Intelligent power distribution network fault analysis and maintenance system based on Internet of Things

Abstract

The invention discloses an intelligent power distribution network fault analysis and maintenance system based on the Internet of Things, which belongs to the technical field of power distribution network fault analysis and maintenance and comprises a multi-source data acquisition preprocessing module, a dynamic digital twin modeling module, an edge cloud fusion analysis module, a transient fault topology positioning module, a self-healing control module and an intelligent maintenance scheduling module. According to the method, fault features are extracted through edge-side AI reasoning, global anomaly detection is realized in combination with cloud big data analysis, anomaly modes under different working conditions are adapted through a dynamic balance coefficient, and fault type tags are quickly generated through weighted fusion of feature vectors in edge-side real-time reasoning, so that the fault detection accuracy is improved. The global anomaly scoring mechanism greatly improves the fault classification accuracy, updates the node state and the probability density function by injecting a dynamic digital twin model, quantifies the distribution characteristics of the faults in space coordinates in combination with the topological weight, the fault influence function and the space attenuation coefficient, and accurately recognizes the fault high-incidence area.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network fault analysis and maintenance, and specifically refers to an intelligent distribution network fault analysis and maintenance system based on the Internet of Things. Background Art

[0002] With the acceleration of urbanization and the diversification of electricity demand, higher requirements are placed on the reliability and intelligence level of the distribution network;

[0003] However, there are still certain defects in the existing distribution network fault analysis and maintenance. The existing distribution network fault analysis model uses a static topology model and builds a network structure based on fixed device connection relationships. It cannot reflect the topology adjustments caused by distributed power grid connection, load fluctuations or equipment status changes in real time. It mostly relies on cloud-based centralized analysis or edge-side local reasoning, lacks a global anomaly detection mechanism, and uses a fixed-rule fault location algorithm. It does not combine the topological weights and fault impact functions of the dynamic digital twin model, and cannot quantify the distribution characteristics of the fault in spatial coordinates, resulting in large positioning deviations. The existing switch action instructions are transmitted through traditional communication protocols, which lack anti-tampering and traceability mechanisms, resulting in security risks in the execution process. Therefore, an intelligent distribution network fault analysis and maintenance system based on the Internet of Things is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent distribution network fault analysis and maintenance system based on the Internet of Things to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: an IoT-based intelligent distribution network fault analysis and maintenance system, comprising a multi-source data acquisition and preprocessing module, a dynamic digital twin modeling module, an edge-cloud fusion analysis module, a transient fault topology location module, a self-healing control module, and an intelligent maintenance scheduling module;

[0006] The multi-source data acquisition preprocessing module is used to collect electrical parameters, environmental data and spatial displacement of the distribution network, and perform denoising and standardization processing to generate a spatiotemporal fusion feature matrix;

[0007] The dynamic digital twin modeling module constructs a dynamic digital twin model of the distribution network topology based on the spatiotemporal fusion feature matrix, and updates the node status and network parameters in real time;

[0008] The edge-cloud fusion analysis module is used to extract fault features through edge-side AI reasoning, and combines cloud-side big data analysis to perform global anomaly detection on the dynamic digital twin model to generate fault type labels;

[0009] The instantaneous fault topology location module is used to combine the fault type label and the topological structure data of the dynamic digital twin modeling module to calculate the spatial position probability distribution of the fault section and output the fault location coordinates;

[0010] The self-healing control module is used to generate a switch action instruction sequence according to the coordinates of the fault section, and execute isolation and power restoration operations after verification through the blockchain trusted instruction chain;

[0011] The intelligent maintenance scheduling module is used to obtain the health status and historical fault data of the equipment monitored by the self-healing control module, generate a maintenance task priority queue and provide feedback.

[0012] Preferably, the multi-source data acquisition and preprocessing module collects electrical parameters, environmental data and spatial displacement of the distribution network in real time through IoT devices, integrates heterogeneous data streams through edge computing nodes, and performs data cleaning, outlier detection, correction and denoising;

[0013] Feature extraction is performed on the preprocessed data, and the high-dimensional features are reduced in dimension through principal component analysis. The reduced spatial features are spliced ​​with the temporal features to generate a spatiotemporal fusion feature matrix.

