An intelligent fault diagnosis method and system for power distribution terminal equipment based on the Internet of Things

Abstract

The application discloses a power distribution terminal equipment fault intelligent diagnosis method and system based on an Internet of Things, which comprises the following steps: S1, collecting multiple operation data of a power distribution terminal to construct a time sequence sample; S2, extracting power disturbance features based on a sliding window, generating a behavior coupling matrix and a topological connection matrix, and calculating node redundancy; S3, fusing two types of relations to construct a dynamic atlas; S4, inputting a graph diffusion network, fusing structure and behavior diffusion results, and generating a node state feature vector; S5, inputting a multi-label classification model to identify fault types and probabilities; S6, combining fault types and redundancy to generate a response instruction; S7, executing the response and updating a relation matrix and a classification model, and closing loop optimization of a diagnosis process. The application realizes accurate identification and rapid response of power distribution terminal faults, and improves intelligent operation and maintenance level of a power supply system.

Description

Technical Field

[0001] This invention relates to the field of power distribution automation technology, and in particular to an intelligent fault diagnosis method and system for power distribution terminal equipment based on the Internet of Things. Background Technology

[0002] With the increasing intelligence of power distribution networks, distribution terminal equipment plays an increasingly crucial role in the system. These terminals are mainly used to collect electrical parameters, sense operating status, and execute control commands, serving as a vital bridge between distribution areas and the upper-level system. In recent years, distribution terminals have generally acquired multi-dimensional data acquisition capabilities, including voltage, current, and power, and connect to IoT platforms via wireless or wired communication, enabling basic information reporting and remote management. However, in actual operation, relying solely on the terminal itself or single-point analysis methods to identify faults is insufficient to meet the current system's requirements for diagnostic accuracy and response efficiency.

[0003] Traditional methods often rely on manually set thresholds, such as judging whether the voltage is too high or whether the power is fluctuating abnormally. While these methods may work in some stable scenarios, they are prone to false alarms or missed alarms in power distribution systems with widely distributed terminals and frequent state changes. Moreover, most existing methods only analyze data from a single terminal and lack an understanding of the relationships between multiple terminals, especially when there are ring network structures or backup paths, making it impossible to effectively assess the influence between nodes.

[0004] Another common problem is the disconnect between diagnosis and control. Existing systems typically only issue alarms after detecting anomalies, leaving manual confirmation and operational decisions for subsequent processing. This is inefficient and susceptible to human error. While some existing research has attempted to introduce graph neural networks or other algorithms to improve diagnostic capabilities, these methods often overlook the dynamic coupling between the physical wiring relationships and operational behaviors of power distribution terminals, resulting in limited improvement in diagnostic effectiveness and hindering practical application.

[0005] Especially when multiple types of faults occur simultaneously, traditional classification methods fail to adequately consider the dependencies between labels, easily leading to inaccurate identification results. Furthermore, most existing control strategies remain fixed, lacking mechanisms for adjusting strategies based on real-time information such as network topology and node redundancy. The execution level also lacks closed-loop feedback, impacting overall control effectiveness.

[0006] Based on the aforementioned problems, there is an urgent need for a novel diagnostic method that comprehensively considers the physical connections and operational status correlations between power distribution terminals. This method should not only improve identification accuracy but also achieve linkage between diagnostic results and control strategies, truly realizing a closed-loop process of "identification-response-optimization." Therefore, how to provide an intelligent fault diagnosis method and system for power distribution terminal equipment based on the Internet of Things is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose an intelligent fault diagnosis method and system for power distribution terminal equipment based on the Internet of Things. This invention integrates dynamic graph modeling, multi-scale graph diffusion reasoning, and multi-label classification, and describes in detail the joint modeling and intelligent response mechanism of the physical connection relationship and disturbance behavior coupling relationship between power distribution terminals. It has the advantages of high fault identification accuracy, strong response control closed-loop performance, and strong adaptability to dynamic topology structures.

[0008] A method for intelligent fault diagnosis of power distribution terminal equipment based on the Internet of Things according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data;

[0010] S2. Perform sliding window processing on the running data, extract the node power change trend, calculate the power disturbance synchronization degree between any pair of nodes, generate a behavior coupling relationship matrix, generate a topology connection relationship matrix based on the terminal physical connection information, and calculate the number of redundant connection paths to obtain the redundancy index of each node.

[0011] S3. By fusing the topological connection matrix and the behavioral coupling matrix, a dynamic graph containing structural edges and behavioral edges is constructed.

