A user phase analysis method, system, device and storage medium

Abstract

The present application discloses a user phase analysis method, system, device and storage medium, the method includes: collecting the power consumption data of the gateway meter and the user in the substation area and generating a sample curve image of the voltage data changing over time; reconstructing the sample curve image through the VAE network to obtain a large number of diverse curve images, and using them as training sample data; training the SNN model through the training sample data, optimizing the hyperparameters of the SNN model by using the grid search optimization parameter method, thereby constructing a twin neural network to extract and process image features, and output the feature vectors of the voltage of the gateway meter and the voltage of the user; calculating the similarity of the feature vectors of the voltage data of the user and the different phases of the gateway meter, and obtaining the probability value of the phase to which the user belongs, and determining the phase analysis results of the single-phase users in the low-voltage substation area according to the probability value. Thus, the technical problem of poor accuracy of the user phase analysis in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a user phase analysis method, system, device and storage medium. Background Art

[0002] With the increase in the number of loads and load expansion and reconnection work in the low-voltage power supply system, the phase information of a large number of single-phase loads is unclear, which can easily lead to problems such as difficulty in managing the imbalance of three-phase loads in the power station, increased line losses, and reduced power quality. In order to achieve refined marketing management of power grid companies, reduce losses and increase efficiency, and correctly calculate line losses, accurate household-transformer relationships are the basis. Therefore, user phase reasoning work is carried out to clarify the phase information of each single-phase user, so as to facilitate the adjustment of the phase load in the power station.

[0003] At present, the main methods for determining the phase difference of users in the substation area include identification methods based on substation identification instruments, identification methods based on carrier communication, correlation analysis, and widely used clustering algorithms. In practical applications, phase difference determination with the help of substation identification instruments requires a lot of manpower and time, and the carrier communication identification signal is susceptible to interference, interruption or other unexpected situations, which will affect the accuracy of identification. There are also phase difference identification methods using correlation analysis, that is, identification based on the correlation between the electrical quantity of the single-phase meter and the substation gateway meter, and identification based on clustering algorithms, that is, user clustering, comparing the electrical quantity of each type of user with the substation transformer to obtain the phase difference connected to each type of user. For substations with small three-phase imbalance, the identification accuracy of these two methods is not ideal, and the synchronization and integrity of the data collection of the gateway meter and each user meter are required to be high. Summary of the invention

[0004] The present application provides a user phase analysis method, system, device and storage medium, which are used to solve the technical problem of poor accuracy of user phase analysis in the prior art.

[0005] In view of this, the first aspect of the present application provides a user phase analysis method, the method comprising:

[0006] Collect the area gateway table and user power consumption data and generate a sample curve image of voltage data changing over time;

[0007] Reconstructing the sample curve images through the VAE network to obtain a large number of diverse curve images as training sample data;

[0008] The SNN model is trained by the training sample data, and the hyperparameters of the SNN model are optimized by using a grid search optimization parameter method, so as to construct a twin neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage;

[0009] Calculate the similarity of the characteristic vectors of the user voltage and the voltage data of different phases in the gateway table, and obtain the probability value of the phase to which the user belongs. Determine the phase analysis result of the single-phase user in the low-voltage area according to the probability value.

[0010] Optionally, the sample reconstruction of the sample curve image by using a VAE variational autoencoder to obtain a large number of diverse curve images as training sample data specifically includes:

[0011] The sample curve image is converted into a statistical distribution in the latent space through the encoder of the VAE network, and the statistical distribution parameters are the mean and variance; the statistical distribution parameters are effectively output variable distribution through the decoder of the VAE network to obtain training sample data.

[0012] Optionally, the SNN model is trained by the training sample data, and the hyperparameters of the SNN model are optimized by a grid search optimization parameter method, so as to construct a convolutional neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage, specifically including:

[0013] Constructing an SNN model based on single sample learning, and training the SNN model through the training sample data;

[0014] The grid search method and cross-validation method are combined to cross-validate several parameter combinations of the SNN model to determine the optimal parameter combination, thereby constructing two twin neural networks with shared weights and consistent structures to extract and process image features and output feature vectors of gateway voltage and user voltage.

