Power distribution network concealment fault identification method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a method and a device for identifying a hidden fault of a power distribution network, electronic equipment and a storage medium. The method comprises the following steps: acquiring a fault current signal generated when each feeder line of the power distribution network runs in a historical period; performing data expansion on the fault current signal by using a least square generative adversarial network to obtain a sample set; training a preset basic recognition model according to the sample set to obtain a fault recognition model; and in response to a fault identification operation, identifying the hidden fault of the power distribution network based on the fault identification model. According to the scheme, the efficiency and the accuracy of the hidden fault identification of the power distribution network can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network detection, and in particular to a method, device, electronic device and storage medium for identifying hidden faults in a distribution network. Background Art

[0002] Electricity plays a key role in modern society. As an indispensable form of energy, it is vital to the country's economic development and people's daily life. As an important link connecting the transmission network and power users, the distribution network is responsible for the effective distribution and transmission of electric energy. Its operating reliability is directly related to the normal production and life of power users. In the operation of the distribution network, common problems include instantaneous hidden faults, permanent hidden faults and other types. Hidden faults refer to problems that exist in the power system but are not easily detected. They may pose a potential threat to the normal operation of the system. They are problems that occur briefly and disappear quickly in the power system. Although they last for a short time, they may still affect the stability of the system and the normal operation of the equipment. In order to deeply study the laws of hidden faults in the distribution network, it is necessary to classify and analyze the faults in detail in order to better understand and scientifically manage various types of faults. Classification research can be carried out based on the nature, cause, impact and other aspects of the fault, providing an objective basis for preventing and solving faults.

[0003] However, the hidden faults in the distribution network have weak fault signals, are repeated at a single point, and are difficult to locate, gradually damaging system performance, ultimately leading to equipment failure or system collapse, and reducing the reliability of power supply in the distribution network system. In addition, there are many types of hidden faults, and the fault representations vary greatly. It is impossible to summarize the unified characteristics of hidden faults, and it is even more impossible to quickly and intelligently identify hidden faults. The precise maintenance of daily distribution lines is difficult, the personnel input efficiency is low, the investment is large, the system status recovery time is long, and the evaluation is difficult. After a fault occurs, distribution automation can isolate the fault in a limited section, but further search currently relies mainly on experience judgment, which is inefficient, time-consuming, and labor-intensive. More than 50% of hidden faults still cannot be accurately checked, which buries hidden dangers for subsequent faults. It can be seen that the current efficiency of identifying hidden faults in the distribution network is low and the accuracy is poor. Summary of the invention

[0004] The embodiments of the present application provide a distribution network hidden fault identification system, method, electronic device and storage medium, which can improve the efficiency and accuracy of distribution network hidden fault identification.

[0005] In a first aspect, an embodiment of the present application provides a method for identifying hidden faults in a distribution network, comprising:

[0006] Collect the fault current signals generated by the operation of each feeder in the distribution network during the historical period;

[0007] Performing data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set;

[0008] Training a preset basic recognition model according to the sample set to obtain a fault recognition model;

[0009] In response to the fault identification operation, hidden faults of the power distribution network are identified based on the fault identification model.

[0010] Optionally, in some embodiments of the present application, training a preset basic recognition model according to the sample set to obtain a fault recognition model includes:

[0011] Convert each fault current signal in the sample set into a time-frequency characteristic diagram;

[0012] The preset basic recognition model is trained based on the time-frequency feature graph to obtain a fault recognition model.

[0013] Optionally, in some embodiments of the present application, the training of a preset basic recognition model based on the time-frequency feature graph to obtain a fault recognition model includes:

[0014] Adjusting the size of the time-frequency feature map to a preset size;

[0015] Based on the multi-head self-attention structure and the shifted window multi-head self-attention structure of the basic recognition model, the adjusted video feature map is recognized to obtain the estimated fault corresponding to the fault current signal;

[0016] The preset basic recognition model is trained according to the actual fault and the estimated fault corresponding to the fault current signal to obtain a fault recognition model.

[0017] Optionally, in some embodiments of the present application, before performing data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set, the method further includes:

[0018] Get the preset generator and discriminator;

[0019] A least squares generative adversarial network is constructed based on the generator, the discriminator and the preset least squares loss function.

[0020] Optionally, in some embodiments of the present application, constructing a least squares generative adversarial network according to the generator, the discriminator and a preset least squares loss function includes:

[0021] Outputting generation samples generated by the operation of each feeder based on the generator;

[0022] Determine, based on the discriminator, the probability that the generated sample is real data;

[0023] According to the probability that the generated sample is real data and a preset least squares loss function, a least squares generative adversarial network is constructed.