[0012] S4. Input the dynamic graph into the dual-scale graph diffusion inference network, and perform local diffusion based on structural edges and global diffusion based on behavioral edges respectively. The diffusion path is fused through a gating mechanism to output the node state feature vector.

[0013] S5. Input the node state feature vector into a trainable multi-label classification model to identify the fault type and probability, and output the classification label.

[0014] S6. Based on the fault type and redundancy index, generate response instructions that include control actions and scheduling objectives.

[0015] S7. Execute response commands, record control results, update the parameters of the behavior coupling matrix and the multi-label classification model, and complete the diagnostic closed-loop optimization.

[0016] Optionally, S2 specifically includes:

[0017] S21. Active power time series for each node of the distribution terminal. Perform sliding window processing, and set the window step size as... Calculate nodes within each time window power change rate ,in, Indicates the first in the window Each sampling time;

[0018] S22, For any two distribution terminals, the nodes and Extract the power change rate sequence within the same time window. and ,in, Indicates sampling At each moment, calculate the power disturbance synchronization degree. Define the behavioral coupling matrix The Item for ;

[0019] S23. Based on the physical wiring relationships between nodes, extract the direct connection states between nodes and construct a topology connection matrix. Topological connection matrix Let be a symmetric adjacency matrix, representing the structural connectivity elements of a power distribution network, defined as follows:

[0020] ;

[0021] S24, Based on the topological connection matrix A depth-first search algorithm is used to enumerate each node. All reachable paths, count the number of reachable redundant paths. And calculate the node redundancy index. ,in, This represents the maximum number of redundant paths among all nodes in the current power distribution network. Used for metric nodes The relative connectivity redundancy level indicates that a node has more alternative paths to carry out failover tasks.

[0022] Optionally, in step S24, a depth-first search algorithm is used to enumerate each node. The process of identifying all reachable paths includes: starting with each distribution terminal as the starting node, constructing a connection graph based on the topology connection matrix, traversing all paths connected to the node using a depth-first search algorithm, limiting the path length to no more than the maximum allowed number of hops set in the system during the traversal, skipping any paths containing duplicate nodes or forming closed loops, and counting each path that satisfies the conditions of no repetition, no loops, and reachability to other nodes within a limited length as a valid path. The total number of valid paths that can be enumerated starting from the distribution terminal is the number of reachable paths for the node.

[0023] Optionally, the process of constructing the dynamic graph in S3 includes: establishing structural edges between power distribution terminals based on the topological connection matrix as the connection basis of structural edges; establishing behavioral edges reflecting the synchronization relationship of power disturbances between nodes based on the behavioral coupling matrix as the weight reference; and after labeling the structural edges and behavioral edges with attribute categories, integrating them into a heterogeneous graph structure. In the heterogeneous graph, each node represents a power distribution terminal, and the edge types are distinguished as physical connection edges and behavioral coupling edges. Among them, structural edges indicate that there is an actual wiring connection between nodes, and behavioral edges indicate that the power disturbances between nodes are synchronous.

[0024] Optionally, S4 specifically includes:

[0025] S41. Using the structural edges corresponding to the topological connection matrix as the local diffusion path input channel, perform local graph state propagation based on the set of adjacent nodes of each node to generate structural edge diffusion output.

[0026] S42. Using the behavioral edges corresponding to the behavioral coupling relationship matrix as the input channel for the global diffusion path, the behavioral information is propagated between non-adjacent nodes based on the weights of the synchronization between nodes, generating the behavioral edge diffusion output.

[0027] S43. Generate node state vectors for the structural edge diffusion output and the behavioral edge diffusion output, respectively, and denote them as structural state vectors. With behavioral state vector , the structure state vector With behavioral state vector In the input gating mechanism, the structure state vector is calculated. With behavioral state vector The correlation distribution, self-information, and mutual information feature density in the vector space are adaptively generated by the gating mechanism to form a fusion weight coefficient vector, which dynamically controls the fusion ratio of the structural edge diffusion output and the behavioral edge diffusion output.

[0028] S44. Using the fusion weight coefficient vector output by the gating mechanism, the diffusion results of structural edges and behavioral edges are weighted and fused to generate the fused node state feature vector. .

