collect data through the power system sensor network, perform data augmentation and train convolutional neural networks. The enhanced data set is constructed using Gaussian noise and salt-pigment noise, combined with ReLU activation function and maximum pooling operation, and trained using Adam optimization algorithm to output the on-off probability of the output line.
It improves the accuracy and stability of topological identification, and can maintain high-precision topological structure identification in a noisy environment, and is suitable for power systems containing distributed energy.
Smart Images
Figure CN120408137A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of distribution networks, and in particular to a method and system for identifying the topology of a distribution network by anti-noise convolutional deep learning. Background Art
[0002] With the continuous increase in the penetration rate of renewable energy in the distribution network, the identification of the distribution network topology has become a crucial part of distribution network management. Traditional topology identification methods are difficult to adapt to the frequent access and withdrawal of distributed energy sources (such as photovoltaic and wind power) in the power system, and the identification accuracy is easily affected by noise, resulting in misjudgment of topology identification and affecting the reliability of topology identification. Developing a topology identification method suitable for a large number of distributed energy accesses and with strong anti-noise ability has become an important challenge in current research. Therefore, proposing a new deep learning model based on a convolutional neural network to solve this problem has significant innovation and advantages. Summary of the Invention
[0003] In view of the above existing problems, the present invention is proposed.
[0004] Therefore, the problem to be solved by the present invention is: how to solve the problem that traditional topology identification methods are difficult to adapt to the frequent access and withdrawal of distributed energy sources (such as photovoltaic and wind power) in the power system, and the identification accuracy is easily affected by noise, resulting in misjudgment of topology identification and affecting the reliability of topology identification.
[0005] To solve the above technical problems, the present invention provides the following technical solution: A method for identifying the topology of a distribution network by anti-noise convolutional deep learning, including collecting voltage amplitude data of each node through a sensor network in the power system, recording topology structure labels, and constructing an initial data set; performing data augmentation on the voltage amplitude data to construct an augmented data set; training a convolutional neural network model using the augmented data set; after training is completed, applying the trained convolutional neural network to real-time power system data, outputting the identified power system topology structure, and obtaining the line topology connection and disconnection information of the distribution network.
[0006] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the data augmentation includes adding Gaussian noise and salt-and-pepper noise; adding Gaussian noise to the voltage amplitude data to construct a Gaussian noise augmented data set; adding salt-and-pepper noise to the voltage amplitude data to construct a salt-and-pepper noise augmented data set; combining the Gaussian noise data set, the salt-and-pepper noise data set, and the original data set to construct an augmented data set for training the convolutional neural network model; Gaussian noise perturbs the input data by adjusting the mean and standard deviation of the data, while salt-and-pepper noise simulates accidental failures of sensors by randomly changing the values of some data points.
[0007] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the identified topology of the power system includes the status of each line in the distribution network, including the on / off status of the line; the power system is a distribution network containing high-penetration renewable energy; the on / off status of the line is determined according to the threshold of 0.5. When the output probability is greater than 0.5, it is determined to be on, and when it is less than 0.5, it is determined to be off, balancing the risk of misjudgment; the trained convolutional neural network is applied to the operating power system in real time to identify the on-line topology and output the status change of the distribution network in real time.
[0008] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the convolutional neural network model includes using the Sigmoid activation function as the output layer to output the on / off probability of each line. The convolutional neural network model includes a one-dimensional convolutional layer 1, a one-dimensional convolutional layer 2, a one-dimensional convolutional layer 3, and a fully connected layer; the one-dimensional convolutional layer 1 uses 8 convolutional kernels to perform preliminary feature extraction on the input data; the one-dimensional convolutional layer 2 uses 16 convolutional kernels to further extract higher-level features; the one-dimensional convolutional layer 3 uses 8 convolutional kernels to continue compressing the features; the fully connected layer flattens the features extracted by the convolutional layer and makes a final decision output through the fully connected layer to complete the topology identification; the size of the convolutional kernel used in the convolutional layer is set to 5 to capture sufficient local information in the input data and balance the computational efficiency and feature extraction ability.