[0024] Optionally, in some embodiments of the present application, the collecting of fault current signals generated when each feeder of the distribution network is operating during a historical period includes:

[0025] Collect fault data generated by the operation of each feeder in the distribution network during the historical period;

[0026] A high-pass filter is used to eliminate asymmetric components in the fault data, and a fault current signal generated when each feeder is running during the historical period is obtained.

[0027] Optionally, in some embodiments of the present application, eliminating the asymmetric component in the fault data by a high-pass filter to obtain the fault current signal generated when each feeder is running during the historical period includes:

[0028] Based on the frequency information of the zero-sequence current after the fault and the frequency information of the zero-sequence current before the fault of each feeder, the asymmetric component in the fault data is eliminated to obtain the fault current signal generated when each feeder was operated during the historical period.

[0029] In a second aspect, an embodiment of the present application provides a device for identifying hidden faults in a distribution network, comprising:

[0030] A collection module is used to collect fault current signals generated by the operation of each feeder in the distribution network during the historical period;

[0031] An expansion module, used for performing data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set;

[0032] A training module, used to train a preset basic recognition model according to the sample set to obtain a fault recognition model;

[0033] The identification module is used to identify hidden faults of the distribution network based on the fault identification model in response to the fault identification operation.

[0034] Correspondingly, the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of any of the above methods when executing the program.

[0035] The present application also provides a storage medium, wherein the storage medium stores a computer program, and the computer program implements any of the above methods when the computer program is executed by a processor.

[0036] The present application provides a method, device, electronic device and storage medium for identifying hidden faults in a distribution network. After collecting the fault current signals generated by the operation of each feeder in the distribution network during the historical period, the fault current signals are expanded using a least squares generative adversarial network to obtain a sample set. Then, a preset basic recognition model is trained according to the sample set to obtain a fault recognition model. Finally, in response to the fault recognition operation, the hidden faults in the distribution network are identified based on the fault recognition model. In the scheme for identifying hidden faults in the distribution network provided by the present application, the fault current signals are expanded using a least squares generative adversarial network, and the preset basic recognition model is trained using the expanded training set to obtain a fault recognition model. When performing fault recognition, the hidden faults in the distribution network can be directly identified based on the fault recognition model, without relying on artificial experience to judge and identify the hidden faults in the distribution network. Thus, the efficiency and accuracy of identifying hidden faults in the distribution network can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 It is a flow chart of a method for identifying hidden faults in a distribution network provided in an embodiment of the present application;

[0039] Figure 2 This is the overall model framework of a distribution network hidden fault identification method based on LSGAN-Swin Transformer in this application;

[0040] Figure 3 This is a specific implementation flow chart of a distribution network hidden fault identification method based on LSGAN-Swin Transformer provided by this application;

[0041] Figure 4 is the discriminator and generator loss function provided in the embodiment of the present application;

[0042] Figure 5 It is a time-frequency diagram of simulation data provided in an embodiment of the present application;

[0043] Figure 6 It is a time-frequency diagram of generated data provided in an embodiment of the present application;

[0044] Figure 7 It is a visualization analysis diagram of raw data provided by an embodiment of the present application;

[0045] Figure 8 This is a visualization analysis diagram of the results after Swin-Transformer model processing provided in the embodiment of the present application;

[0046] Fig. 9 It is a structural schematic diagram of a hidden fault identification device for a distribution network provided in an embodiment of the present application;

[0047] Fig.10 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of systems and methods consistent with the examples detailed in the attached claims, or with some aspects of the present application.

[0049] It should be noted that, in this article, the terms "include", "comprises" or any other variants thereof are intended to cover descriptions such as non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.

[0050] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present application, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.

[0052] Embodiments of the present application provide a method, device, electronic device, and storage medium for identifying hidden faults in a distribution network.

[0053] Among them, the distribution network hidden fault identification device can be deployed in a terminal, and the terminal can be a personal computer. As mentioned above, the above-mentioned terminal is often equipped with a display device, and the display device can also be a display, a display screen, a touch screen, etc. The touch screen can also be a touch screen, a touch panel, etc., but is not limited to this. The terminal can communicate with the server in a wired or wireless manner. The server can include an independently running server or a distributed server, or a server cluster composed of multiple servers. The terminal can include a mobile phone, a tablet computer or a personal computer (PC, Personal Computer).

