Power distribution network on-line fault studying and judging system based on Internet of Things

Abstract

The invention discloses a power distribution network online fault research and judgment system based on the Internet of Things, which belongs to the technical field of the Internet of Things and comprises an Internet of Things sensing module, a data acquisition preprocessing module, a topology modeling analysis module, a fault research and judgment module, a fault recovery strategy module and a communication interaction module. A power distribution network topology structure model is constructed through the topology modeling analysis module, the fault position can be accurately calculated through the fault detection, positioning and isolation unit in the fault research and judgment module, the optimal switch operation sequence is determined to isolate a fault area, the accurate positioning and isolation are beneficial to reducing the influence range of the fault on the power distribution network, and the fault diagnosis accuracy is improved. And the fault research and judgment module comprises a fault detection unit, a fault positioning unit and a fault isolation unit, fault positioning is carried out through an abnormal score function, fault possibility score calculation and combination with a graph-level representation vector, and the fault position can be rapidly and accurately determined.

Description

Technical Field

[0001] The present invention belongs to the technical field of the Internet of Things, and specifically refers to an online fault judgment system for a distribution network based on the Internet of Things. Background Technique

[0002] With the rapid development of smart grid and Internet of Things technologies, as a key component of the power grid system, the fault location and judgment technology of the distribution network has received increasing attention. The traditional method of troubleshooting the distribution network has problems such as long time consumption, low efficiency, and high cost, and it is difficult to meet the requirements of modern power systems for rapid fault response.

[0003] However, there are still certain defects in the existing online fault judgment system for the distribution network. The existing online fault judgment system for the distribution network relies on traditional fault detection methods and cannot accurately locate the specific location of the fault, resulting in an unnecessary expansion of the power outage range. It relies on a relatively simplified topological model and fails to fully collect or update in real time the geographical location, connection relationship, and detailed operation status information of the equipment in the distribution network, resulting in a large difference between the model and the actual distribution network, affecting the accuracy and real-time nature of fault judgment. When formulating a fault recovery strategy, it may lack sufficient intelligence and flexibility and is difficult to make the best decision based on the actual situation and fault type of the distribution network. Therefore, an online fault judgment system for the distribution network based on the Internet of Things is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an online fault judgment system for a distribution network based on the Internet of Things to solve the problems raised in the above background technique.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An online fault judgment system for a distribution network based on the Internet of Things, including an Internet of Things perception module, a data acquisition and preprocessing module, a topological modeling and analysis module, a fault judgment module, a fault recovery strategy module, and a communication and interaction module.

[0006] The Internet of Things perception module monitors the status of the distribution network in real time through deployed Internet of Things sensors.

[0007] The data acquisition and preprocessing module is used to collect the Internet of Things perception module in real time and perform preprocessing.

[0008] The topological modeling and analysis module is used to construct a topological structure model of the distribution network and perform topological structure analysis.

[0009] The fault judgment module determines the specific location of the fault by monitoring the abnormal situation of the distribution network in real time and calculates the optimal solution to isolate the fault area.

[0010] The fault recovery strategy module analyzes according to the fault judgment result and formulates a fault recovery strategy.

[0011] The communication interaction module is used to provide an intuitive operation interface, enabling operation and maintenance personnel to monitor the status of the distribution network and perform control operations.

[0012] Among them, the topology modeling and analysis module collects the basic data of the distribution network from the data acquisition and preprocessing module, obtains the location coordinates, geographical distribution information, and line data information of all devices in the distribution network, determines the key nodes in the distribution network according to the collected basic data, defines attributes for each key node, determines the physical connections between nodes according to the actual connection method of the distribution network, defines attributes for each edge, designs the model of the graphic database according to the definitions of nodes and edges, imports the collected data into the graphic database, and establishes the connection relationship between nodes in the graph database according to the actual connection method between devices to form the distribution network topology structure model;

[0013] According to the Internet of Things perception module, the status of the distribution network is monitored in real time, and the real-time monitoring data is combined with the constructed distribution network topology structure model to update the status of nodes and edges in the distribution network topology structure model.