[0015] Optionally, the similarity between the characteristic vectors of the user voltage and the voltage data of different phases in the gateway table is calculated, and the probability value of the phase to which the user belongs is obtained, and the phase analysis result of the single-phase user in the low-voltage area is determined according to the probability value, specifically including:

[0016] The similarity calculation classifier of the twin neural network is used to calculate the similarity of the feature vectors of the user voltage and the voltage data of different phases in the gateway table;

[0017] In the calculation classifier, the softmax activation function is matched with the cross entropy loss function to predict the probability value of the phase difference, and finally the phase difference analysis result of the user to be identified is output.

[0018] A second aspect of the present application provides a user phase analysis system, the system comprising:

[0019] The sampling unit is used to collect the power consumption data of the substation and the user and generate a sample curve image of the voltage data changing with time;

[0020] A reconstruction unit, used for reconstructing the sample curve image through a VAE network to obtain a large number of diverse curve images as training sample data;

[0021] An extraction unit, used to train the SNN model through the training sample data, optimize the SNN model hyperparameters by using a grid search optimization parameter method, thereby constructing a twin neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage;

[0022] The analysis unit is used to calculate the similarity between the user voltage and the characteristic vectors of different phase voltage data of the gateway table, and obtain the probability value of the phase to which the user belongs, and determine the phase analysis result of the single-phase user in the low-voltage station area according to the probability value.

[0023] Optionally, the reconstruction unit is specifically used to:

[0024] The sample curve image is converted into a statistical distribution in the latent space through the encoder of the VAE network, and the statistical distribution parameters are the mean and variance; the statistical distribution parameters are effectively output variable distribution through the decoder of the VAE network to obtain training sample data.

[0025] Optionally, the extraction unit is specifically used to:

[0026] Constructing an SNN model based on single sample learning, and training the SNN model through the training sample data;

[0027] The grid search method and cross-validation method are combined to cross-validate several parameter combinations of the SNN model to determine the optimal parameter combination, thereby constructing two twin neural networks with shared weights and consistent structures to extract and process image features and output feature vectors of gateway voltage and user voltage.

[0028] Optionally, the analysis unit is specifically used to:

[0029] The similarity calculation classifier of the twin neural network is used to calculate the similarity of the feature vectors of the user voltage and the voltage data of different phases in the gateway table;

[0030] In the calculation classifier, the softmax activation function is matched with the cross entropy loss function to predict the probability value of the phase difference, and finally the phase difference analysis result of the user to be identified is output.

[0031] A third aspect of the present application provides a user phase analysis device, the device comprising a processor and a memory:

[0032] The memory is used to store program code and transmit the program code to the processor;

[0033] The processor is used to execute the steps of the user phase analysis method as described in the first aspect according to the instructions in the program code.

[0034] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method described in the first aspect above.

[0035] It can be seen from the above technical solutions that this application has the following advantages:

[0036] The present application provides a user phase analysis method, including: collecting the gateway meter of the substation and the user's electricity consumption data and generating a sample curve image of the voltage data changing over time; reconstructing the sample curve image through the VAE network to obtain a large number of diverse curve images, and using them as training sample data; training the SNN model through the training sample data, and optimizing the SNN model hyperparameters using the grid search optimization parameter method, thereby constructing a twin neural network to extract and process image features, and output the feature vectors of the gateway meter voltage and the user voltage; calculating the similarity of the feature vectors of the user voltage and the different phase voltage data of the gateway meter, and obtaining the probability value of the phase to which the user belongs, and determining the phase analysis results of single-phase users in the low-voltage substation according to the size of the probability value.

[0037] Compared with the prior art, the user phase difference reasoning method based on grid search and VAE-SNN provided by this application combines user phase difference reasoning with grid search and VAE-SNN model for the first time, and proposes a single-phase user phase difference reasoning method in low-voltage substation area based on grid search and VAE-SNN. The training sample data is reconstructed by VAE to generate network training sample data, and nonlinear sample features are accurately extracted. In the process of model optimization, the grid search algorithm is used to finely optimize the hyperparameters of the network to achieve the best model effect. The present invention expands the number of samples through VAE and extracts the nonlinear characteristics of the samples. The two sub-networks of SNN avoid the subjectivity of manually extracting voltage features. Thereby solving the technical problem of poor accuracy of user phase difference analysis in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of a flow chart of a user phase analysis method embodiment provided in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of the structure of a twin neural network provided in an embodiment of the present application;

[0040] Figure 3 A schematic diagram of a network search parameter optimization algorithm provided in an embodiment of the present application;