[0009] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the convolutional neural network model further includes using the ReLU activation function after each convolutional layer, and using the max pooling operation after the one-dimensional convolutional layer 3 to increase the non-linear expression ability of the network and reduce the feature dimension.
[0010] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the training includes using the Adam optimization algorithm, updating the parameters of the convolutional neural network by calculating the gradient of the weighted binary cross-entropy loss function and performing backpropagation to obtain the optimal solution.
[0011] As a preferred solution of the method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to the present invention, wherein: the training further includes the loss function expressed as:
[0012] L = L0 + ω1·L1 + ω2·L2
[0013] Among them, L0 is the loss of the original data set, L1 is the loss of the Gaussian noise data set, L2 is the loss of the salt-and-pepper noise data set, and ω1 and ω2 are the weights of the losses corresponding to the Gaussian noise data set and the salt-and-pepper noise data set respectively; each loss L X is calculated using the cross-entropy function. The subscript X takes values of 0, 1, and 2 respectively, which is expressed as:
[0014]
[0015] Among them, N is the total number of samples of the input data, M is the number of lines, and y ij is the true label of the jth branch on the ith sample, and 0 or 1 is used to represent the on-off state of the line, where 1 represents the line is on and 0 represents the line is off. is the predicted probability of the jth branch on the ith sample. A predicted probability greater than 0.5 indicates that the line is predicted to be operating normally, and less than 0.5 indicates that the predicted line is disconnected.
[0016] Another object of the present invention is to provide a system for a distribution network topology identification method based on anti-noise convolutional deep learning, which can solve the problem of a distribution network topology identification based on anti-noise convolutional deep learning by constructing a system for a distribution network topology identification based on anti-noise convolutional deep learning.
[0017] To solve the above technical problems, the present invention provides the following technical solutions: A distribution network topology identification system based on anti-noise convolutional deep learning includes a data acquisition module, a data augmentation module, a training module, and a prediction module; the data acquisition module is used to collect the voltage amplitude data of each node through a sensor network in the power system, record the topology structure label, and construct an initial data set; the data augmentation module is used to perform data augmentation on the voltage amplitude data to construct an augmented data set; the training module is used to train a convolutional neural network model using the augmented data set; the prediction module is used to apply the trained convolutional neural network to real-time power system data after training and output the identified power system topology structure to obtain the line topology on-off information of the distribution network.
[0018] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned distribution network topology identification method based on anti-noise convolutional deep learning are implemented.
[0019] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned distribution network topology identification method based on anti-noise convolutional deep learning are implemented.
[0020] The beneficial effects of the present invention are as follows: A method for identifying the topology of a distribution network based on anti-noise convolutional deep learning provided by the present invention introduces data augmentation methods based on Gaussian noise and salt-and-pepper noise. Gaussian noise is used to simulate common random measurement errors in the power system, while salt-and-pepper noise is used to simulate data loss or outlier interference in extreme cases, improving the robustness of the model in a noisy environment and enabling the method to effectively reduce the impact of noise on the accuracy of topology identification in practical applications. The present invention is applicable to the identification of meshed and radial topology structures, and has good universality and adaptability. Through the feature extraction and data augmentation of the convolutional neural network, the present invention can maintain a high accuracy of topology identification in different noisy environments. Description of the Drawings
[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.
[0022] Figure 1 It is a flowchart of a method for identifying the topology of a distribution network based on anti-noise convolutional deep learning provided by the first embodiment of the present invention.
[0023] Figure 2 It is an implementation flowchart of a method for identifying the topology of a distribution network based on anti-noise convolutional deep learning provided by the first embodiment of the present invention.
[0024] Figure 3 It is a structural diagram of a system for identifying the topology of a distribution network based on anti-noise convolutional deep learning provided by the second embodiment of the present invention. Detailed Embodiments
[0025] To make the above objects, features, and advantages of the present invention more obvious and understandable, the detailed embodiments of the present invention will be described in detail below with reference to the drawings in the specification.
[0026] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention, so the present invention is not limited by the specific embodiments disclosed below.