[0014] The Internet of Things perception module can continuously monitor the status of the distribution network in real time through the deployed sensors, including key parameters such as voltage, current, and temperature, ensuring the real-time and accuracy of the data.

[0015] Among them, the topology modeling and analysis module performs distribution network topology structure analysis according to the constructed distribution network topology structure model, and extracts the features of nodes and edges from the distribution network topology structure model; Let an undirected graph be G=(V, E), where V represents the node set and E represents the edge set, and each node has an associated node feature vector , and each edge has an associated edge feature vector , and the new node feature vector of each node i at the l+1 layer is calculated through the graph neural network, and is expressed as:

[0016] ,

[0017] In the formula, represents the weight matrix of the l-th layer, represents the feature vector of node i, represents the weighted sum of the feature vectors of the neighbor node j of node i, represents the activation function, N(i) represents the neighbor set of node i, represents the normalization factor, represents the bias vector of the l-th layer.

[0018] Among them, according to the obtained new node feature vectors, the new node feature vectors of the last layer of all nodes are aggregated into a representation vector of the entire graph through pooling. The implementation formula of the graph-level representation vector g is as follows:

[0019] ,

[0020] In the formula, represents the set of new node feature vectors of all nodes in the last layer, represents the new node feature vector of node i in the last layer, represents the pooling function, represents the normalization coefficient, represents the summation of the feature vectors of all nodes, represents the total number of nodes, represents the new node feature vector of node i at the L-th layer.

[0021] The data acquisition and preprocessing module cleans, filters, and standardizes the original data. The topology modeling and analysis module then accurately models the topology of the distribution network through the graph neural network algorithm to ensure the accuracy of fault judgment.

[0022] Among them, the fault judgment module includes a fault detection unit, a fault location unit, and a fault isolation unit;

[0023] The fault detection unit collects data from the data acquisition and preprocessing module in real time for fault anomaly detection;

[0024] The fault location unit calculates the fault probability score based on the detection result of the fault detection unit, combines the fault probability score with the graph-level representation vector to determine the specific location of the fault;

[0025] The fault isolation unit calculates the optimal switch operation sequence according to the specific location of the fault;

[0026] For the fault detection unit, assume that the data collected from the data acquisition and preprocessing module in real time is , and define an anomaly score function to measure the degree to which the data at a given time point deviates from the normal range. The implementation formula is as follows:

[0027] ,

[0028] In the formula, represents the data point at time t 's anomaly score, represents the data point at time t, represents the weight of the i-th distribution network state feature, represents the value of the i-th distribution network state feature at time t, represents the mean value of the i-th distribution network state feature, represents the standard deviation of the i-th distribution network state feature, and n represents the number of distribution network state features.

[0029] Among them, the fault location unit calculates the fault possibility of each node according to the fault anomaly score. There are N nodes, and the possibility score of each node j for a fault The formula for realizing the fault possibility score is:

[0030] ,

[0031] In the formula, represents the anomaly score of node , N represents the total number of nodes, represents the natural exponential function, which is used to convert the anomaly score into a positive value and amplify the difference, represents the sum of the exponents of the anomaly scores of all nodes;

[0032] The specific fault location is determined by combining the obtained fault possibility score with the graph-level representation vector. The realization formula is:

[0033] ,

[0034] In the formula, represents the fault score of node j, represents the new node feature vector of node i in the last layer, g represents the graph-level representation vector, represents the similarity function, which calculates the similarity between vectors. According to the fault score all nodes are sorted, and the node with the highest score is the specific fault location.

[0035] The fault judgment module includes three units: fault detection, location, and isolation. Through the anomaly score function, calculation of the fault possibility score, and combination with the graph-level representation vector for fault location, the fault location can be determined quickly and accurately.