[0029] Optionally, S5 specifically includes: the process of inputting the node state feature vector into the multi-label classification model to identify the fault type in S5 includes: inputting the node state feature vector generated by the gating mechanism into the multi-label classification model to identify the fault type. The input is fed into a trainable multi-label classification model, which includes an input layer, a feature compression layer, a label association modeling layer, and an output layer. In the feature compression layer, the node state feature vectors are processed... Dimensional compression is performed to obtain compressed feature vectors. In the label association modeling layer, a label co-occurrence matrix based on historical sample statistics is constructed. Multi-label feature mapping is then performed based on the label co-occurrence matrix to generate multi-label embedding vectors. In the output layer, multi-label embedding vectors Execute the activation operation and output the fault type and probability score corresponding to each node. The fault types include power failure, overvoltage abnormality, undervoltage abnormality, current imbalance, reactive power abnormality, communication abnormality, and status code abnormality.

[0030] Optionally, S6 specifically includes: retrieving the corresponding control action based on the fault type output by the multi-label classification model, adjusting the control action in conjunction with the redundancy index of the node, determining the scheduling target parameters based on the fault type and the node's position in the topology, and generating a response instruction containing the control action and the scheduling target.

[0031] An intelligent fault diagnosis system for power distribution terminal equipment based on the Internet of Things according to an embodiment of the present invention includes the following modules:

[0032] The operation data acquisition module is used to collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data;

[0033] The relation matrix construction module is used to generate behavioral coupling relation matrices and topological connection relation matrices based on runtime data.

[0034] The redundancy calculation module is used to calculate the number of redundant paths for each node based on the topology connection matrix.

[0035] The dynamic graph construction module is used to fuse structural edges and behavioral edges to generate a heterogeneous graph structure representing the state of the power distribution terminal.

[0036] The graph diffusion fusion module is used to perform local diffusion of structural edges and global diffusion of behavioral edges, and outputs the fusion state vector through a gating mechanism.

[0037] The multi-label classification module is used to input the fused state vector into the multi-label classification model to identify the fault type and probability score;

[0038] The response strategy generation module is used to generate control actions and scheduling target parameters based on fault type and redundancy index;

[0039] The closed-loop update module is used to execute response commands and provide feedback on control results, updating the parameters of the behavioral coupling relationship and the multi-label classification model.

[0040] The beneficial effects of this invention are:

[0041] (1) This invention constructs a dynamic graph model that integrates topological connection relationship and power disturbance synchronization relationship, and combines a dual-scale graph diffusion mechanism and a gating fusion strategy to extract structural and behavioral correlation features between power distribution terminals, thereby improving the fault location accuracy under complex topological structures, and has good identification robustness, especially in concurrent multi-fault scenarios.

[0042] (2) This invention introduces a multi-label classification model to identify specific fault types, combines the redundancy index of power distribution terminals, generates structured response instructions from the preset strategy library, clarifies the control action type and scheduling target, so that the diagnostic results can directly drive the control behavior, and constructs a closed-loop mechanism for the whole process from identification, response to model update, which solves the problem of "disconnection between identification and response" in traditional solutions.

[0043] (3) This invention supports dynamic adjustment of graph structure and classification model parameters during operation. By executing feedback to update the coupling relationship of behavior and classification model weights, online retraining and self-optimization can be achieved. The system has the ability to quickly adapt to changes in wiring methods and fluctuations in electrical state, which can help build an adaptive intelligent power distribution terminal management solution for the Internet of Things environment. Attached Figure Description

[0044] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0045] Figure 1 This is an overall flowchart of an intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things proposed in this invention;

[0046] Figure 2 This is a module connection diagram of an intelligent fault diagnosis system for power distribution terminal equipment based on the Internet of Things proposed in this invention. Detailed Implementation

[0047] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0048] refer to Figure 1 and Figure 2 A method for intelligent fault diagnosis of power distribution terminal equipment based on the Internet of Things includes the following steps:

[0049] S1. Collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data;

[0050] S2. Perform sliding window processing on the running data, extract the node power change trend, calculate the power disturbance synchronization degree between any pair of nodes, generate a behavior coupling relationship matrix, generate a topology connection relationship matrix based on the terminal physical connection information, and calculate the number of redundant connection paths to obtain the redundancy index of each node.

[0051] S3. By fusing the topological connection matrix and the behavioral coupling matrix, a dynamic graph containing structural edges and behavioral edges is constructed.

[0052] S4. Input the dynamic graph into the dual-scale graph diffusion inference network, and perform local diffusion based on structural edges and global diffusion based on behavioral edges respectively. The diffusion path is fused through a gating mechanism to output the node state feature vector.

[0053] S5. Input the node state feature vector into a trainable multi-label classification model to identify the fault type and probability, and output the classification label.

[0054] S6. Based on the fault type and redundancy index, generate response instructions that include control actions and scheduling objectives.