[0054] It should be noted that the order of description of the following embodiments is not intended to limit the priority order of the embodiments.

[0055] The present application provides a method for identifying hidden faults in a distribution network, comprising: collecting fault current signals generated when each feeder of the distribution network is in operation during a historical period; performing data expansion on the fault current signals using a least squares generative adversarial network to obtain a sample set; training a preset basic recognition model according to the sample set to obtain a fault recognition model; and identifying hidden faults in the distribution network based on the fault recognition model in response to a fault recognition operation.

[0056] See also Figure 1 , Figure 1 : is a flow chart of a method for identifying hidden faults in a distribution network provided by the present application, and the method for identifying hidden faults in a distribution network specifically includes the following process:

[0057] 101. Collect the fault current signals generated by the operation of each feeder in the distribution network during the historical period.

[0058] Distribution network refers to the power network in the power system that transmits electric energy from substations or power stations to end users (such as households, commercial and industrial users). The distribution network is an important part of the power system, which is responsible for the final distribution and supply of electric energy.

[0059] First, you can identify the types of faults that need to be collected, such as transient hidden faults, permanent hidden faults, etc. At the same time, determine the distribution network areas and feeders that need to be monitored. Install current sensors at key nodes of the distribution network, and then use the current sensors to monitor the current of the feeders in real time. Use a data acquisition card (DAQ) or other similar hardware devices to collect sensor data.

[0060] Optionally, in some embodiments of the present application, the step of “collecting fault current signals generated by the operation of each feeder of the distribution network during the historical period” may specifically include:

[0061] Collect fault data generated by the operation of each feeder in the distribution network during the historical period;

[0062] The asymmetric components in the fault data are eliminated by a high-pass filter, and the fault current signals generated when each feeder was in operation during the historical period are obtained.

[0063] Install current and voltage sensors on each feeder of the distribution network to monitor and record electrical parameters in real time. Use a data acquisition system (such as a data acquisition card or smart meter) to collect the current and voltage data recorded by the sensors. Ensure that the time stamps of all sensors are synchronized so that they can accurately correspond during subsequent data analysis. Store the collected data in a database or data lake for easy retrieval and analysis. Check the integrity and accuracy of the data and remove outliers and erroneous records. Perform necessary conditioning on the original signal, such as amplification and digitization, to meet the needs of subsequent analysis. Select a suitable high-pass filter: Select a suitable filter based on the frequency range of the fault characteristics. For example, if the fault signal mainly contains high-frequency components, select a high-pass filter to eliminate the asymmetric components of the low frequency.

[0064] Optionally, in some embodiments of the present application, the step of “eliminating asymmetric components in the fault data by using a high-pass filter to obtain a fault current signal generated when each feeder is running during the historical period” may specifically include:

[0065] Based on the frequency information of the zero-sequence current after the fault and the frequency information of the zero-sequence current before the fault of each feeder, the asymmetric components in the fault data are eliminated, and the fault current signals generated when each feeder was in operation during the historical period are obtained.

[0066] Extract the zero-sequence current component from the three-phase current. The zero-sequence current is the vector sum of the three-phase currents, which usually changes significantly when a fault occurs. Calculate the fundamental frequency of the current signal and determine the number of data points within one cycle. Select the zero-sequence current data of one or more complete cycles before the fault as a reference. Select the zero-sequence current data of one or more complete cycles after the fault. For each cycle, calculate the difference between the zero-sequence current after the fault and the zero-sequence current before the fault. Specifically, subtract the data corresponding to one cycle of the zero-sequence current after the fault of each feeder from the one cycle before the fault, and then subtract the data of the steady-state fault quantity of one cycle after 10 cycles of the fault from the obtained data. According to this algorithm, the asymmetric component and the steady-state power frequency component of the fault are eliminated to obtain the high-frequency transient pure fault component of the zero-sequence current of each feeder. Optionally, in some embodiments, the hidden faults of the distribution network can be divided into tree barriers, small animal bridging, and insulator flashover according to different waveform characteristics.

[0067] 102. Perform data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set.

[0068] First, the fault current signal data is normalized to a standard range, such as [-1, 1]. The existing fault current signal dataset is divided into a training set and a validation set.

[0069] Define a generator that receives a random noise vector z sampled from a Gaussian distribution as input. Fully connected layers or convolutional transpose layers can be used to gradually transform the noise vector into samples of the same dimension as the fault current signal. Batch normalization and ReLU activation functions can be used to enhance the stability and performance of the model.