[0014] Preferably, the multi-source data acquisition and preprocessing module is wirelessly connected to the edge-cloud fusion analysis module and the dynamic digital twin modeling module, the dynamic digital twin modeling module is wirelessly connected to the edge-cloud fusion analysis module, the instantaneous fault topology positioning module and the self-healing control module, the edge-cloud fusion analysis module is wirelessly connected to the instantaneous fault topology positioning module and the self-healing control module, the instantaneous fault topology positioning module is wirelessly connected to the self-healing control module, the self-healing control module is wirelessly connected to the intelligent maintenance scheduling module, and the intelligent maintenance scheduling module is wirelessly connected to the multi-source data acquisition and preprocessing module.

[0015] Preferably, the dynamic digital twin modeling module, the topology structure initialization step includes: the spatiotemporal fusion feature matrix output by the multi-source data acquisition preprocessing module, including electrical parameters, environmental data, spatial displacement and timestamp;

[0016] An initial topology graph is constructed based on BIM data, where nodes represent devices and edges represent electrical connection relationships. The geometric properties of the topology graph are calibrated by combining the spatial displacement in the spatiotemporal fusion feature matrix, and a graph convolutional network model is constructed. The spatiotemporal fusion feature matrix is ​​input and the device status prediction is output.

[0017] Preferably, the dynamic digital twin modeling module builds a dynamic digital twin model of the distribution network topology, comprising: simulating the voltage distribution and power flow physical characteristics of the distribution network through electromagnetic transient simulation, using the electrical parameters in the time-space fusion feature matrix as simulation input, and dynamically adjusting the model parameters;

[0018] The spatiotemporal fusion feature matrix is ​​input into the graph convolutional network model to learn the topological dependencies between devices. The node weights are dynamically assigned through the attention mechanism to capture the impact of key devices.

[0019] Perform weighted fusion of physical simulation results and the output of the graph convolutional network model to generate a hybrid digital twin model;

[0020] When a distributed power source is detected to be connected to the grid, the topology is dynamically adjusted and the adjacency matrix is ​​updated through a dynamic graph neural network;

[0021] Update node status according to real-time electrical parameters, dynamically adjust model parameters based on equipment health status, calculate power distribution of distribution network through forward-backward substitution method based on real-time load data and topology structure, calculate the impact of key parameters on voltage stability, and dynamically adjust control strategy.

[0022] Preferably, the edge-cloud fusion analysis module extracts fault features by performing feature extraction on the spatiotemporal fusion feature matrix on the edge side and performing weighted fusion to form a fault feature vector. The implementation formula is:

[0023]

[0024] In the formula, F represents the fault feature vector, α i represents the dynamic weight of the i-th category feature, f i represents the characteristics of the i-th type of source data, β represents the global characteristics of the spatiotemporal correlation characteristics, and g(x,y) represents the spatiotemporal correlation function, which maps the spatial coordinate x and the time series y into high-level features.

[0025] Preferably, the edge-cloud fusion analysis module implements the dynamic threshold anomaly scoring step: calculating the global anomaly score according to the fault feature vector F, and the implementation formula is:

[0026]

[0027] In the formula, S represents the global anomaly score, γ represents the dynamic balance coefficient, and μ nl ,σ nl represents the characteristic mean and standard deviation under normal conditions, γ represents the dynamic balance coefficient, d(F,C cy ) represents the minimum Euclidean distance from the current fault feature vector F to the abnormal cluster, d max represents the normalization factor.

[0028] Preferably, the edge-cloud fusion analysis module generates a fault type label based on the global anomaly score, and the implementation formula is:

[0029]

[0030] In the formula, R represents the fault type label, L represents the predefined fault type set, and δ l represents the confidence weight of the lth type of fault, represents the reference anomaly score of the lth type fault, θ l Represents the tolerance threshold of the lth type; the generated fault type label is injected into the dynamic digital twin model to update the node status.

[0031] Preferably, the instantaneous fault topology location model, the spatial location probability of the fault section is realized by the following steps: the spatial location probability of the fault section is calculated according to the topological structure and fault type label of the dynamic digital twin modeling module, and the realization formula is:

[0032]

[0033] In the formula, P(x) represents the probability density of the fault at the spatial coordinate x, W i Represents the topological weight of node and edge i, I(R i ) represents the influence function of the fault type label R on the node and edge i, d(x,x i ) represents the geometric distance between the spatial coordinate x and the nodes and edges, λ represents the spatial attenuation coefficient, and N is the total number of nodes and edges in the dynamic topology.