[0041] Figure 4A schematic diagram of the structure of a single-phase user phase-specific reasoning SNN model provided in an embodiment of the present application;

[0042] Figure 5 A schematic diagram of the structure of a user phase analysis system embodiment provided in an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] See also Figure 1 , a user phase analysis method provided in an embodiment of the present application includes:

[0045] Step 101, collecting the area gateway table and user power consumption data and generating a sample curve image of voltage data changing over time;

[0046] It should be noted that in the substation power consumption information collection system of this embodiment, substation gateway tables and user power consumption data are collected, the sampling period is 60 minutes, and 24-point voltage data is generated every day. In order to analyze the characteristics of voltage data changing over time, it is displayed in the form of images.

[0047] Step 102: reconstruct the sample curve image through the VAE network to obtain a large number of diverse curve images, and use them as training sample data;

[0048] It should be noted that considering that the existing samples cannot meet the twin network's requirement for the number of training samples, the VAE network is first used to reconstruct the existing sample curve images. The VAE sample generation process is specifically to receive an image and add statistical principles to the autoencoder to enable it to learn a continuous, highly structured latent space. It converts the image into parameters of a statistical distribution (mean and variance), and considers randomness during the encoding and decoding process, forcing each point sampled in the latent space to be decoded into a valid output.

[0049] VAE first converts the input sample x into the statistical distribution of the latent space through the encoder The two statistical distribution parameters generated in this process are the mean and variance. The decoder performs effective output variable distribution on the statistical distribution features obtained in the encoder. in and θ are all parameters in the encoder and decoder respectively. The similarity between two distributions is determined by KL discreteness, and the parameters are optimized to minimize discreteness. The calculation method is as shown in formula (1).

[0050]

[0051] Step 103: train the SNN model using training sample data, and optimize the hyperparameters of the SNN model using a grid search optimization parameter method, thereby constructing a twin neural network to extract and process image features, and output feature vectors of the gateway voltage and the user voltage;

[0052] It should be noted that the phase difference reasoning model based on the twin neural network is constructed, and its structure diagram is as follows Figure 2 , and optimizes the network model parameters based on the grid search algorithm. The twin neural network, also known as the twin neural network, is one of the commonly used single-sample learning models. The network contains two sub-networks with the same weight parameters, which extract features from the input data to ensure that two extremely similar images cannot be mapped to very different positions in the feature space by their respective networks, so as to maximize the representation of different labels and minimize the representation of the same label. The twin neural network mainly extracts and transforms data features through network 1 and network 2 to obtain a one-dimensional feature vector of the two data, calculates the similarity of the two feature vectors through a linear classifier, and determines the possibility of the same category of the input data.

[0053] The twin neural network sub-network is composed of two convolutional neural networks with the same parameters. The convolutional neural network structure includes input layer, convolution layer, pooling layer, fully connected layer and output layer. The convolutional neural network performs a series of feature extraction operations on the input image, and continuously optimizes the weights and other hyperparameters in the model to ensure that the set neural network can obtain the best results.

[0054] The convolution layer is the core structure of the convolutional neural network. Its function is to extract the feature map of the sample data in the form of a "sliding window". The size of the feature map is determined by the input data, the convolution kernel and the sliding step size. These parameters are all given at the beginning of the model establishment. The calculation process of the convolution layer is shown in formula (2).

[0055]

[0056] in, represents the output layer l of the i-th channel, represents the convolution kernel vector, Represents the bias, f is the activation function, and each convolutional layer uses the ReLU function as the activation function.

[0057] After the convolution layer is processed, the feature map is subjected to a series of parameter reduction operations such as dimension reduction, removal of redundant information, and compression through the pooling layer, thereby shortening the program running time. The maximum pooling method is usually selected in the convolutional neural network, and the calculation process of the pooling layer is shown in formula (3).

[0058]

[0059] in, represents the output feature map obtained by downsampling operation, They represent the weight and bias of the input image respectively.

[0060] The function of the fully connected layer is to transform the dimension of the feature map extracted by the convolution layer and the pooling layer, and to expand the feature map into a one-dimensional vector.

[0061] In this embodiment, two convolutional neural networks are selected, each having 3 layers of Conv2D modules, 3 layers of MaxPool2D layers, and 1 layer of fully connected layer. The detailed parameters of each layer of the network are shown in Table 1. The parameters and weights of the two sub-networks are set as shared parameters to ensure that they can be updated synchronously.