[0070] Define a discriminator, which receives fault current signal samples from the real training set or samples generated by the generator. Convolutional layers and LeakyReLU activation functions can be used to gradually reduce the spatial dimension of the data. Finally, a probability value is output through a fully connected layer, indicating the probability that the input sample is real data.

[0071] Alternately train the discriminator and generator. First optimize the discriminator, fix the generator, minimize the discriminator loss, then optimize the generator, fix the discriminator, and minimize the generator loss.

[0072] Use optimization algorithms such as Adam or SGD to update the weights of the network. In each iteration, generate new noise vectors, generate new samples through the generator, and use them for training. After training, use the generator to generate new fault current signal samples. Perform necessary post-processing on the generated samples to ensure that they are comparable in quality to the real samples. Evaluate the quality and diversity of the generated samples. Use the generated samples to expand the original dataset and use them to train the fault diagnosis model.

[0073] Through these steps, LSGAN can effectively expand the dataset of fault current signals, thereby improving the generalization ability and accuracy of the fault diagnosis model.

[0074] Optionally, in some embodiments of the present application, before the step of “using a least squares generative adversarial network to perform data expansion on the fault current signal to obtain a sample set”, the following may be specifically included:

[0075] Get the preset generator and discriminator;

[0076] Based on the generator, discriminator and the preset least squares loss function, a least squares generative adversarial network is constructed.

[0077] Optionally, in some embodiments of the present application, the step of “constructing a least squares generative adversarial network according to the generator, the discriminator and the preset least squares loss function” may specifically include:

[0078] Based on the generator output, the generated samples are generated when each feeder is running;

[0079] Based on the discriminator, the probability of the generated sample being real data;

[0080] According to the probability that the generated sample is real data and the preset least squares loss function, a least squares generative adversarial network is constructed.

[0081] There are three different types of hidden fault events in the distribution network, and 200 samples are generated for each type, totaling 600 samples. Subsequently, these samples are divided into training sets and test sets in a ratio of 4:1 for training the LSGAN model. Figure 3 The figure shows the changes in the loss functions of the generator and the discriminator during the training process. During the training of the LSGAN model, the loss functions of the generator and the discriminator showed large fluctuations in the initial iteration stage. However, after 50 iterations, the loss of the generator was significantly reduced. When the training reached 100 iterations, the loss of both the generator and the discriminator dropped below 0.5 and remained stable.

[0082] 103. Train the preset basic recognition model according to the sample set to obtain a fault recognition model.

[0083] For example, divide the sample set into training set, validation set, and test set. Extract features from the fault current signal that help identify faults. Ensure that each sample has the correct label, which should be provided by domain experts or fault detection algorithms. Select a suitable model based on the fault type and data characteristics, such as support vector machine (SVM), random forest, neural network, etc. Select a machine learning library or framework, such as TensorFlow, PyTorch, Scikit-learn, etc. Determine hyperparameters such as learning rate, batch size, and number of iterations. Train the model using the training set data and adjust the model parameters through optimization algorithms such as gradient descent. Use the validation set data to evaluate the model performance and perform hyperparameter adjustment. Apply techniques such as cross-validation and L1 / L2 regularization to prevent overfitting. Use the test set to evaluate the model's accuracy, recall, F1 score, and other indicators. Generate a confusion matrix to see how the model performs on different categories. Adjust the model architecture or hyperparameters based on the evaluation results. Export the trained model to a specific format, such as PMML, ONNX, etc., and deploy the model to a production environment, which can be a cloud server, local server, or edge device

[0084] Optionally, in some embodiments of the present application, the step of “training a preset basic recognition model according to the sample set to obtain a fault recognition model” may specifically include:

[0085] Convert each fault current signal in the sample set into a time-frequency characteristic diagram;

[0086] The preset basic recognition model is trained based on the time-frequency feature graph to obtain a fault recognition model.

[0087] Commonly used time-frequency analysis methods include short-time Fourier transform (STFT), wavelet transform (WT) and continuous wavelet transform (CWT). Short-time Fourier transform (STFT): divide the signal into short time frames and perform Fourier transform on each frame. Wavelet transform (WT): use wavelet function to decompose the signal at multiple scales. Continuous wavelet transform (CWT): perform wavelet decomposition on the signal at continuous scales.

[0088] The results of the time-frequency transform are visualized as feature maps, which usually include amplitude spectrum, phase spectrum, etc. Select an appropriate machine learning model according to the nature of the problem, such as convolutional neural network (CNN), support vector machine (SVM), random forest, etc. Set the model's parameters such as the number of layers, number of neurons, activation function, etc. Divide the sample set into a training set, a validation set, and a test set. Extract features from the time-frequency feature map that help identify faults. Ensure that each sample has the correct label. Train the model using the training set data. : Use the validation set data to adjust the model parameters.