[0034] Preferably, the self-healing control module receives the fault type label and fault section coordinates output by the edge-cloud fusion analysis module, maps the fault coordinates to the topological structure in the dynamic digital twin model, determines the fault impact range, generates isolation instructions and recovery instructions based on the fault type and topological relationship, plans the shortest recovery path through a graph theory algorithm, generates a switch action sequence, and encapsulates the switch action instruction sequence as a blockchain transaction, including:

[0035] Command content: switch ID, action type, and expected execution time;

[0036] Signature information: digital signature of the control center;

[0037] Timestamp: ensure the timeliness of instructions;

[0038] Hash value: the hash value of the transaction content;

[0039] The verified instruction transactions are synchronously stored through blockchain nodes to ensure that the data cannot be tampered with, and the verified instructions are sent to the target switch device through the MQTT protocol.

[0040] Preferably, the intelligent maintenance scheduling module obtains the equipment health status output by the self-healing control module, reads historical fault data from the database or log system, and sorts the real-time data and historical data by timestamp to ensure time consistency;

[0041] Based on real-time and historical data, the system extracts equipment health indicators, uses a neural network model, and combines historical fault data to train a health prediction model. This outputs the current health score of the equipment and prioritizes tasks.

[0042] Task allocation is optimized through genetic algorithms to ensure maximum resource utilization, and a balance is struck between minimizing power outage time and maximizing resource utilization. The particle swarm optimization algorithm is used to generate the optimal scheduling plan, and the execution results are fed back to the multi-source data acquisition preprocessing module and the dynamic digital twin modeling module for optimization.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This invention constructs an initial topology map using BIM data and calibrates geometric properties by combining the spatial displacement in the spatiotemporal fusion feature matrix. This addresses the limitation of traditional static topology models that cannot dynamically adapt to grid changes. By introducing graph convolutional networks and attention mechanisms, and through hybrid modeling of physical simulation and graph neural networks, it achieves device state prediction and dynamic allocation of topological weights. It calculates power distribution through forward and backward substitution, and dynamically adjusts control strategies based on real-time load data, significantly improving the prediction accuracy of the digital twin model on the impact of voltage stability.

[0045] 2. This invention extracts fault features through edge-side AI reasoning and combines it with cloud-based big data analysis to achieve global anomaly detection. It adapts to abnormal patterns under different working conditions through dynamic balancing coefficients, while real-time reasoning on the edge quickly generates fault type labels by weighted fusion feature vectors. The global anomaly scoring mechanism significantly improves the accuracy of fault classification. At the same time, by injecting a dynamic digital twin model to update node status, the intelligent level of distribution network fault diagnosis is significantly improved.

[0046] 3. The present invention calculates the spatial location probability distribution of the fault section by associating the fault type label with the dynamic digital twin topology structure. Based on the probability density function, combined with the topological weight, fault impact function and spatial attenuation coefficient, it quantifies the distribution characteristics of the fault in the spatial coordinates, accurately identifies the high-incidence areas of faults, and adapts to the uncertainty brought about by the change of distribution network topology through dynamic topological weight adjustment. The spatial attenuation coefficient and geometric distance parameters are used to significantly improve the positioning accuracy. The output fault location coordinates can be directly used for the switching action planning of the self-healing control module, shortening the time window for fault isolation and power restoration, and effectively reducing the scope of power outages.

