A distribution network equipment diagnosis and analysis method based on measurement data
Through the diagnostic analysis method based on measurement data, deep learning model and graph theory algorithm, the problem of fault monitoring of distribution network equipment is solved, the safe and stable operation of distribution network and the accuracy of fault diagnosis are improved, and the equipment status is realized Effective management.
Patent Information
- Application Number
- CN202411036239.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-07-31
AI Technical Summary
Distribution grid equipment failures may lead to safety hazards and economic losses in the power system. It is difficult for the existing technology to effectively monitor and manage the status of distribution grid equipment, affecting operating efficiency and energy utilization.
Using diagnostic analysis methods based on measurement data, the equipment topology relationship diagram is constructed by collecting and preprocessing real-time monitoring data, historical device status data and GIS data, and analyzing using graph theory algorithms and deep learning models (variant autoencoder VAE) Equipment abnormality, abnormality detection and fault prediction are performed.
It improves the monitoring and operation management of distribution network equipment, ensures the safe and stable operation of distribution network, enhances the accuracy and efficiency of fault diagnosis, and can better capture the complex characteristics of equipment status data.
Smart Images

Figure CN118569507B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of distribution network detection, and in particular to a distribution network equipment diagnosis and analysis method based on measurement data. Background Art
[0002] In order to improve the safety and economy of distribution network operation and assist distribution network operation and maintenance personnel in making correct analysis and decisions, it is necessary to monitor distribution network equipment and fully understand the various states of distribution network equipment. The discovery of distribution network equipment failures is closely related to the equipment operating status, health status, and environmental status.
[0003] Failure of distribution network equipment may lead to safety hazards in the power system, such as short circuits and overloads, which may cause fires, electrical accidents and other safety issues. It may also lead to power outages in certain areas of the distribution network, affecting the normal electricity demand of users, causing economic losses and inconvenience, affecting the operating efficiency of the distribution network, and may lead to increased energy loss and reduced energy utilization, affecting the economic efficiency of the distribution network. Summary of the invention
[0004] In order to solve the above problems, the purpose of the present invention is to provide a distribution network equipment diagnosis and analysis method based on measurement data, which can effectively improve the monitoring and operation management of distribution network equipment and further ensure the safe and stable operation of the distribution network.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] A distribution network equipment diagnosis and analysis method based on measurement data comprises the following steps:
[0007] S1: Collect comprehensive data from the distribution network system, including real-time monitoring data, historical equipment status data, fault event data and GIS data, and pre-process them;
[0008] S2: constructing a device topology relationship diagram based on the pre-processed distribution network GIS data, wherein the pre-processed distribution network GIS data includes the location and connection relationship of lines, transformers and switchgear, and using a graph theory algorithm to analyze the connection relationship and path between devices to obtain device topology relationship data;
[0009] S3: Based on the mining algorithm, analyze the correlation characteristics between the equipment topology relationship data and the historical equipment status data and fault event data, and build a training data set;
[0010] S4: Based on the deep learning model, build a device anomaly detection model and train it based on the training data set;
[0011] The device anomaly detection model is built based on the variational autoencoder VAE, as follows:
[0012] The training data set is used as input data. The VAE uses a multi-layer neural network structure as an encoder. The input data x is converted into a mean vector μ and a variance vector σ of the latent space through the encoder.