[0036] Among them, the fault isolation unit calculates the optimal switch operation sequence according to the specific fault location. Let the optimal switch state be The formula for optimal isolation is:

[0037] ,

[0038] Subject to:

[0039] ,

[0040] ,

[0041] In the formula, represents the switch state connecting node i and node j represents the switch cost, and λ represents the adjustment parameter represents the sum of costs of all switches represents the sum of products of fault scores and switch states and represents that each node can only be connected to one node

[0042] Among them, the fault recovery strategy module formulates a fault recovery strategy according to the location result and the optimal isolation result provided by the fault judgment, based on the states of the fault area and the non-fault area, executes the fault recovery strategy and monitors the quasi-state of the distribution network in real time, and obtains a graph-level representation vector through pooling aggregation, improving the accuracy and efficiency of fault judgment

[0043] Among them, the communication interaction module provides an intuitive operation interface, enabling operation and maintenance personnel to monitor the state of the distribution network and execute control operations, generating instructions according to the output of the fault recovery strategy module, sending the instructions to the control center, collecting the distribution network state and recovery progress information, and sending a status report to the control center

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows

[0045] 1. To solve the problems of insufficient real-time performance, difficult fault location, and low fault isolation and recovery efficiency of traditional distribution networks, and improve real-time performance, accuracy, and efficiency, the present invention constructs a distribution network topology structure model through the topology modeling and analysis module, and through the fault detection, location, and isolation units in the fault judgment module, can accurately calculate the fault location and determine the optimal switch operation sequence to isolate the fault area. This precise location and isolation helps to reduce the impact range of the fault on the distribution network and shorten the recovery time

[0046] 2. The present invention collects information such as the geographical location, connection relationship, and operating status of each device in the distribution network through the topology modeling and analysis module, constructs a topology model highly consistent with the actual distribution network, tracks the changes in the device status in the distribution network in real time, and automatically updates the topology model to ensure the accuracy of fault judgment

[0047] 3. The present invention uses a fault judgment module, which includes three units: fault detection, location, and isolation. By means of an anomaly score function, calculation of fault probability scores, and combination with graph-level representation vectors for fault location, it can quickly and accurately determine the fault location. The fault detection unit can analyze the collected data in real time and trigger the fault judgment process immediately once an anomaly is detected, shortening the time interval from fault discovery to judgment. The fault location unit can accurately locate the specific position where the fault occurs by calculating the fault probability scores and combining with graph-level representation vectors, reducing the possibility of misjudgment and missed judgment. The fault isolation unit calculates the optimal switch operation sequence based on the fault location to quickly isolate the fault area and prevent the fault from expanding and affecting non-fault areas.

[0048] 4. The present invention can intelligently formulate a fault recovery strategy according to the fault judgment result and the actual situation of the distribution network through the fault recovery strategy module, including steps such as fault isolation, load transfer, and power source recovery, ensuring the efficiency and safety of fault recovery. During the fault recovery process, the system can monitor the state of the distribution network in real time, including changes in parameters such as voltage, current, and load, to ensure the smooth implementation of the recovery strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic structural diagram of the online fault judgment system for distribution network based on the Internet of Things of the present invention;

[0050] Figure 2 It is a flowchart of the operation of the topology modeling analysis module of the online fault judgment system for distribution network based on the Internet of Things of the present invention;

[0051] Figure 3 It is a flowchart of the operation of the fault judgment module of the online fault judgment system for distribution network based on the Internet of Things of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment

[0054] Please refer to Figures 1-3 As shown, the present invention provides a technical solution: including an Internet of Things perception module, a data acquisition and preprocessing module, a topology modeling analysis module, a fault judgment module, a fault recovery strategy module, and a communication interaction module;

[0055] The Internet of Things (IoT) sensing module monitors the status of the distribution network in real time through deployed IoT sensors;

[0056] The data acquisition and preprocessing module is used to collect the IoT sensing module in real time and perform preprocessing;

[0057] The topology modeling and analysis module is used to construct a distribution network topology structure model and perform topology structure analysis;

[0058] The fault judgment module determines the specific location of the fault by monitoring the abnormal conditions of the distribution network in real time, and calculates the optimal solution to isolate the fault area;

[0059] The fault recovery strategy module analyzes according to the fault judgment result and formulates a fault recovery strategy;

[0060] The communication and interaction module is used to provide an intuitive operation interface, enabling operation and maintenance personnel to monitor the status of the distribution network and execute control operations.