[0089] Optionally, in some embodiments of the present application, the step of “training a preset basic recognition model based on the time-frequency feature graph to obtain a fault recognition model”

[0090] Adjust the size of the time-frequency feature map to a preset size;

[0091] The adjusted video feature map is identified based on the multi-head self-attention structure and the shifted window multi-head self-attention structure of the basic recognition model to obtain the estimated fault corresponding to the fault current signal;

[0092] The preset basic recognition model is trained according to the actual fault and the estimated fault corresponding to the fault current signal to obtain a fault recognition model.

[0093] Determine the target size of the time-frequency feature map based on the input requirements of the model. Use interpolation methods (such as bilinear interpolation, bicubic interpolation) or cropping / padding methods to adjust the size of the time-frequency feature map. Introduce a multi-head self-attention mechanism in the model to allow the model to process different positions and subspace information in the feature map simultaneously. Introduce a shifted window self-attention mechanism to increase the receptive field of the model by applying self-attention at different positions. Compose a dataset with the resized time-frequency feature maps and the corresponding fault labels. Divide the dataset into training, validation, and test sets. Train the model using the training set data and adjust the model parameters using an optimization algorithm (such as Adam). Use the validation set data to adjust the model parameters to prevent overfitting.

[0094] 104. In response to the fault identification operation, identify hidden faults of the distribution network based on a fault identification model.

[0095] For example, load a trained fault identification model. Ensure that the software environment and hardware resources required for the model to run are available. Collect current signal data from the distribution network in real time. Perform the same preprocessing operations as those used in training on the real-time collected data, such as normalization and denoising. Convert the preprocessed data into a time-frequency feature map and adjust it to the preset size of the model input. Input the preprocessed and feature-extracted data into the fault identification model for inference. Make a fault judgment based on the probability output by the model and the set threshold to determine whether a fault exists. The model outputs the probability of each possible fault category. Determine the specific fault type based on the fault type with the highest probability. If the model identifies a fault, the system will notify the technician or automatically trigger the corresponding response measures. Combine the identification results and possible positioning information to determine the approximate location of the fault.

[0096] The above is the distribution network hidden fault identification process provided in the embodiment of the present application.

[0097] In summary, the embodiment of the present application uses a least squares generative adversarial network to perform data expansion on the fault current signal after collecting the fault current signal generated by the operation of each feeder in the distribution network during the historical period, and obtains a sample set. Then, the preset basic recognition model is trained according to the sample set to obtain a fault recognition model. Finally, in response to the fault recognition operation, the hidden faults of the distribution network are identified based on the fault recognition model. In the scheme for identifying hidden faults in the distribution network provided by the present application, the least squares generative adversarial network is used to perform data expansion on the fault current signal, and the preset basic recognition model is trained using the expanded training set to obtain a fault recognition model. When performing fault identification, the hidden faults of the distribution network can be directly identified based on the fault recognition model, without relying on artificial experience to judge and identify the hidden faults of the distribution network. Thus, the efficiency and accuracy of identifying hidden faults in the distribution network can be improved.