[0027] Example 1, refer to Figure 1 and Figure 2, which is the first embodiment of the present invention. This embodiment provides a method for identifying the topology of a distribution network based on anti-noise convolutional deep learning, including: collecting the voltage amplitude data of each node through a sensor network in the power system, recording the topology structure labels, and constructing an initial data set; performing data augmentation on the voltage amplitude data to construct an augmented data set; training a convolutional neural network model using the augmented data set; after the training is completed, applying the trained convolutional neural network to the real-time power system data to output the identified topology structure of the power system, and obtaining the line topology connection and disconnection information of the distribution network.
[0028] The present invention provides a method for identifying the topology of a distribution network based on anti-noise convolutional deep learning, which can accurately identify the topology structure of the distribution network in a noisy environment. By combining data augmentation techniques, such as adding Gaussian noise and salt-and-pepper noise, an augmented data set is constructed to improve the robustness of the model to noise. In addition, a convolutional neural network is used for training, enabling this method to not only adapt to different distribution network structures but also perform efficient topology identification during real-time operation. As Figure 1 and Figure 2 shown.
[0029] S1. Collect the voltage amplitude data of each node through a sensor network in the power system, record the topology structure labels, and construct an initial data set.
[0030] Collect the voltage amplitude data of each node through a sensor network in the power system, and record the topology structure labels at the same time. The collected data constitutes the initial data set, which reflects the operating state of the distribution network and provides a basis for subsequent topology identification.
[0031] S2. Perform data augmentation on the voltage amplitude data to construct an augmented data set.
[0032] To improve the robustness of the model to noise, perform data augmentation on the voltage amplitude data, specifically by adding Gaussian noise and salt-and-pepper noise. In this way, an augmented data set is constructed to enhance the adaptability of the neural network to noise interference in the real environment.
[0033] In the noise setting, Gaussian noise perturbs the input data by adjusting the mean and standard deviation of the data, while salt-and-pepper noise simulates the accidental failure of sensors by randomly changing the values of some data points.
[0034] Add Gaussian noise and salt-and-pepper noise; add Gaussian noise to the voltage amplitude data to construct a Gaussian noise augmented data set; add salt-and-pepper noise to the voltage amplitude data to construct a salt-and-pepper noise augmented data set; combine the Gaussian noise data set, the salt-and-pepper noise data set, and the original data set to construct an augmented data set for training the convolutional neural network model.
[0035] The parameters of the Gaussian noise are a mean of 0 and a standard deviation of 5%, and the noise ratio of the salt-and-pepper noise is set to 5%.
[0036] S3. For the convolutional neural network model, train it using the augmented dataset.
[0037] Train the convolutional neural network using the augmented dataset. This network contains multiple convolutional layers and fully connected layers, including one-dimensional convolutional layers for extracting features at different levels: the first layer uses a convolutional kernel of size 5 and number 8 for initial feature extraction of the input data; the second layer uses a convolutional kernel of size 5 and number 16 to further extract higher-level features; the third layer uses a convolutional kernel of size 5 and number 8 to continue compressing the features. Finally, the features extracted by the convolutional layers are flattened through the fully connected layer and decision output is performed to complete the recognition of the topological structure.
[0038] After each convolutional layer, the ReLU activation function is used, and the max pooling operation is used after the last convolutional layer to increase the non-linear expression ability of the network and reduce the feature dimension. The output layer uses the Sigmoid activation function to output the on-off probability of each line. By setting a threshold of 0.5, it is determined whether the line is in a normal operating state (line closed) or a disconnected state (line open).
[0039] During the training process, the optimization algorithm uses the Adam optimizer, and backpropagation is performed by calculating the gradient of the weighted binary cross-entropy loss function to update the weights of the convolutional neural network, ensuring the stability of the training process and the high accuracy of the model.