[0013] ;
[0014] ;
[0015] ;
[0016] Among them, h is the extracted feature; W enc is the weight matrix of the encoder, which is used to map the input data to the feature space; b enc is the bias term of the encoder, used to offset the data in the feature space; W μ The weight matrix for the encoder output mean vector; b μ The bias term for the encoder output mean vector; W σ The weight matrix for the encoder output variance vector; b σ is the bias term of the encoder output variance vector; ReLU is the activation function;
[0017] The VAE uses a multi-layer neural network structure as a decoder to reconstruct the sampled values of the latent space into input data;
[0018] ;
[0019] ;
[0020] Among them, z is the latent variable, is the decoder feature, is the reconstructed input data; W dec is the weight matrix of the decoder, which is used to map the latent variables back to the feature space; b dec is the bias term of the decoder, used to offset the data in the feature space; W out is the weight matrix of the decoder output layer, used to reconstruct data; b out is the bias term of the decoder output layer;
[0021] The overall loss is constructed by combining the reconstruction error and the distribution loss of the latent space:
[0022] ;
[0023] ;
[0024] ;
[0025] in, is the reconstruction error; N is the number of samples, It is The original data of samples; It is Reconstructed data of samples; represents the square of Euclidean distance; is the distribution loss of the latent space, J is the dimension of the latent space, and They are dimensional mean and variance; β is a hyperparameter used to balance the reconstruction error and the distribution loss of the latent space;
[0026] Use stochastic gradient descent to minimize the loss function, iteratively train the model, and adjust the model parameters to maximize the variational lower bound:
[0027] ;
[0028] in, are the model parameters for the tth iteration, is the learning rate, is the gradient of the loss function with respect to the parameters;
[0029] S5: Obtain real-time device status data, and perform detection based on the device anomaly detection model to obtain abnormal devices;
[0030] S6: Based on the abnormal equipment and fault type, visualization is performed and displayed in the equipment topology diagram, and the probability of other equipment failures is predicted based on the association model.
[0031] Furthermore, S1 is specifically as follows: obtaining real-time monitoring data, historical equipment status data, fault event data and GIS data from the distribution network system; integrating various types of data into a unified data warehouse and establishing data association relationships; cleaning the data, including removing outliers and filling missing values; and converting and standardizing the data format.
[0032] Further, the S2 is specifically:
[0033] S21: Based on the locations and connection relationships of the lines, transformers, and switchgear in the preprocessed distribution network GIS data, a topological relationship diagram between the devices is constructed, the devices are represented as nodes, the connection relationships are represented as edges, and a graph data structure is constructed;
[0034] S22: using the Floyd-Warshall algorithm to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices;
[0035] S23: Use visualization tools to display the device topology diagram in graphical form, including device nodes and connection lines.
[0036] Further, the S22 is specifically:
[0037] Convert the device topology graph data into the required format of the adjacency matrix table, including the connection relationship and path cost between nodes;
[0038] Traverse the device topology graph data, build an adjacency matrix based on the connection relationship and path cost between nodes, and fill the adjacency matrix according to the following formula Elements of represent the connection relationship and path cost between nodes;
[0039] ;
[0040] in, represents the path cost between node i and node j;
[0041] The Floyd-Warshall algorithm is used to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices:
[0042] (1) Initialize the path length and path cost between nodes;
[0043] (2) By traversing all nodes, update the path length and path cost between nodes;
[0044] (3) Repeat steps (1) and (2) until the shortest path and optimal path costs between all nodes remain stable;
[0045] ;
[0046] ;
[0047] in, is the shortest path between node i and node j, is the optimal path cost between node i and node j; k is the intermediate node between node i and node j;
[0048] Output the shortest path length and path cost matrix between all nodes, as well as path information.
[0049] Further, the S3 is specifically:
[0050] S31: integrating device topology relationship data, historical device status data, and fault event data into a transaction data set;
[0051] S32: construct an FP tree based on the transaction data set, compress the data set and retain the frequent item set information;
[0052] S33: Mining frequent itemsets and association rules from FP-trees.
[0053] Furthermore, during the VAE model training process, a mixed Gaussian distribution is used as a prior distribution to better capture the potential structure of the data, as follows:
[0054] The mixed Gaussian distribution consists of multiple Gaussian components, each of which represents a cluster of data;
[0055] Define the parameters of the mixed Gaussian distribution, including the mean, covariance matrix, and mixing coefficients of each Gaussian component;
[0056] When training the VAE model, the latent variable z is sampled from the mixed Gaussian distribution;
[0057] A Gaussian component k is randomly selected according to the mixing coefficient of the mixed Gaussian distribution, where , K represents the number of Gaussian components in the mixed Gaussian distribution, then the latent variable z is sampled from the distribution of Gaussian component k, and the mean of Gaussian component k is ,but
[0058]
[0059] in, is the Cholesky decomposition matrix of Gaussian component k, is a random noise term adopted from a standard normal distribution;
[0060] Calculate the KL divergence between the reconstructed distribution and the mixed Gaussian prior distribution to optimize the VAE model;
[0061] The probability density of the mixed Gaussian distribution is expressed as:
[0062]
[0063] in, is the mixing coefficient of the kth Gaussian component, and are the mean and covariance matrices of the kth Gaussian component respectively; represents the probability density function of the kth Gaussian component;
[0064] ;
[0065] Where D represents the dimension of the data. represents the mean vector of the kth Gaussian component, represents the determinant of the covariance matrix; exp() represents the natural exponential function; Represents the covariance matrix The inverse matrix of ; T represents transpose;
[0066] The KL divergence is used to measure the reconstruction distribution The difference between the prior distribution p(z) is:
[0067] .