[0098] In order to facilitate understanding of the distribution network hidden fault identification solution of the present application, the following will further explain, please refer to Figure 2 , Figure 2This is the overall framework of the model of a hidden fault identification method for distribution network based on LSGAN-Swin Transformer in this application. Specifically, Swin Transformer is applied to the field of fault identification to construct an efficient classification model. The backbone architecture of the model includes a patch partition layer, a stacking module, a normalization layer, a global pooling layer, and a fully connected layer. Through continuous wavelet transform, the current signal of the hidden fault of the distribution network is converted into a time-frequency feature map of RGB three channels. The generated time-frequency map is input into the patch partition module of the model for block and flattening operations to achieve resolution compression. Specifically, the size of the time-frequency map is converted from the original [H, W, C] to [H / 4, W / 4, 16C] after partitioning and flattening. In the model structure, a window multi-head self-attention structure (W-MSA) and a shifted window multi-head self-attention structure (SW-MSA) are used, and the feature map is further processed by multi-layer SwinTransformer Blocks. The moving window multi-head self-attention mechanism (MSA) includes three main structures: MSA, W-MSA, and SW-MSA. By introducing the W-MSA and SW-MSA structures, the moving window multi-head self-attention mechanism (MSA) significantly improves the computational efficiency. The MSA structure achieves efficient time-frequency graph calculation by combining the self-attention mechanism and the moving window technology. Specifically, W-MSA divides the time-frequency graph into small windows and calculates the attention independently in each small window, which not only reduces the overall computational workload but also improves the parallelism and efficiency of the calculation. On the other hand, SW-MSA further combines the sliding window technology and local convolution operations to expand the receptive field while maintaining high computational efficiency. SW-MSA uses a sliding window method to enable each local convolution operation to cover more time-frequency graph areas, thereby greatly optimizing the calculation process of the attention mechanism without much increase in computational requirements. Through this design, MSA not only achieves a significant improvement in computational efficiency, but also significantly reduces the consumption of computing resources while maintaining accuracy. In order to solve the problem of large-scale data processing, the shifting window multi-head self-attention structure (SW-MSA) and efficient batch calculation methods are adopted. In this method, the concepts of overlapping windows and non-overlapping windows are first introduced. Overlapping windows create partially overlapping data segments in time series data, so that each window can utilize the information of adjacent windows, thereby improving the continuity and consistency of data processing. In contrast, non-overlapping windows divide the data into independent blocks, reducing the redundancy of calculations. The shifted window multi-head self-attention structure (SW-MSA) uses the multi-head self-attention mechanism (MSA) to perform self-attention calculations. MSA performs parallel calculations on the input data through multiple attention heads to capture different features and patterns in the data. This mechanism significantly improves the efficiency and accuracy of self-attention calculations.In addition, by adopting efficient batch computing, SW-MSA can process multiple data blocks simultaneously, greatly improving the computing speed and resource utilization.

[0099] Figure 3 This is a specific implementation flow chart of a distribution network hidden fault identification method based on LSGAN-Swin Transformer provided by this application, such as Figure 3 As shown, the specific steps include:

[0100] Step 1: On-site sampling of current signals of each feeder in the distribution network operation, and establishment of a feature sample library based on the waveform data obtained from the field test, so as to classify the hidden faults of the distribution network according to different waveform characteristics;

[0101] Among them, during the continuous burning process of the AC arc, the arc will extinguish near the current zero point, making the current present a tortuous alternating waveform, that is, the "zero-rest" phenomenon. The properties of the arc gap resistance will affect the voltage recovery speed, resulting in differences in the arc gap energy balance process, which in turn affects the length of the zero-rest time. The type of concealment is identified according to the difference in zero-rest time. During the stable burning period of the grounding arc, the zero-sequence current has two zero-rest processes per cycle. In order to reduce the influence of abnormal waveform distortion caused by the external environment, it is considered that the zero-rest time Tzero is the time average of 10 cycles, that is, 20 zero-rest processes in the steady-state stage of the current, that is:

[0102]

[0103] Where: k is the kth cycle in the steady-state arcing stage; N is the number of cycles; Tk and Tk' are the two zero-rest times of the kth cycle respectively.

[0104] Step 2: Use the least squares generative adversarial network (LSGAN) to expand the waveform sample data collected on site to enhance the diversity and representativeness of the fault data set;

[0105] There are three different types of hidden fault events in the distribution network, and 200 samples are generated for each type, totaling 600 samples. Subsequently, these samples are divided into training sets and test sets in a ratio of 4:1 for training the LSGAN model. Figure 4The figure shows the changes in the loss functions of the generator and the discriminator during the training process. During the training of the LSGAN model, the loss functions of the generator and the discriminator showed large fluctuations in the initial iteration stage. However, after 50 iterations, the loss of the generator was significantly reduced. When the training reached 100 iterations, the losses of the generator and the discriminator both dropped below 0.5 and remained stable. This phenomenon shows that the network has reached Nash equilibrium, indicating that the generator and the discriminator have reached a state of equilibrium in the competition. After 500 iterations, the time-frequency feature map of the generated data and the real sample data is compared. Figure 5 and Figure 6 As shown in the figure, the match between the two is good. This shows that the sample data generated by LSGAN training is highly consistent with the real fault data in terms of time-frequency characteristics, providing reliable data support for the accurate detection and diagnosis of single-phase grounding development faults.

[0106] Step 3: Build a Swin Transformer network and configure related parameters. Input the training set into the parameter-configured network for training to generate a hidden fault identification model for the distribution network.