[0040] During the training process, the cross-entropy loss function is used to calculate the loss of each sample to measure the difference between the prediction result and the true label, and the losses of the original dataset and the augmented dataset are comprehensively considered. The calculation formula of the loss function in the training process is:
[0041] L = L0 + ω1·L1 + ω2·L2
[0042] where L0 is the loss of the original dataset, L1 is the loss of the Gaussian noise dataset, L2 is the loss of the salt-and-pepper noise dataset, and ω1 and ω2 are the weights corresponding to the losses of the Gaussian noise dataset and the salt-and-pepper noise dataset respectively.
[0043] Each loss L X is calculated using the cross-entropy function, and the subscript X takes values of 0, 1, and 2 respectively, which is expressed as:
[0044]
[0045] where N is the total number of samples of the input data, M is the number of lines, y ij$y_{ij}$ is the true label of the $j$-th branch in the $i$-th sample, represented by 0 or 1 to indicate the on / off state of the line, where 1 indicates the line is on and 0 indicates the line is off. $p_{ij}$ is the predicted probability of the $j$-th branch in the $i$-th sample. A predicted probability greater than 0.5 indicates that the line is predicted to be operating normally, and less than 0.5 indicates that the line is predicted to be disconnected.
[0046] The topology identification model is applicable to radial and meshed distribution networks.
[0047] S4. After training is completed, apply the trained convolutional neural network to real-time power system data, output the identified power system topology structure, and obtain the line topology on / off information of the distribution network.
[0048] After training is completed, the obtained convolutional neural network can be applied to the operating power system in real time. By predicting real-time voltage amplitude data, the network can output the topology structure of the distribution network and identify the on / off state of each line. This method can monitor the state changes of the distribution network in real time, ensuring the timeliness and accuracy of topology identification, and is particularly suitable for power systems with a large number of distributed energy accesses.
[0049] A method for identifying the topology of a distribution network based on anti-noise convolutional deep learning according to the present invention is applicable to power systems with a large number of distributed energy accesses. This method can effectively identify the topology structure of the distribution network, especially in the presence of noise interference, and has strong anti-noise ability.
[0050] This method processes voltage amplitude data through a convolutional neural network to realize the identification of the power system topology structure in scenarios with frequent fluctuations of distributed energy. Specifically, in the data preprocessing stage, a dual-modal noise injection technique, including adding Gaussian noise and salt-and-pepper noise, is used to construct an enhanced data set; in network design, based on the convolutional neural network, combined with the ReLU activation function and max-pooling operation, the non-linear expression ability of the network is increased; in the training process, the Adam optimization algorithm is used to achieve anti-noise training, thereby improving the identification ability of the model under different noise conditions. Compared with the prior art, this method can significantly improve the accuracy and stability of topology identification, and has strong application prospects and advantages.
[0051] Example 2, refer to Figure 3 , which is the second embodiment of the present invention. Different from the previous embodiment, it provides a system for identifying the topology of a distribution network based on anti-noise convolutional deep learning, including: a data acquisition module, a data enhancement module, a training module, and a prediction module.
[0052] The data acquisition module is used to collect the voltage amplitude data of each node through the sensor network in the power system, record the topology structure label, and construct an initial data set.
[0053] The data augmentation module is used to perform data augmentation on the voltage amplitude data and construct an augmented dataset.
[0054] The training module is used to train the convolutional neural network model using the augmented dataset.
[0055] After the training is completed, the prediction module applies the trained convolutional neural network to the real-time power system data, outputs the identified power system topology structure, and obtains the line topology connection and disconnection information of the distribution network.
[0056] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.
[0057] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or used in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.
[0058] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connections (electronic devices) having one or more wirings, portable computer diskettes (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber devices, and portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.
[0059] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.
[0060] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.