[0068] Furthermore, the association model is constructed as follows:
[0069] The association relationship between devices is represented as a graph structure, where each node represents a device and the edge represents the association relationship between devices. Abnormal device data and other device status data are used as node features to construct a device association graph with features.
[0070] Use graph neural networks to learn device association graphs to capture association information between devices and dependencies between features.
[0071] Use abnormal device data and other device status data as training data, train the graph neural network model, optimize the model parameters to maximize the prediction accuracy, and obtain the association model
[0072] After the training is completed, the association model is used to predict the failure probability of other devices, which is expressed as:
[0073] (equipment failure | equipment status data) = (equipment status data) P (equipment failure | equipment status data) = f (equipment status data);
[0074] Where f is the mapping function learned by the graph neural network model, which is used to predict the probability of equipment failure.
[0075] The present invention has the following beneficial effects:
[0076] 1. The present invention effectively improves the monitoring and operation management of distribution network equipment, and further ensures the safe and stable operation of the distribution network;
[0077] 2. The present invention constructs a device topology diagram and analyzes the connection relationship and path between devices, which helps to deeply understand the correlation between distribution network devices, provides a more accurate basis for fault diagnosis, and constructs a device anomaly detection model based on a deep learning model, which can better capture the complex characteristics of device status data and improve the accuracy and efficiency of anomaly detection;
[0078] 3. By introducing the mixed Gaussian distribution as the prior distribution, the VAE model can better capture the potential structure of the data and improve the effects of anomaly detection and data generation. During the training process, it is necessary to ensure that the parameters of the mixed Gaussian distribution are appropriately selected and the KL divergence loss is optimized to achieve more accurate potential representation learning. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] Figure 1 is a flow chart of the method of the present invention;
[0080] Figure 2 It is a schematic diagram of the variational autoencoder VAE structure in one embodiment of the present invention. DETAILED DESCRIPTION
[0081] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0082] refer to Figure 1 The present invention provides a distribution network equipment diagnosis and analysis method based on measurement data, comprising the following steps:
[0083] S1: Collect comprehensive data from the distribution network system, including real-time monitoring data, historical equipment status data, fault event data and GIS data, and pre-process them;
[0084] S2: constructing a device topology relationship diagram based on the pre-processed distribution network GIS data, wherein the pre-processed distribution network GIS data includes the location and connection relationship of lines, transformers and switchgear, and using a graph theory algorithm to analyze the connection relationship and path between devices to obtain device topology relationship data;
[0085] S3: Based on the mining algorithm, analyze the correlation characteristics between the equipment topology relationship data and the historical equipment status data and fault event data, and build a training data set;
[0086] S4: Based on the deep learning model, build a device anomaly detection model and train it based on the training data set;
[0087] The device anomaly detection model is built based on the variational autoencoder VAE, as follows:
[0088] The training data set is used as input data. The VAE uses a multi-layer neural network structure as an encoder. The input data x is converted into a mean vector μ and a variance vector σ of the latent space through the encoder.