[0107] The training set, validation set, and test set were set up with a ratio of 7:2:1. In the case of an imbalanced sample ratio, more fault samples were generated to balance the dataset by verifying the data generation capability of LSGAN. Then, the Swin-T model was used for classification, and the recognition accuracy of the model was observed under different sample numbers by changing the training ratio of fault samples to normal samples.

[0108] Step 4: Use the trained model to identify various hidden fault events of distribution networks in the validation set;

[0109] A data set of hidden faults in the distribution network operation was generated. By applying the Swin-T model to the data set, single-phase ground fault events can be accurately classified and identified. With the help of the t-SNE method, the classification results are clearly visualized. Figure 7 and Figure 8 As shown in the figure, the t-SNE method provides an intuitive means of verifying the classification performance of the Swin-T model, which helps to understand the performance of the model in high-dimensional data.

[0110] In order to facilitate better implementation of the distribution network hidden fault identification method of the embodiment of the present application, the embodiment of the present application also provides a distribution network hidden fault identification device. The meanings of the terms are the same as those in the above-mentioned distribution network hidden fault identification system, and the specific implementation details can refer to the description in the system embodiment.

[0111] See also Fig. 9 , Fig. 9 A schematic diagram of the structure of a hidden fault identification device for a distribution network provided in an embodiment of the present application, wherein the hidden fault identification device for a distribution network may specifically include a collection module 201, an expansion module 202, a training module 203 and an identification module 204, which may be specifically as follows:

[0112] The acquisition module 201 is used to acquire the fault current signal generated by the operation of each feeder in the distribution network during the historical period;

[0113] An expansion module 202 is used to perform data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set;

[0114] The training module 203 is used to train the preset basic recognition model according to the sample set to obtain a fault recognition model;

[0115] The identification module 204 is used to identify hidden faults of the distribution network based on the fault identification model in response to the fault identification operation.

[0116] The embodiment of the present application provides a device for identifying hidden faults in a distribution network. After the acquisition module 201 acquires the fault current signal generated by each feeder in the distribution network during the historical period, the expansion module 202 uses the least squares generative adversarial network to perform data expansion on the fault current signal to obtain a sample set. Then, the training module 203 trains a preset basic recognition model according to the sample set to obtain a fault recognition model. Finally, the recognition module 204 responds to the fault recognition operation and identifies the hidden faults in the distribution network based on the fault recognition model. In the scheme for identifying hidden faults in the distribution network provided in the present application, the least squares generative adversarial network is used to perform data expansion on the fault current signal, and the training set obtained by the expansion is used to train the preset basic recognition model to obtain a fault recognition model. When performing fault recognition, the hidden faults in the distribution network can be directly identified based on the fault recognition model, without relying on artificial experience to judge and identify the hidden faults in the distribution network. Thus, the efficiency and accuracy of identifying hidden faults in the distribution network can be improved.

[0117] In addition, the present application also provides an electronic device, such as Fig.10 As shown, it shows a schematic diagram of the structure of the electronic device involved in the embodiment of the present application, specifically:

[0118] The electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will appreciate that Fig.10The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0119] The processor 301 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 302, and calling data stored in the memory 302, it executes various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 301.

[0120] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and a method for identifying hidden faults in a power distribution network by running the software programs and modules stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 302 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.

[0121] The electronic device also includes a power supply 303 for supplying power to each component. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions. The power supply 303 can also include one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators and other arbitrary components.

[0122] The electronic device may further include an input unit 304, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0123] Although not shown, the electronic device may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 301 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 302 according to the following instructions, and the processor 301 will run the application programs stored in the memory 302, thereby realizing various functions, as follows:

[0124] The fault current signals generated by the operation of each feeder in the distribution network during the historical period are collected; the fault current signals are expanded using the least squares generative adversarial network to obtain a sample set; a preset basic recognition model is trained according to the sample set to obtain a fault recognition model; in response to the fault recognition operation, hidden faults of the distribution network are identified based on the fault recognition model.

[0125] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0126] After collecting the fault current signals generated by the operation of each feeder in the distribution network during the historical period, the embodiment of the present application uses the least squares generative adversarial network to perform data expansion on the fault current signals to obtain a sample set, then trains the preset basic recognition model according to the sample set to obtain a fault recognition model, and finally, in response to the fault recognition operation, identifies the hidden faults of the distribution network based on the fault recognition model. In the scheme for identifying hidden faults in the distribution network provided by the present application, the least squares generative adversarial network is used to perform data expansion on the fault current signals, and the preset basic recognition model is trained using the expanded training set to obtain a fault recognition model. When performing fault recognition, the hidden faults of the distribution network can be directly identified based on the fault recognition model, without relying on artificial experience to judge and identify the hidden faults of the distribution network, thereby improving the efficiency and accuracy of identifying hidden faults in the distribution network.