[0055] S7. Execute response commands, record control results, update the parameters of the behavior coupling matrix and the multi-label classification model, and complete the diagnostic closed-loop optimization.

[0056] This invention achieves end-to-end automated fault diagnosis of power distribution terminal equipment by constructing a complete process from power distribution terminal operation data acquisition, coupling relationship identification, graph modeling, graph diffusion fusion, fault identification to response strategy feedback. Compared with existing technical approaches that rely on manual threshold setting and single-point analysis, this invention significantly improves the accuracy and system adaptability of diagnosis through dynamic data-driven and structural behavior fusion, making it particularly suitable for actual operating environments with complex topologies and frequent power disturbances, and exhibiting higher practicality and robustness.

[0057] In this embodiment, S2 specifically includes:

[0058] S21. Active power time series for each node of the distribution terminal. Perform sliding window processing, and set the window step size as... Calculate nodes within each time window power change rate :

[0059] ;

[0060] in, Indicates the first in the window At each sampling time, a total of samples were collected. At that moment;

[0061] S22, For any two distribution terminals, the nodes and Extract the power change rate sequence within the same time window. and Calculate power disturbance synchronization Define the behavioral coupling matrix The Item for :

[0062] ;

[0063] in, and Representing nodes respectively With nodes The average value of the power change rate sequence within the current time window, and the behavioral coupling matrix. Characterizing the consistency of perturbation trends among different nodes, when A value close to 1 indicates a highly synchronized power disturbance trend between the two nodes;

[0064] This invention introduces a formula for calculating the inter-node disturbance synchronization degree based on the power disturbance change rate, which can characterize the dynamic behavioral similarity between any pair of distribution terminal nodes within a sliding time window. This formula constructs a behavioral coupling relationship matrix with more physical behavioral significance by modeling the correlation of active power change trends. Compared with traditional correlation determination methods based on static values ​​or fixed thresholds, this formula has stronger dynamic adaptability and anomaly sensitivity. Especially in scenarios such as fault propagation, fluctuation interference, and disturbance coordination, it can more accurately capture potential influence chains between nodes, providing high-quality edge weight support for subsequent dynamic graph modeling and graph diffusion inference, significantly improving the accuracy of fault identification and causal analysis capabilities.

[0065] S23. Based on the physical wiring relationships between nodes, extract the direct connection states between nodes and construct a topology connection matrix. Topological connection matrix Let be a symmetric adjacency matrix, representing the structural connectivity elements of a power distribution network, defined as follows:

[0066] ;

[0067] S24, Based on the topological connection matrix A depth-first search algorithm is used to enumerate each node. All reachable paths, count the number of reachable redundant paths. And calculate the node redundancy index. ,in, This represents the maximum number of redundant paths among all nodes in the current power distribution network. Used for metric nodes The relative connectivity redundancy level indicates that a node has more alternative paths to carry out failover tasks.

[0068] This invention introduces a mechanism for calculating disturbance synchronization and constructing a relationship matrix, enabling quantitative modeling of the operational behavior between terminals. Compared to traditional methods that rely solely on physical connections or static numerical judgments, this invention dynamically identifies the degree of behavioral correlation between distribution terminals based on power change trends. This reflects the characteristics of multi-terminal coordinated response or interference propagation, demonstrating significant advantages in identifying potential fault propagation paths and constructing a more realistic terminal relationship network.

[0069] In this embodiment, step S24 uses a depth-first search algorithm to enumerate each node. The process of identifying all reachable paths includes: starting with each distribution terminal as the starting node, constructing a connection graph based on the topology connection matrix, traversing all paths connected to the node using a depth-first search algorithm, limiting the path length to no more than the maximum allowed number of hops set in the system during the traversal, skipping any paths containing duplicate nodes or forming closed loops, and counting each path that satisfies the conditions of no repetition, no loops, and reachability to other nodes within a limited length as a valid path. The total number of valid paths that can be enumerated starting from the distribution terminal is the number of reachable paths for the node.

[0070] This invention proposes a depth-first search method based on the topology connectivity matrix to accurately calculate the number of redundant paths at distribution terminal nodes, thereby obtaining a node redundancy index. This index can be used to assist in generating more feasible and secure response strategies, solving the problems of existing response mechanisms that ignore topology redundancy and lack control and fault tolerance capabilities, and providing decision support for strategy optimization and high-reliability power supply.