[0047] 4. This invention verifies switch action commands through a blockchain-based trusted instruction chain, addressing the issues of command tampering and untraceable execution in traditional self-healing control systems. Its core process includes fault segment coordinate mapping, isolation and recovery command generation, graph-based path planning, and blockchain transaction encapsulation and verification. Command content is digitally signed and timestamped to ensure timeliness and legitimacy, while transaction hashes ensure data immutability. Ultimately, commands are distributed to physical devices via the MQTT protocol for execution. By combining graph-based algorithms with dynamic topology, the system optimizes switch action sequences and significantly improves the efficiency of fault isolation and recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a schematic diagram of the structure of the intelligent distribution network fault analysis and maintenance system based on the Internet of Things of the present invention;

[0049] Figure 2 This is the operating process of the intelligent distribution network fault analysis and maintenance system based on the Internet of Things of the present invention Figure 1 ;

[0050] Figure 3 This is the operating process of the intelligent distribution network fault analysis and maintenance system based on the Internet of Things of the present invention Figure 2 ;

[0051] Figure 4 This is the operating process of the intelligent distribution network fault analysis and maintenance system based on the Internet of Things of the present invention Figure 3 . DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] Example

[0054] See also Figures 1-4 As shown, the present invention provides a technical solution: including a multi-source data acquisition and preprocessing module, a dynamic digital twin modeling module, an edge-cloud fusion analysis module, a transient fault topology location module, a self-healing control module and an intelligent maintenance scheduling module;

[0055] The multi-source data acquisition and preprocessing module is used to collect electrical parameters, environmental data and spatial displacement of the distribution network, and perform denoising and standardization processing to generate a spatiotemporal fusion feature matrix;

[0056] The dynamic digital twin modeling module constructs a dynamic digital twin model of the distribution network topology based on the spatiotemporal fusion feature matrix, and updates the node status and network parameters in real time;

[0057] The edge-cloud fusion analysis module is used to extract fault features through edge-side AI reasoning, and combines cloud-side big data analysis to perform global anomaly detection on the dynamic digital twin model to generate fault type labels;

[0058] The instantaneous fault topology location module is used to combine the fault type label and the topological structure data of the dynamic digital twin modeling module to calculate the spatial position probability distribution of the fault section and output the fault location coordinates;

[0059] The self-healing control module is used to generate a switch action instruction sequence according to the coordinates of the fault section, and execute isolation and power restoration operations after verification through the blockchain trusted instruction chain;

[0060] The intelligent maintenance scheduling module is used to obtain the health status and historical fault data of the equipment monitored by the self-healing control module, generate a maintenance task priority queue and provide feedback.

[0061] Preferably, the multi-source data acquisition and preprocessing module collects electrical parameters, environmental data and spatial displacement of the distribution network in real time through IoT devices, integrates heterogeneous data streams through edge computing nodes, and performs data cleaning, outlier detection, correction and denoising;

[0062] Feature extraction is performed on the preprocessed data, and the high-dimensional features are reduced in dimension through principal component analysis. The reduced spatial features are spliced ​​with the temporal features to generate a spatiotemporal fusion feature matrix.

[0063] In this embodiment, the multi-source data acquisition and preprocessing module is wirelessly connected to the edge-cloud fusion analysis module and the dynamic digital twin modeling module, the dynamic digital twin modeling module is wirelessly connected to the edge-cloud fusion analysis module, the instantaneous fault topology positioning module and the self-healing control module, the edge-cloud fusion analysis module is wirelessly connected to the instantaneous fault topology positioning module and the self-healing control module, the instantaneous fault topology positioning module is wirelessly connected to the self-healing control module, the self-healing control module is wirelessly connected to the intelligent maintenance scheduling module, and the intelligent maintenance scheduling module is wirelessly connected to the multi-source data acquisition and preprocessing module.

[0064] In this embodiment, the dynamic digital twin modeling module, the topology structure initialization step includes: the spatiotemporal fusion feature matrix output by the multi-source data acquisition preprocessing module, including electrical parameters, environmental data, spatial displacement and timestamp;

[0065] An initial topology graph is constructed based on BIM data, where nodes represent devices and edges represent electrical connection relationships. The geometric properties of the topology graph are calibrated by combining the spatial displacement in the spatiotemporal fusion feature matrix, and a graph convolutional network model is constructed. The spatiotemporal fusion feature matrix is ​​input and the device status prediction is output.

[0066] In this embodiment, the dynamic digital twin modeling module builds a dynamic digital twin model of the distribution network topology, including the following steps: simulating the voltage distribution and power flow physical characteristics of the distribution network through electromagnetic transient simulation, using the electrical parameters in the time-space fusion feature matrix as simulation input, and dynamically adjusting the model parameters;

[0067] The spatiotemporal fusion feature matrix is ​​input into the graph convolutional network model to learn the topological dependencies between devices. The node weights are dynamically assigned through the attention mechanism to capture the impact of key devices.