[0089] ;
[0090] ;
[0091] ;
[0092] Among them, h is the extracted feature; W enc is the weight matrix of the encoder, which is used to map the input data to the feature space; b enc is the bias term of the encoder, used to offset the data in the feature space; W μ The weight matrix for the encoder output mean vector; b μ The bias term for the encoder output mean vector; W σ The weight matrix for the encoder output variance vector; b σ is the bias term of the encoder output variance vector; ReLU is the activation function;
[0093] The VAE uses a multi-layer neural network structure as a decoder to reconstruct the sampled values of the latent space into input data;
[0094] ;
[0095] ;
[0096] Among them, z is the latent variable, is the decoder feature, is the reconstructed input data; W dec is the weight matrix of the decoder, which is used to map the latent variables back to the feature space; b dec is the bias term of the decoder, used to offset the data in the feature space; W out is the weight matrix of the decoder output layer, used to reconstruct data; b out is the bias term of the decoder output layer;
[0097] The overall loss is constructed by combining the reconstruction error and the distribution loss of the latent space:
[0098] ;
[0099] ;
[0100] ;
[0101] in, is the reconstruction error; N is the number of samples, It is The original data of samples; It is Reconstructed data of samples; represents the square of Euclidean distance; is the distribution loss of the latent space, J is the dimension of the latent space, and They are dimensional mean and variance; β is a hyperparameter used to balance the reconstruction error and the distribution loss of the latent space;
[0102] Use stochastic gradient descent to minimize the loss function, iteratively train the model, and adjust the model parameters to maximize the variational lower bound:
[0103] ;
[0104] in, are the model parameters for the tth iteration, is the learning rate, is the gradient of the loss function with respect to the parameters;
[0105] S5: Obtain real-time device status data, and perform detection based on the device anomaly detection model to obtain abnormal devices;
[0106] S6: Based on the abnormal equipment and fault type, visualization is performed and displayed in the equipment topology diagram, and the probability of other equipment failures is predicted based on the association model.
[0107] In this embodiment, S1 specifically includes: obtaining real-time monitoring data, historical equipment status data, fault event data and GIS data from the distribution network system; integrating various types of data into a unified data warehouse, establishing data association relationships (connection relationships between devices, equipment type information, etc. Use primary keys and foreign keys to ensure the association and consistency between data); cleaning the data, including removing outliers and filling missing values; converting and standardizing the data.
[0108] In this embodiment, the S2 is specifically:
[0109] S21: Based on the locations and connection relationships of the lines, transformers, and switchgear in the preprocessed distribution network GIS data, a topological relationship diagram between the devices is constructed, the devices are represented as nodes, the connection relationships are represented as edges, and a graph data structure is constructed;
[0110] S22: using the Floyd-Warshall algorithm to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices;
[0111] S23: Use visualization tools to display the device topology diagram in graphical form, including device nodes and connection lines.
[0112] In this embodiment, the S22 is specifically:
[0113] Convert the device topology graph data into the required format of the adjacency matrix table, including the connection relationship and path cost between nodes;
[0114] Traverse the device topology graph data, build an adjacency matrix based on the connection relationship and path cost between nodes, and fill the adjacency matrix according to the following formula Elements of represent the connection relationship and path cost between nodes;
[0115] ;
[0116] in, represents the path cost between node i and node j;
[0117] The Floyd-Warshall algorithm is used to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices:
[0118] (1) Initialize the path length and path cost between nodes;
[0119] (2) By traversing all nodes, update the path length and path cost between nodes;
[0120] (3) Repeat steps (1) and (2) until the shortest path and optimal path costs between all nodes remain stable;
[0121] ;
[0122] ;
[0123] in, is the shortest path between node i and node j, is the optimal path cost between node i and node j; k is the intermediate node between node i and node j;
[0124] Output the shortest path length and path cost matrix between all nodes, as well as path information.
[0125] In this embodiment, the S3 is specifically:
[0126] S31: integrating device topology relationship data, historical device status data, and fault event data into a transaction data set;
[0127] S32: construct an FP tree based on the transaction data set, compress the data set and retain the frequent item set information;
[0128] S33: Mining frequent itemsets and association rules from FP-trees.
[0129] In this embodiment, during the VAE model training process, a mixed Gaussian distribution is used as a prior distribution to better capture the potential structure of the data, as follows:
[0130] The mixed Gaussian distribution consists of multiple Gaussian components, each of which represents a cluster of data;
[0131] Define the parameters of the mixed Gaussian distribution, including the mean, covariance matrix, and mixing coefficients of each Gaussian component;
[0132] When training the VAE model, the latent variable z is sampled from the mixed Gaussian distribution;
[0133] A Gaussian component k is randomly selected according to the mixing coefficient of the mixed Gaussian distribution, where , K represents the number of Gaussian components in the mixed Gaussian distribution, then the latent variable z is sampled from the distribution of Gaussian component k, and the mean of Gaussian component k is ,but
[0134]
[0135] in, is the Cholesky decomposition matrix of Gaussian component k, is a random noise term adopted from a standard normal distribution;
[0136] Calculate the KL divergence between the reconstructed distribution and the mixed Gaussian prior distribution to optimize the VAE model;
[0137] The probability density of the mixed Gaussian distribution is expressed as:
[0138]
[0139] in, is the mixing coefficient of the kth Gaussian component, and are the mean and covariance matrices of the kth Gaussian component respectively; represents the probability density function of the kth Gaussian component;
[0140] ;
[0141] Where D represents the dimension of the data. represents the mean vector of the kth Gaussian component, represents the determinant of the covariance matrix; exp() represents the natural exponential function; Represents the covariance matrix The inverse matrix of ; T represents transpose;
[0142] The KL divergence is used to measure the reconstruction distribution The difference between the prior distribution p(z) is:
[0143] .