[0061] Among them, the topology modeling and analysis module collects the basic data of the distribution network from the data acquisition and preprocessing module, obtains the position coordinates, geographical distribution information and line data information of all devices in the distribution network, determines the key nodes in the distribution network according to the collected basic data, defines attributes for each key node, determines the physical connections between nodes according to the actual connection method of the distribution network, defines attributes for each edge, designs the model of the graph database according to the definitions of nodes and edges, imports the collected data into the graph database, and establishes the connection relationship between nodes in the graph database according to the actual connection method between devices to form a distribution network topology structure model;

[0062] According to the status of the distribution network monitored in real time by the IoT sensing module, the real-time monitoring data is combined with the constructed distribution network topology structure model to update the status of nodes and edges in the distribution network topology structure model.

[0063] The IoT sensing module can continuously monitor the status of the distribution network in real time through deployed sensors, including key parameters such as voltage, current, and temperature, ensuring the real-time and accuracy of data.

[0064] Among them, the topology modeling and analysis module performs distribution network topology structure analysis according to the constructed distribution network topology structure model, and extracts the characteristics of nodes and edges from the distribution network topology structure model; Let an undirected graph be G=(V, E), where V represents the node set and E represents the edge set, and each node has an associated node feature vector , and each edge has an associated edge feature vector , and calculates the new node feature vector of each node i at the l+1 layer through a graph neural network At the l + 1 layer, it is represented as:

[0065] ,

[0066] In the formula, represents the weight matrix of the l-th layer, represents the feature vector of node i, represents the feature vector of neighbor node j of node i is the weighted sum of represents the activation function, and N(i) represents the neighbor set of node i, represents the normalization factor, represents the bias vector of the l-th layer.

[0067] Among them, according to the obtained new node feature vectors, the new node feature vectors of all nodes in the last layer are aggregated into the whole graph through pooling, and the formula for realizing the graph-level representation vector g is:

[0068] ,

[0069] In the formula, represents the set of new node feature vectors of all nodes in the last layer, represents the new node feature vector of node i in the last layer, represents the pooling function, represents the normalization coefficient, represents the summation of the feature vectors of all nodes, represents the total number of nodes, represents the new node feature vector of node i in the L-th layer.

[0070] The data acquisition and preprocessing module cleans, filters, and standardizes the original data, while the topology modeling and analysis module accurately models the topology structure of the distribution network through the graph neural network algorithm to ensure the accuracy of fault judgment.

[0071] Among them, the fault judgment module includes a fault detection unit, a fault location unit, and a fault isolation unit;

[0072] The fault detection unit collects data from the data acquisition and preprocessing module in real time for fault anomaly detection;

[0073] The fault location unit calculates the fault probability score based on the detection results of the fault detection unit, combines the fault probability score with the graph-level representation vector to determine the specific location of the fault;

[0074] The fault isolation unit calculates the optimal switch operation sequence according to the specific location of the fault;

[0075] The fault detection unit is configured to collect data in real time from the data acquisition and preprocessing module as , and define an anomaly score function to measure the degree to which the data at a given time point deviates from the normal range. The implementation formula is:

[0076] ,

[0077] In the formula, represents the anomaly score of the data point at time t , represents the data point at time t, represents the weight of the i-th distribution network state feature, represents the value of the i-th distribution network state feature at time t, represents the mean value of the i-th distribution network state feature, represents the standard deviation of the i-th distribution network state feature, and n represents the number of distribution network state features.

[0078] Among them, the fault location unit calculates the fault possibility of each node according to the fault anomaly score. There are N nodes, and the possibility score of each node j having a fault is calculated by the following formula:

[0079] ,

[0080] In the formula, represents the anomaly score of node , N represents the total number of nodes, represents the natural exponential function, which is used to convert the anomaly score into a positive value and amplify the difference, represents the sum of the exponents of the anomaly scores of all nodes;

[0081] The specific fault location is determined by combining the obtained fault possibility score with the graph-level representation vector. The implementation formula is:

[0082] ,

[0083] In the formula, represents the fault score of node j, represents the new node feature vector of node i at the last layer, g represents the graph-level representation vector, represents the similarity function, which calculates the similarity between vectors. According to the fault score all nodes are sorted, and the node with the highest score is the specific fault location.