[0068] Perform weighted fusion of physical simulation results and the output of the graph convolutional network model to generate a hybrid digital twin model;

[0069] When a distributed power source is detected to be connected to the grid, the topology is dynamically adjusted and the adjacency matrix is ​​updated through a dynamic graph neural network;

[0070] Update node status according to real-time electrical parameters, dynamically adjust model parameters based on equipment health status, calculate power distribution of distribution network through forward-backward substitution method based on real-time load data and topology structure, calculate the impact of key parameters on voltage stability, and dynamically adjust control strategy.

[0071] In this embodiment, the edge-cloud fusion analysis module extracts fault features by performing feature extraction on the spatiotemporal fusion feature matrix on the edge side and performing weighted fusion to form a fault feature vector;

[0072] Specifically, the implementation formula is:

[0073]

[0074] In the formula, F represents the fault feature vector, α i represents the dynamic weight of the i-th category feature, f i represents the characteristics of the i-th type of source data, β represents the global characteristics of the spatiotemporal correlation characteristics, and g(x,y) represents the spatiotemporal correlation function, which maps the spatial coordinate x and the time series y into high-level features.

[0075] In this embodiment, the edge-cloud fusion analysis module implements the dynamic threshold anomaly scoring steps: calculating the global anomaly score based on the fault feature vector F;

[0076] Specifically, the implementation formula is:

[0077]

[0078] In the formula, S represents the global anomaly score, γ represents the dynamic balance coefficient, and μ nl ,σ nl represents the characteristic mean and standard deviation under normal conditions, γ represents the dynamic balance coefficient, d(F,C cy ) represents the minimum Euclidean distance from the current fault feature vector F to the abnormal cluster, d max represents the normalization factor.

[0079] In this embodiment, the edge-cloud fusion analysis module generates a fault type label based on the global anomaly score, and the implementation formula is:

[0080]

[0081] In the formula, R represents the fault type label, L represents the predefined fault type set, and δ l represents the confidence weight of the lth type of fault, represents the reference anomaly score of the lth type fault, θ l Represents the tolerance threshold of the lth type; the generated fault type label is injected into the dynamic digital twin model to update the node status.

[0082] In this embodiment, the instantaneous fault topology location model and the spatial location probability of the fault section are implemented by calculating the spatial location probability of the fault section according to the topological structure and fault type label of the dynamic digital twin modeling module. The implementation formula is:

[0083]

[0084] In the formula, P(x) represents the probability density of the fault at the spatial coordinate x, W i Represents the topological weight of node and edge i, I(R i ) represents the influence function of the fault type label R on the node and edge i, d(x,x i ) represents the geometric distance between the spatial coordinate x and the nodes and edges, λ represents the spatial attenuation coefficient, and N is the total number of nodes and edges in the dynamic topology.

[0085] In this embodiment, the self-healing control module receives the fault type label and fault section coordinates output by the edge-cloud fusion analysis module, maps the fault coordinates to the topological structure in the dynamic digital twin model, determines the fault impact range, generates isolation instructions and recovery instructions based on the fault type and topological relationship, plans the shortest recovery path through a graph theory algorithm, generates a switch action sequence, and encapsulates the switch action instruction sequence as a blockchain transaction, including:

[0086] Command content: switch ID, action type, and expected execution time;

[0087] Signature information: digital signature of the control center;

[0088] Timestamp: ensure the timeliness of instructions;

[0089] Hash value: the hash value of the transaction content;

[0090] The verified instruction transactions are synchronously stored through blockchain nodes to ensure that the data cannot be tampered with, and the verified instructions are sent to the target switch device through the MQTT protocol.

[0091] In this embodiment, the intelligent maintenance scheduling module obtains the equipment health status output by the self-healing control module, reads historical fault data from the database or log system, and sorts the real-time data and historical data by timestamp to ensure time consistency;

[0092] Based on real-time and historical data, the system extracts equipment health indicators, uses a neural network model, and combines historical fault data to train a health prediction model. This outputs the current health score of the equipment and prioritizes tasks.