[0144] In this embodiment, the association model is constructed as follows:
[0145] The association relationship between devices is represented as a graph structure, where each node represents a device and the edge represents the association relationship between devices. Abnormal device data and other device status data are used as node features to construct a device association graph with features.
[0146] Use graph neural networks to learn device association graphs to capture association information between devices and dependencies between features.
[0147] Use abnormal device data and other device status data as training data, train the graph neural network model, optimize the model parameters to maximize the prediction accuracy, and obtain the association model;
[0148] After the training is completed, the association model is used to predict the failure probability of other devices, which is expressed as:
[0149] (equipment failure | equipment status data) = (equipment status data) P (equipment failure | equipment status data) = f (equipment status data);
[0150] Where f is the mapping function learned by the graph neural network model, which is used to predict the probability of equipment failure.
[0151] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0152] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0153] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0155] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.
Claims
1. A distribution network equipment diagnosis and analysis method based on measurement data, characterized in that: The following steps are involved: S1: Collect comprehensive data from the distribution network system, including real-time monitoring data, historical equipment status data, fault event data and GIS data, and pre-process them; S2: constructing a device topology relationship diagram based on the pre-processed distribution network GIS data, wherein the pre-processed distribution network GIS data includes the location and connection relationship of lines, transformers and switchgear, and using a graph theory algorithm to analyze the connection relationship and path between devices to obtain device topology relationship data; S3: Based on the mining algorithm, analyze the correlation characteristics between the equipment topology relationship data and the historical equipment status data and fault event data, and build a training data set; S4: Based on the deep learning model, build a device anomaly detection model and train it based on the training data set; The device anomaly detection model is built based on the variational autoencoder VAE, as follows: The training data set is used as input data. The VAE uses a multi-layer neural network structure as an encoder. The input data x is converted into a mean vector μ and a variance vector σ of the latent space through the encoder. ; ; ; Among them, h is the extracted feature; W enc is the weight matrix of the encoder, which is used to map the input data to the feature space; b enc is the bias term of the encoder, used to offset the data in the feature space; W μ The weight matrix for the encoder output mean vector; b μ The bias term for the encoder output mean vector; W σ The weight matrix for the encoder output variance vector; b σ is the bias term of the encoder output variance vector; ReLU is the activation function; The VAE uses a multi-layer neural network structure as a decoder to reconstruct the sampled values of the latent space into input data; ; ; Among them, z is the latent variable, is the decoder feature, is the reconstructed input data; W dec is the weight matrix of the decoder, which is used to map the latent variables back to the feature space; b dec is the bias term of the decoder, used to offset the data in the feature space; W out is the weight matrix of the decoder output layer, used to reconstruct data; b out is the bias term of the decoder output layer; The overall loss is constructed by combining the reconstruction error and the distribution loss of the latent space: ; ; ; in, is the reconstruction error; N is the number of samples, It is The original data of samples; It is Reconstructed data of samples; represents the square of Euclidean distance; is the distribution loss of the latent space, J is the dimension of the latent space, and They are dimensional mean and variance; β is a hyperparameter used to balance the reconstruction error and the distribution loss of the latent space; Use stochastic gradient descent to minimize the loss function, iteratively train the model, and adjust the model parameters to maximize the variational lower bound: ; in, are the model parameters for the tth iteration, is the learning rate, is the gradient of the loss function with respect to the parameters; S5: Obtain real-time device status data, and perform detection based on the device anomaly detection model to obtain abnormal devices; S6: Based on the abnormal equipment and fault type, visualization is performed and displayed in the equipment topology diagram, and the probability of other equipment failures is predicted based on the association model.
2. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 1, characterized in that: The S1 specifically includes: obtaining real-time monitoring data, historical equipment status data, fault event data and GIS data from the distribution network system; integrating various types of data into a unified data warehouse and establishing data association relationships; cleaning the data, including removing outliers and filling missing values; and converting and standardizing the data format.
3. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 1, characterized in that: The S2 is specifically: S21: Based on the locations and connection relationships of the lines, transformers, and switchgear in the preprocessed distribution network GIS data, a topological relationship diagram between the devices is constructed, the devices are represented as nodes, the connection relationships are represented as edges, and a graph data structure is constructed; S22: using the Floyd-Warshall algorithm to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices; S23: Use visualization tools to display the device topology diagram in graphical form, including device nodes and connection lines.
4. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 3, characterized in that: The S22 is specifically: Convert the device topology graph data into the required format of the adjacency matrix table, including the connection relationship and path cost between nodes; Traverse the device topology graph data, build an adjacency matrix based on the connection relationship and path cost between nodes, and fill the adjacency matrix according to the following formula Elements of represent the connection relationship and path cost between nodes; ; in, represents the path cost between node i and node j; The Floyd-Warshall algorithm is used to analyze the device topology diagram to obtain the connection relationship, path length and path cost between devices: (1) Initialize the path length and path cost between nodes; (2) By traversing all nodes, update the path length and path cost between nodes; (3) Repeat steps (1) and (2) until the shortest path and optimal path costs between all nodes remain stable; ; ; in, is the shortest path between node i and node j, is the optimal path cost between node i and node j; k is the intermediate node between node i and node j; Output the shortest path length and path cost matrix between all nodes, as well as path information.
5. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 1, characterized in that: The S3 is specifically: S31: integrating device topology relationship data, historical device status data, and fault event data into a transaction data set; S32: construct an FP tree based on the transaction data set, compress the data set and retain the frequent item set information; S33: Mining frequent itemsets and association rules from FP-trees.
6. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 1, characterized in that: During VAE training, a mixed Gaussian distribution is used as a prior distribution to better capture the potential structure of the data, as follows: The mixed Gaussian distribution consists of multiple Gaussian components, each of which represents a cluster of data; Define the parameters of the mixed Gaussian distribution, including the mean, covariance matrix, and mixing coefficients of each Gaussian component; When training the VAE model, the latent variable z is sampled from the mixed Gaussian distribution; A Gaussian component k is randomly selected according to the mixing coefficient of the mixed Gaussian distribution, where , K represents the number of Gaussian components in the mixed Gaussian distribution, then the latent variable z is sampled from the distribution of Gaussian component k, and the mean of Gaussian component k is ,but in, is the Cholesky decomposition matrix of Gaussian component k, is a random noise term adopted from a standard normal distribution; Calculate the KL divergence between the reconstructed distribution and the mixed Gaussian prior distribution to optimize the VAE model; The probability density of the mixed Gaussian distribution is expressed as: in, is the mixing coefficient of the kth Gaussian component, and are the mean and covariance matrices of the kth Gaussian component respectively; represents the probability density function of the kth Gaussian component; ; Where D represents the dimension of the data. represents the mean vector of the kth Gaussian component, represents the determinant of the covariance matrix; exp() represents the natural exponential function; Represents the covariance matrix The inverse matrix of ; T represents transpose; The KL divergence is used to measure the reconstruction distribution The difference between the prior distribution p(z) is: 。 7. A distribution network equipment diagnosis and analysis method based on measurement data according to claim 1, characterized in that: The association model is constructed as follows: The association relationship between devices is represented as a graph structure, where each node represents a device and the edge represents the association relationship between devices. Abnormal device data and other device status data are used as node features to construct a device association graph with features. Use graph neural networks to learn device association graphs to capture association information between devices and dependencies between features. Use abnormal device data and other device status data as training data, train the graph neural network model, optimize the model parameters to maximize the prediction accuracy, and obtain the association model; After the training is completed, the association model is used to predict the failure probability of other devices, which is expressed as: (Equipment failure | Equipment status data) = (Equipment status data) P (Equipment failure | Equipment status data) = f (Equipment status data) Where f is the mapping function learned by the graph neural network model, which is used to predict the probability of equipment failure.