[0084] The fault judgment module includes three units: fault detection, location, and isolation. It can quickly and accurately determine the fault location by means of an anomaly score function, calculation of fault probability scores, and combination with graph-level representation vectors for fault location.

[0085] Among them, the fault isolation unit calculates the optimal switch operation sequence according to the specific fault location. Let the optimal switch state be , and the optimal isolation implementation formula is:

[0086] ,

[0087] Subject to:

[0088] ,

[0089] ,

[0090] In the formula, represents the switch state connecting node i and node j, represents the cost of switch , λ represents the adjustment parameter, represents the sum of the costs of all switches, represents the sum of the products of the fault score and the switch state, and represent that each node can only be connected to one node.

[0091] Among them, the fault recovery strategy module formulates a fault recovery strategy according to the location result and the optimal isolation result provided by the fault judgment, based on the states of the fault area and the non-fault area, executes the fault recovery strategy and monitors the quasi-state of the distribution network in real time, and obtains the graph-level representation vector through pooling aggregation, improving the accuracy and efficiency of the fault judgment.

[0092] Among them, the communication and interaction module provides an intuitive operation interface, enabling operation and maintenance personnel to monitor the state of the distribution network and execute control operations, generating instructions according to the output of the fault recovery strategy module, sending the instructions to the control center, collecting the state and recovery progress information of the distribution network, and sending a status report to the control center.

[0093] In this example, specifically: The topology modeling and analysis module performs topology structure analysis of the distribution network according to the constructed distribution network topology structure model, and extracts the features of nodes and edges from the distribution network topology structure model; Let an undirected graph be G=(V, E), where V represents the set of nodes, E represents the set of edges, and each node has an associated node feature vector , and each edge has an associated edge feature vector , calculate the new node feature vector of each node i at the l+1 layer through the graph neural network , expressed at the l+1 layer as:

[0094] ,

[0095] In the formula, represents the weight matrix of the l-th layer, represents the feature vector of node i, represents the feature vector of neighbor node j of node i weighted sum of represents the activation function, N(i) represents the neighbor set of node i, represents the normalization factor, is , represents the bias vector of the l-th layer.

[0096] In this example, specifically: According to the obtained new node feature vector, the new node feature vectors of the last layer of all nodes are aggregated into the whole graph through pooling, and the formula for realizing the graph-level representation vector g is:

[0097] ,

[0098] In the formula, represents the set of new node feature vectors of all nodes in the last layer, represents the new node feature vector of node i in the last layer, represents the pooling function, represents the normalization coefficient, represents the sum of the feature vectors of all nodes, represents the total number of nodes, represents the new node feature vector of node i at the L-th layer.

[0099] In this example, specifically: The fault location unit calculates the fault possibility of each node according to the fault anomaly score. There are N nodes, and the possibility score of each node j having a fault , and the formula for realizing the fault possibility score is:

[0100] ,

[0101] In the formula, represents the anomaly score of node j, N represents the total number of nodes, represents the natural exponential function, which is used to convert the anomaly score into a positive value and amplify the difference, represents the sum of the exponents of all node anomaly scores;

[0102] Determine the specific fault location based on the obtained fault possibility score in combination with the graph-level representation vector. The implementation formula is as follows:

[0103] ,

[0104] In the formula, represents the fault score of node j, represents the new node feature vector of node i in the last layer, g represents the graph-level representation vector, represents the similarity function, which calculates the similarity between vectors. According to the fault score sort all nodes, and the node with the highest score is the specific fault location.

[0105] In this example, specifically: The fault isolation unit calculates the optimal switch operation sequence according to the specific fault location. Let the optimal switch state be , and the optimal isolation implementation formula is:

[0106] ,

[0107] Subject to:

[0108] ,

[0109] ,

[0110] In the formula, represents the switch state connecting node i and node j, represents the switch cost, λ represents the adjustment parameter, represents the sum of the costs of all switches, represents the sum of the products of the fault score and the switch state, and means that each node can only be connected to one node.