[0062] Table 1

[0063]

[0064] In a neural network, the setting of hyperparameters has a great influence on the effect of the model and is also affected by the input data. By continuously optimizing the weights and other hyperparameters in the model, the grid search method is selected in this embodiment to optimize the hyperparameters in the network. Grid search is an algorithm that automatically finds the optimal network hyperparameters. The grid search parameter optimization algorithm combines the grid search method with the cross-validation method, performs cross-validation on each parameter combination, calculates the model accuracy of the parameter combination, determines the best parameter combination, and completes the optimization of the neural network model. In this embodiment, the hyperparameters that need to be optimized, such as the learning rate, L2 regularization parameter, batch size, and the number of model training iterations, are all stored in the dictionary, such as Figure 3 ,The training sample data of single-phase users are used to evaluate the model performance of all hyperparameters to ensure that the set neural network can obtain the optimal results.

[0065] According to the detailed parameters of each layer of the convolutional neural network in Table 1, the final results of the single-phase user phase inference model are as follows: Figure 4 In this model, the input data are the household meter voltage curve image to be identified and the different phase voltage curves of the gateway meter in the substation.

[0066] Step 104, calculate the similarity between the characteristic vectors of the user voltage and the voltage data of different phases in the gateway table, and obtain the probability value of the user's phase, and determine the phase analysis result of the single-phase user in the low-voltage station area according to the probability value.

[0067] It should be noted that this embodiment uses Euclidean distance for similarity judgment and outputs the final recognition state of the sample. Euclidean distance measures the absolute distance between two points in a multidimensional space. The smaller the Euclidean distance, the more similar the two points are. The two feature vectors finally obtained after feature extraction by convolutional neural network are Vi = [vi1, vi2, ..., vin] and Vj = [vj1, vj2, ..., vjn], and the Euclidean distance calculation formula is as shown in formula (4).

[0068]

[0069] The similarity matrix sx = [s1, s2, s3] between the voltage curve image of the user meter to be identified and the voltage curves of different phases of the gateway meter in the substation is obtained through the twin neural network. The softmax activation function is matched with the cross entropy loss function in the classifier to predict the probability value of the phase, and finally the phase difference reasoning result of the user to be identified is output. The calculation formula of the softmax classifier is as shown in formula (5).

[0070]

[0071] This embodiment provides a user phase difference reasoning method based on grid search and VAE-SNN. Based on the power consumption data of the distribution area obtained from the distribution area automation database, the voltage curve image of the three-phase voltage of the gateway meter of the distribution area and the single-phase user on the low-voltage side is generated; the image feature extraction function and image recognition advantages of the neural network are used to construct a single-phase user phase difference reasoning model based on grid search and VAE-SNN. VAE reconstructs the original sample data through the autoencoder to generate a large number of diverse training samples, which increases the number of training samples and can accurately extract nonlinear sample features. Secondly, based on the optimization and determination of hyperparameters by grid search, an SNN model based on single sample learning is constructed, and the generated data is used to train the SNN model. According to the characteristics of the voltage curve image, two convolutional neural networks with shared weights and consistent structures are constructed by using the grid search optimization parameter method to extract and process the image features, and the feature vectors of the gateway meter voltage and the user voltage are output. A similarity calculation classifier is designed at the end of the network. The similarity of the feature vectors of the user voltage data and the different phase voltage data of the gateway meter is first calculated, and the probability value of the user's phase is obtained. The phase difference reasoning result of the single-phase user in the low-voltage area is determined according to the probability value.

[0072] The above is an embodiment of a user phase analysis method provided in an embodiment of the present application, and the following is an embodiment of a user phase analysis system provided in an embodiment of the present application.

[0073] See also Figure 5 , a user phase analysis system provided in an embodiment of the present application includes:

[0074] The sampling unit 201 is used to collect the gateway table of the substation area and the user's power consumption data and generate a sample curve image of the voltage data changing over time;

[0075] A reconstruction unit 202 is used to reconstruct the sample curve image through the VAE network to obtain a large number of diverse curve images as training sample data;

[0076] The extraction unit 203 is used to train the SNN model through the training sample data, optimize the hyperparameters of the SNN model by using the grid search optimization parameter method, thereby constructing a twin neural network to extract and process the image features, and output the feature vectors of the gate meter voltage and the user voltage;

[0077] The analysis unit 204 is used to calculate the similarity between the user voltage and the characteristic vectors of the different phase voltage data of the gateway table, and obtain the probability value of the user's phase, and determine the phase analysis result of the single-phase user in the low-voltage station area according to the probability value.