[0093] Task allocation is optimized through genetic algorithms to ensure maximum resource utilization, and a balance is struck between minimizing power outage time and maximizing resource utilization. The particle swarm optimization algorithm is used to generate the optimal scheduling plan, and the execution results are fed back to the multi-source data acquisition preprocessing module and the dynamic digital twin modeling module for optimization.

[0094] Working Principle: The system collects electrical parameters, environmental data, and spatial displacement of the distribution network in real time through IoT devices. It then uses outlier detection and denoising algorithms to eliminate noise and invalid data, cleans the data, and extracts key features from the cleaned data, including current mutations, temperature trends, and vibration spectra, to form a multidimensional feature vector. Principal component analysis is then used to reduce the dimensionality of high-dimensional features. After retaining the main information, the spatial features are combined with the temporal features to generate a spatiotemporal fusion feature matrix.

[0095] Through the spatiotemporal fusion feature matrix of multi-source data, a dynamic digital twin model of the distribution network is constructed, and the initial topology diagram is constructed through BIM (Building Information Model) data. Nodes represent devices, and edges represent electrical connection relationships. The geometric properties of the topology diagram are calibrated in combination with spatial displacement to ensure that the spatial position of the model is consistent with that of the physical equipment. Electromagnetic transient simulation is used to simulate physical characteristics such as voltage distribution and power flow. Electrical parameters are used as input to dynamically adjust model parameters. The graph convolutional network model is combined with the attention mechanism to dynamically assign node weights and capture the topological dependencies between devices. The physical simulation results are weightedly fused with the GCN output to generate a hybrid digital twin model. When a distributed power source is detected to be connected to the grid or a load change is detected, the model is dynamically adjusted. The graph neural network updates the adjacency matrix, adjusts the node status according to the real-time electrical parameters, calculates the power distribution based on the real-time load data and topology structure through the forward and back substitution method, evaluates the impact of key parameters, and dynamically adjusts the control strategy; the edge-cloud fusion analysis module extracts local features from the spatiotemporal fusion feature matrix, combines dynamic weight allocation, generates a fault feature vector, maps the spatial coordinates and time series into high-dimensional features through the spatiotemporal correlation function, calculates the deviation between the current data and the normal state based on the fault feature vector, and generates a global anomaly score based on the distance index of the anomaly cluster. According to the anomaly score and the predefined fault type set, the confidence weight of each type of fault is calculated. If the score exceeds the tolerance threshold, the corresponding fault label is generated. , and inject the dynamic digital twin model to update the node status; the instantaneous fault topology positioning module quantifies the fault possibility of each node and edge according to the topological weights of the nodes and edges in the dynamic digital twin model, combined with the influence function of the fault type, introduces the spatial attenuation coefficient, and calculates the relationship between the fault coordinates and the geometric distance of the equipment. The closer the distance, the higher the probability density; otherwise, it attenuates. The probability density distribution of the fault at the spatial coordinate x is generated through the total number of nodes and edges and the spatial attenuation coefficient, and the most likely fault section is determined. The self-healing control module maps the fault coordinates to the topological structure of the dynamic digital twin model, determines the fault impact range, and plans the shortest recovery path based on the fault type and topological relationship through the graph theory algorithm to generate the switch action sequence. , the verified instructions are sent to the target switch device through the MQTT protocol to perform isolation and power restoration operations; the intelligent maintenance scheduling module obtains the equipment health status from the self-healing control module, reads historical fault data from the database, and trains the health prediction model through the neural network model combined with historical data to output the current health score of the equipment. According to the health score, fault probability and power outage risk, a multi-level feedback queue strategy is used to divide the task priority, and the genetic algorithm is used to optimize task allocation, balance resource utilization and minimize power outage time. The particle swarm optimization algorithm is used to generate the optimal scheduling plan, and the execution results are fed back to the multi-source data acquisition preprocessing module and the dynamic digital twin modeling module to continuously optimize the model parameters and scheduling strategy.