[0111] Working principle: The status of the distribution network is monitored in real time through the deployed Internet of Things sensors. The data acquisition and preprocessing module collects the data output by the Internet of Things sensing module in real time and performs preliminary processing such as data cleaning, denoising, and compression;

[0112] The topology modeling and analysis module constructs the topology structure model of the distribution network based on the basic data of the distribution network. Through technologies such as graph databases, the connection relationships between devices are represented in a graphical manner. The real-time monitoring data is combined with the topology structure model to dynamically update the states of nodes and edges. The fault judgment module includes a fault detection unit, a fault location unit, and a fault isolation unit. The fault detection unit uses algorithms such as the anomaly score function to detect anomalies in the data in real time and determine whether a fault has occurred in the distribution network. The fault location unit calculates the fault probability score and determines the specific location of the fault based on the detection results of the fault detection unit and in combination with algorithms such as graph neural networks. The fault isolation unit calculates the optimal switch operation sequence based on the fault location to isolate the fault area and prevent the spread of the fault. The fault recovery strategy module formulates a fault recovery strategy based on the location results and the optimal isolation results provided by the fault judgment module, executes the recovery strategy, monitors the state of the distribution network in real time, evaluates the recovery effect, and adjusts the recovery strategy as needed. The communication and interaction module provides an intuitive operation interface, enabling operation and maintenance personnel to conveniently monitor the state of the distribution network and execute control operations, receive the output instructions of the fault recovery strategy module, and send them to the control center.

[0113] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

[0114] The above describes the present invention and its embodiments. Such a description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and, without departing from the purpose of the present invention, design similar structural forms and embodiments to this technical solution without creative efforts, they should all fall within the protection scope of the present invention.

Claims

1. An online fault judgment system for a distribution network based on the Internet of Things, characterized in that: It includes an Internet of Things (IoT) perception module, a data acquisition and preprocessing module, a topology modeling and analysis module, a fault diagnosis module, a fault recovery strategy module, and a communication and interaction module; The IoT perception module monitors the status of the distribution network in real time through deployed IoT sensors; The data acquisition and preprocessing module is used to collect data from the IoT perception module in real time and perform preprocessing; The topology modeling and analysis module is used to construct a topology structure model of the distribution network and perform topology structure analysis; The fault diagnosis module monitors the abnormal conditions of the distribution network in real time, determines the specific location of the fault, and calculates the optimal solution to isolate the fault area; The fault recovery strategy module analyzes according to the fault diagnosis result and formulates a fault recovery strategy; The communication and interaction module is used to provide an intuitive operation interface for operation and maintenance personnel to monitor the status of the distribution network and execute control operations.

2. The online fault judgment system for a distribution network based on the Internet of Things according to claim 1, characterized in that: The topology modeling and analysis module collects the basic data of the distribution network from the data acquisition and preprocessing module, obtains the location coordinates, geographical distribution information, and line data information of all devices in the distribution network, determines the key nodes in the distribution network according to the collected basic data, defines attributes for each key node, determines the physical connections between nodes according to the actual connection method of the distribution network, defines attributes for each edge, designs the model of the graph database according to the definitions of nodes and edges, imports the collected data into the graph database, and establishes the connection relationship between nodes in the graph database according to the actual connection method between devices to form a topology structure model of the distribution network; According to the status of the distribution network monitored in real time by the IoT perception module, combines the real-time monitoring data with the constructed topology structure model of the distribution network, and updates the status of nodes and edges in the topology structure model of the distribution network.