[0078] Furthermore, in an embodiment of the present application, a user phase analysis device is provided, the device comprising a processor and a memory:

[0079] The memory is used to store program code and transmit the program code to the processor;

[0080] The processor is used to execute the user phase analysis method described in the above method embodiment according to the instructions in the program code.

[0081] Furthermore, a computer-readable storage medium is provided in an embodiment of the present application, and the computer-readable storage medium is used to store program code, and the program code is used to execute the user phase analysis method described in the above method embodiment.

[0082] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0083] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0084] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0085] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0086] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0088] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0089] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A user phase analysis method, characterized in that: include: Collect the area gateway table and user power consumption data and generate a sample curve image of voltage data changing over time; Reconstructing the sample curve images through the VAE network to obtain a large number of diverse curve images as training sample data; The SNN model is trained by the training sample data, and the hyperparameters of the SNN model are optimized by using a grid search optimization parameter method, so as to construct a twin neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage; The similarity calculation classifier of the twin neural network is used to calculate the similarity of the feature vectors of the user voltage and the voltage data of different phases in the gateway table; in the calculation classifier, the softmax activation function is matched with the cross entropy loss function to predict the probability value of the phase, and finally the phase analysis result of the user to be identified is output.

2. The user phase analysis method according to claim 1, characterized in that: The sample curve image is reconstructed by the VAE variational autoencoder to obtain a large number of diverse curve images, which are used as training sample data, specifically including: The sample curve image is converted into a statistical distribution in the latent space through the encoder of the VAE network, and the statistical distribution parameters are the mean and variance; the statistical distribution parameters are effectively output variable distribution through the decoder of the VAE network to obtain training sample data.

3. The user phase analysis method according to claim 1, characterized in that: The SNN model is trained by the training sample data, and the hyperparameters of the SNN model are optimized by using a grid search optimization parameter method, so as to construct a convolutional neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage, specifically including: Constructing an SNN model based on single sample learning, and training the SNN model through the training sample data; The grid search method and cross-validation method are combined to cross-validate several parameter combinations of the SNN model to determine the optimal parameter combination, thereby constructing two twin neural networks with shared weights and consistent structures to extract and process image features and output feature vectors of gateway voltage and user voltage.

4. A user phase analysis system, characterized in that: include: The sampling unit is used to collect the power consumption data of the substation and the user and generate a sample curve image of the voltage data changing with time; A reconstruction unit, used for reconstructing the sample curve image through a VAE network to obtain a large number of diverse curve images as training sample data; An extraction unit, used to train the SNN model through the training sample data, optimize the SNN model hyperparameters by using a grid search optimization parameter method, thereby constructing a twin neural network to extract and process image features, and output feature vectors of the gate meter voltage and the user voltage; The analysis unit is used to calculate the similarity of the characteristic vectors of the user voltage and the voltage data of different phases of the gateway table through the similarity calculation classifier of the twin neural network; in the calculation classifier, the softmax activation function is matched with the cross entropy loss function to predict the probability value of the phase, and finally the phase analysis result of the user to be identified is output.

5. The user phase analysis system according to claim 4, characterized in that: The reconstruction unit is specifically used for: The sample curve image is converted into a statistical distribution in the latent space through the encoder of the VAE network, and the statistical distribution parameters are the mean and variance; the statistical distribution parameters are effectively output variable distribution through the decoder of the VAE network to obtain training sample data.

6. The user phase analysis system according to claim 4, characterized in that: The extraction unit is specifically used for: Constructing an SNN model based on single sample learning, and training the SNN model through the training sample data; The grid search method and cross-validation method are combined to cross-validate several parameter combinations of the SNN model to determine the optimal parameter combination, thereby constructing two twin neural networks with shared weights and consistent structures to extract and process image features and output feature vectors of gateway voltage and user voltage.

7. A user phase analysis device, characterized in that: The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the user phase analysis method described in any one of claims 1-3 according to the instructions in the program code.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and the program code is used to execute the user phase analysis method described in any one of claims 1-3.

Images