[0127] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0128] To this end, an embodiment of the present application provides a storage medium in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any of the distribution network hidden fault identification methods provided in the embodiments of the present application. For example, the instructions can execute the following steps:

[0129] The fault current signals generated by the operation of each feeder in the distribution network during the historical period are collected; the fault current signals are expanded using the least squares generative adversarial network to obtain a sample set; a preset basic recognition model is trained according to the sample set to obtain a fault recognition model; in response to the fault recognition operation, hidden faults of the distribution network are identified based on the fault recognition model.

[0130] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0131] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0132] Since the instructions stored in the storage medium can execute the steps in any one of the distribution network hidden fault identification methods provided in the embodiments of the present application, the beneficial effects that can be achieved by any one of the distribution network hidden fault identification methods provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0133] The above is a detailed introduction to a method, device, electronic device and storage medium for identifying hidden faults in a distribution network provided in an embodiment of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for identifying hidden faults in a distribution network, characterized in that: include: Collect the fault current signals generated by the operation of each feeder in the distribution network during the historical period; Performing data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set; Training a preset basic recognition model according to the sample set to obtain a fault recognition model; In response to the fault identification operation, hidden faults of the power distribution network are identified based on the fault identification model.

2. The method for identifying hidden faults in a distribution network according to claim 1, characterized in that: The method of training a preset basic recognition model according to the sample set to obtain a fault recognition model includes: Convert each fault current signal in the sample set into a time-frequency characteristic diagram; The preset basic recognition model is trained based on the time-frequency feature graph to obtain a fault recognition model.

3. The method for identifying hidden faults in a distribution network according to claim 2, characterized in that: The method of training a preset basic recognition model based on the time-frequency feature graph to obtain a fault recognition model includes: Adjusting the size of the time-frequency feature map to a preset size; Based on the multi-head self-attention structure and the shifted window multi-head self-attention structure of the basic recognition model, the adjusted video feature map is recognized to obtain the estimated fault corresponding to the fault current signal; The preset basic recognition model is trained according to the actual fault and the estimated fault corresponding to the fault current signal to obtain a fault recognition model.

4. The method for identifying hidden faults in a distribution network according to claim 1, characterized in that: Before the least squares generative adversarial network is used to perform data expansion on the fault current signal to obtain a sample set, the method further includes: Get the preset generator and discriminator; A least squares generative adversarial network is constructed based on the generator, the discriminator and the preset least squares loss function.

5. The method for identifying hidden faults in a distribution network according to claim 4, characterized in that: The least squares generative adversarial network is constructed according to the generator, the discriminator and the preset least squares loss function, including: Outputting generation samples generated by the operation of each feeder based on the generator; Determine, based on the discriminator, the probability that the generated sample is real data; According to the probability that the generated sample is real data and a preset least squares loss function, a least squares generative adversarial network is constructed.

6. The method for identifying hidden faults in a distribution network according to claim 1, characterized in that: The acquisition of fault current signals generated by the operation of each feeder in the distribution network during the historical period includes: Collect fault data generated by the operation of each feeder in the distribution network during the historical period; A high-pass filter is used to eliminate asymmetric components in the fault data, and a fault current signal generated when each feeder is running during the historical period is obtained.

7. The method for identifying hidden faults in a distribution network according to claim 6, characterized in that: The method of eliminating the asymmetric component in the fault data by using a high-pass filter to obtain the fault current signal generated when each feeder was running during the historical period includes: Based on the frequency information of the zero-sequence current after the fault and the frequency information of the zero-sequence current before the fault of each feeder, the asymmetric component in the fault data is eliminated to obtain the fault current signal generated when each feeder was operated during the historical period.

8. A device for identifying hidden faults in a distribution network, characterized in that: include: A collection module is used to collect fault current signals generated by the operation of each feeder in the distribution network during the historical period; An expansion module, used for performing data expansion on the fault current signal using a least squares generative adversarial network to obtain a sample set; A training module, used to train a preset basic recognition model according to the sample set to obtain a fault recognition model; The identification module is used to identify hidden faults of the distribution network based on the fault identification model in response to the fault identification operation.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for identifying hidden faults in a distribution network as described in any one of claims 1 to 7 are performed.

10. A storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes the method for identifying hidden faults in a distribution network as claimed in any one of claims 1 to 7.

Images