3. The online fault judgment system for a distribution network based on the Internet of Things according to claim 2, wherein: The topological modeling analysis module performs topological structure analysis of the distribution network based on the constructed distribution network topological structure model, and extracts the features of nodes and edges from the distribution network topological structure model. Let an undirected graph be G=(V, E), where V represents the set of nodes and E represents the set of edges. Each node has an associated node feature vector , and each edge has an associated edge feature vector . The new node feature vector of each node i at the l+1 layer is calculated through a graph neural network , which is expressed at the l+1 layer as: ， In the formula, represents the weight matrix of the l-th layer, represents the feature vector of node i, represents the feature vector of neighbor node j of node i is the weighted sum of, represents the activation function, N(i) represents the neighbor set of node i, represents the normalization factor, represents the bias vector of the l-th layer.

4. The online fault judgment system for a distribution network based on the Internet of Things according to claim 3, wherein: According to the obtained new node feature vectors, pool the last-layer new node feature vectors of all nodes into an entire graph, and the formula for realizing the graph-level representation vector g is: ， In the formula, represents the set of new node feature vectors of all nodes in the last layer, represents the new node feature vector of node i in the last layer, represents the pooling function, represents the normalization coefficient, represents the summation of the feature vectors of all nodes, represents the total number of nodes, represents the new node feature vector of node i in the L-th layer.

5. The online fault judgment system for distribution network based on Internet of Things according to claim 1, wherein: The fault diagnosis module includes a fault detection unit, a fault location unit, and a fault isolation unit; The fault detection unit collects data from the data acquisition and preprocessing module in real time for fault anomaly detection; The fault location unit calculates the fault probability score according to the detection result of the fault detection unit, combines the fault probability score with the graph-level representation vector, and determines the specific location of the fault; The fault isolation unit calculates the optimal switch operation sequence according to the specific location of the fault; The fault detection unit is configured to collect data in real time from the data acquisition and preprocessing module as , and define an anomaly score function to measure the degree to which the data at a given time point deviates from the normal range. The implementation formula is as follows: ， In the formula, represents the anomaly score of the data point at time t , represents the data point at time t, represents the weight of the i-th distribution network state feature, represents the value of the i-th distribution network state feature at time t, represents the mean value of the i-th distribution network state feature, represents the standard deviation of the i-th distribution network state feature, and n represents the number of distribution network state features.

6. The online fault judgment system for distribution network based on Internet of Things according to claim 5, characterized in that: The fault location unit calculates the fault possibility of each node according to the fault anomaly score. There are N nodes, and the possibility score of each node j having a fault , and the implementation formula of the fault possibility score is: ， In the formula, represents the anomaly score of the node , N represents the total number of nodes, represents the natural exponential function, which is used to convert the anomaly score into a positive value and amplify the difference, represents the sum of the exponents of the anomaly scores of all nodes; Determine the specific location of the fault by combining the obtained fault probability score with the graph-level representation vector, and the formula is: ， In the formula, represents the fault score of node j, represents the new node feature vector of node i in the last layer, and g represents the graph-level representation vector, represents the similarity function, which calculates the similarity between vectors, and sorts all nodes according to the fault score The node with the highest score is the specific location of the node fault.

7. The online fault judgment system for distribution network based on Internet of Things according to claim 6, characterized in that: The fault isolation unit calculates the optimal switch operation sequence according to the specific fault location, and the optimal switch state is set as , and the optimal isolation implementation formula is: ， Subject to: ， ， In the formula, represents the switch state connecting node i and node j, represents the switch cost, λ represents the adjustment parameter, represents the sum of the costs of all switches, represents the sum of the products of the fault scores and the switch states, and represents that each node can only be connected to one node.

8. The online fault judgment system for distribution network based on Internet of Things according to claim 1, characterized in that: The fault recovery strategy module formulates a fault recovery strategy according to the location result and the optimal isolation result provided by the fault diagnosis, and according to the status of the fault area and the non-fault area, executes the fault recovery strategy and monitors the quasi-status of the distribution network in real time.

9. The online fault judgment system for distribution network based on Internet of Things according to claim 1, wherein: The communication and interaction module provides an intuitive operation interface for operation and maintenance personnel to monitor the status of the distribution network and execute control operations, generates instructions according to the output of the fault recovery strategy module, sends the instructions to the control center, collects the status and recovery progress information of the distribution network, and sends a status report to the control center.

Images