Claims
1. A method for identifying the topology of a distribution network based on anti-noise convolutional deep learning, characterized in that: including Collecting the voltage amplitude data of each node through a sensor network in the power system, recording the topology structure tags, and constructing an initial data set Performing data augmentation on the voltage amplitude data to construct an augmented data set Training a convolutional neural network model using the augmented data set After training is completed, applying the trained convolutional neural network to real-time power system data, outputting the identified power system topology structure, and obtaining the line topology connection and disconnection information of the distribution network 2. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 1, wherein: The data augmentation includes adding Gaussian noise and salt-and-pepper noise Adding Gaussian noise to the voltage amplitude data to construct a Gaussian noise augmented data set Adding salt-and-pepper noise to the voltage amplitude data to construct a salt-and-pepper noise augmented data set Combining the Gaussian noise data set, the salt-and-pepper noise data set, and the original data set to construct an augmented data set for training the convolutional neural network model Gaussian noise perturbs the input data by adjusting the mean and standard deviation of the data, while salt-and-pepper noise simulates the accidental failure of sensors by randomly changing the values of some data points 3. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 2, wherein: The identified power system topology structure includes the status of each line in the distribution network, including the line connection and disconnection status The power system is a distribution network with a high penetration of renewable energy The connection and disconnection status of the line is determined according to the threshold of 0.
5. When the output probability is greater than 0.5, it is determined to be connected, and when it is less than 0.5, it is determined to be disconnected, balancing the misjudgment risk The trained convolutional neural network is applied to the operating power system in real time for online topology structure identification and real-time output of the state changes of the distribution network 4. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 3, characterized in that: The convolutional neural network model includes Using the Sigmoid activation function as the output layer to output the connection and disconnection probabilities of each line The convolutional neural network model includes a one-dimensional convolutional layer 1, a one-dimensional convolutional layer 2, a one-dimensional convolutional layer 3, and a fully connected layer The one-dimensional convolutional layer 1 uses 8 convolutional kernels to perform preliminary feature extraction on the input data The one-dimensional convolutional layer 2 uses 16 convolutional kernels to further extract higher-level features The one-dimensional convolutional layer 3 uses 8 convolutional kernels to continue to compress the features The fully connected layer flattens the features extracted by the convolutional layer and makes the final decision output through the fully connected layer to complete the topology identification The size of the convolutional kernel used in the convolutional layer is set to 5 to capture sufficient local information in the input data and achieve a balance between computational efficiency and feature extraction ability 5. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 4, characterized in that: The convolutional neural network model also includes using the ReLU activation function after each convolutional layer, and using the max-pooling operation after the one-dimensional convolutional layer 3 to increase the non-linear expression ability of the network and reduce the feature dimension 6. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 5, wherein: The training includes using the Adam optimization algorithm to update the parameters of the convolutional neural network by calculating the gradient of the weighted binary cross-entropy loss function and performing backpropagation to obtain the optimal solution 7. The method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to claim 6, characterized in that: The training also includes the loss function expressed as L = L0 + ω1·L1 + ω2·L2 where L0 is the loss of the original data set, L1 is the loss of the Gaussian noise data set, L2 is the loss of the salt-and-pepper noise data set, and ω1 and ω2 are the weights corresponding to the losses of the Gaussian noise data set and the salt-and-pepper noise data set respectively Each loss L X is calculated using the cross-entropy function, where the subscript X takes values 0, 1, and 2 respectively, denoted as where N is the total number of samples of the input data, M is the number of lines, and y ij is the true label of the j-th branch on the i-th sample, using 0 or 1 to represent the on / off state of the line, where 1 means the line is on and 0 means the line is off. is the predicted probability of the j-th branch on the i-th sample. A predicted probability greater than 0.5 indicates that the line is predicted to be operating normally, and less than 0.5 indicates that the line is predicted to be disconnected.
8. A system adopting a method for identifying the topology of a distribution network by using anti-noise convolutional deep learning as described in any one of claims 1 to 7, characterized in that: It includes a data acquisition module, a data enhancement module, a training module, and a prediction module; The data acquisition module is used to collect the voltage amplitude data of each node through the sensor network in the power system, record the topology structure label, and construct an initial data set; The data enhancement module is used to perform data enhancement on the voltage amplitude data and construct an enhanced data set; The training module is used to train the convolutional neural network model using the enhanced data set; The prediction module is used to apply the trained convolutional neural network to the real-time power system data after training, output the identified power system topology structure, and obtain the line topology connection and disconnection information of the distribution network.
9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of a method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of a method for identifying the topology of a distribution network by anti-noise convolutional deep learning according to any one of claims 1 to 7.