[0096] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

[0097] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. The intelligent distribution network fault analysis and maintenance system based on the Internet of Things is characterized by: It includes a multi-source data acquisition and preprocessing module, a dynamic digital twin modeling module, an edge-cloud fusion analysis module, a transient fault topology location module, a self-healing control module, and an intelligent maintenance scheduling module; The dynamic digital twin modeling module constructs a dynamic digital twin model of the distribution network topology based on the spatiotemporal fusion feature matrix, and updates the node status and network parameters in real time; The edge-cloud fusion analysis module is used to extract fault features through edge-side AI reasoning, and combines cloud-side big data analysis to perform global anomaly detection on the dynamic digital twin model to generate fault type labels; The instantaneous fault topology location module is used to combine the fault type label and the topological structure data of the dynamic digital twin modeling module to calculate the spatial position probability distribution of the fault section and output the fault location coordinates.

2. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 1, characterized in that: The dynamic digital twin modeling module and the topology structure initialization step include: the spatiotemporal fusion feature matrix output by the multi-source data acquisition preprocessing module, including electrical parameters, environmental data, spatial displacement and timestamp, is used to construct an initial topology diagram based on BIM data. Nodes represent equipment, and edges represent electrical connection relationships. The geometric properties of the topology diagram are calibrated in combination with the spatial displacement in the spatiotemporal fusion feature matrix.

3. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 2, characterized in that: The dynamic digital twin modeling module builds a dynamic digital twin model of the distribution network topology, including the following steps: simulating the voltage distribution and power flow physical characteristics of the distribution network through electromagnetic transient simulation, using the electrical parameters in the spatiotemporal fusion feature matrix as simulation input, dynamically adjusting the model parameters, inputting the spatiotemporal fusion feature matrix into the graph convolutional network model, learning the topological dependencies between devices, dynamically allocating node weights through the attention mechanism, capturing the impact of key devices, weightedly fusing the physical simulation results with the output of the graph convolutional network model, generating a hybrid digital twin model, dynamically adjusting the topology when a distributed power source is detected to be connected to the grid, updating the adjacency matrix through a dynamic graph neural network, updating the node status according to the real-time electrical parameters, dynamically adjusting the model parameters in combination with the health status of the equipment, calculating the power distribution of the distribution network based on real-time load data and topology through the forward-backward substitution method, calculating the impact of key parameters on voltage stability, and dynamically adjusting the control strategy.

4. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 1, characterized in that: The edge-cloud fusion analysis module extracts fault features by performing feature extraction on the spatiotemporal fusion feature matrix on the edge side and performing weighted fusion to form a fault feature vector. The implementation formula is: In the formula, F represents the fault feature vector, α i represents the dynamic weight of the i-th category feature, f i represents the characteristics of the i-th type of source data, β represents the global characteristics of the spatiotemporal correlation characteristics, and g(x,y) represents the spatiotemporal correlation function, which maps the spatial coordinate x and the time series y into high-level features.

5. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 4 is characterized in that: The edge-cloud fusion analysis module implements the dynamic threshold anomaly scoring step: calculate the global anomaly score based on the fault feature vector F, and the implementation formula is: In the formula, S represents the global anomaly score, γ represents the dynamic balance coefficient, and μ nl , σ nl represents the characteristic mean and standard deviation under normal conditions, γ represents the dynamic balance coefficient, d(F,C cy ) represents the minimum Euclidean distance from the current fault feature vector F to the abnormal cluster, d max represents the normalization factor.

6. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 5, characterized in that: The edge-cloud fusion analysis module generates fault type labels based on the global anomaly score, and the implementation formula is: In the formula, R represents the fault type label, L represents the predefined fault type set, and δ l represents the confidence weight of the lth type of fault, represents the reference anomaly score of the lth type fault, θ l Represents the tolerance threshold of the lth type; the generated fault type label is injected into the dynamic digital twin model to update the node status.

7. The IoT-based intelligent distribution network fault analysis and maintenance system according to claim 1, characterized in that: The instantaneous fault topology location model and the spatial location probability of the fault section are implemented in the following steps: the spatial location probability of the fault section is calculated according to the topological structure and fault type label of the dynamic digital twin modeling module, and the implementation formula is: In the formula, P(x) represents the probability density of the fault at the spatial coordinate x, W i Represents the topological weight of node and edge i, I(R i ) represents the influence function of the fault type label R on the node and edge i, d(x,x i ) represents the geometric distance between the spatial coordinate x and the nodes and edges, λ represents the spatial attenuation coefficient, and N is the total number of nodes and edges in the dynamic topology.

Images