[0071] In this embodiment, the process of constructing the dynamic graph in S3 includes: establishing structural edges between power distribution terminals based on the topological connection matrix as the connection basis of structural edges; establishing behavioral edges reflecting the power disturbance synchronization relationship between nodes based on the behavioral coupling matrix as the weight reference; and after labeling the structural edges and behavioral edges with attribute categories, integrating them into a heterogeneous graph structure. In the heterogeneous graph, each node represents a power distribution terminal, and the edge types are distinguished as physical connection edges and behavioral coupling edges. Structural edges indicate that there is an actual wiring connection between nodes, and behavioral edges indicate that the power disturbance between nodes is synchronous.

[0072] This invention constructs a heterogeneous dynamic graph model by fusing structural edges and behavioral edges, effectively integrating the physical connection information and power disturbance coupling relationship between power distribution terminals. Compared with methods that only use static wiring diagrams or single feature graphs, the graph modeling method of this invention can express multi-type and multi-granularity node interaction relationships, has dynamic update capabilities and stronger structural expression capabilities, and provides a high-quality input foundation for graph diffusion inference.

[0073] In this embodiment, S4 specifically includes:

[0074] S41. Using the structural edges corresponding to the topological connection matrix as the local diffusion path input channel, perform local graph state propagation based on the set of adjacent nodes of each node to generate structural edge diffusion output.

[0075] S42. Using the behavioral edges corresponding to the behavioral coupling relationship matrix as the input channel for the global diffusion path, the behavioral information is propagated between non-adjacent nodes based on the weights of the synchronization between nodes, generating the behavioral edge diffusion output.

[0076] S43. Generate node state vectors for the structural edge diffusion output and the behavioral edge diffusion output, respectively, and denote them as structural state vectors. With behavioral state vector , the structure state vector With behavioral state vector In the input gating mechanism, the structure state vector is calculated. With behavioral state vector The correlation distribution, self-information, and mutual information feature density in the vector space are adaptively generated into a fusion weight coefficient vector by a gating mechanism. This dynamically controls the fusion ratio of structural edge diffusion output and behavioral edge diffusion output. The specific process includes: firstly, calculating the structural state vector. With behavioral state vector The correlation distribution in the vector space is used to measure directional consistency. Next, the self-information of each vector is calculated to measure the density of its independent features. Then, the mutual information feature density is calculated to evaluate the degree of feature overlap. Finally, the correlation distribution, self-information, and mutual information feature density are used as input features of the gating function. A trainable gating network is used to generate a fusion weight coefficient vector for the structural state vector. With behavioral state vector By performing element-wise weighted fusion, dynamic proportional control of the output of structural and behavioral diffusion paths can be achieved.

[0077] S44. Using the fusion weight coefficient vector output by the gating mechanism, the diffusion results of structural edges and behavioral edges are weighted and fused to generate the fused node state feature vector. .

[0078] This invention proposes a dual-scale graph diffusion inference network with a gating fusion mechanism. State information is propagated in both structural and behavioral edge paths, and the gating mechanism adaptively calculates the fusion weights, enabling dynamic control of information fusion and feature extraction. Compared to traditional graph neural networks with fixed diffusion paths or average aggregation methods, this method can flexibly adjust the propagation bias in different scenarios, effectively improving the graph model's ability to express complex power disturbance features.

[0079] In this embodiment, S5 specifically includes: the process of inputting the node state feature vector into the multi-label classification model to identify the fault type in S5 includes: inputting the node state feature vector generated by the gating mechanism into the multi-label classification model to identify the fault type. The input is fed into a trainable multi-label classification model, which includes an input layer, a feature compression layer, a label association modeling layer, and an output layer. In the feature compression layer, the node state feature vectors are processed... Dimensional compression is performed to obtain compressed feature vectors. In the label association modeling layer, a label co-occurrence matrix based on historical sample statistics is constructed. Multi-label feature mapping is then performed based on the label co-occurrence matrix to generate multi-label embedding vectors. In the output layer, multi-label embedding vectors Execute the activation operation and output the fault type and probability score corresponding to each node. The fault types include power failure, overvoltage abnormality, undervoltage abnormality, current imbalance, reactive power abnormality, communication abnormality, and status code abnormality.

[0080] This invention designs a trainable multi-label classification model that takes the fused node state vector as input and identifies multiple fault labels, including power outage, overvoltage, undervoltage, current imbalance, reactive power anomaly, communication anomaly, and status code anomaly. It supports probability output and label correlation modeling. This model overcomes the limitations of traditional classification methods in handling label co-occurrence or ignoring multi-class combinations, improving the accuracy and generalization ability of fault identification and providing reliable support for intelligent diagnosis in complex scenarios.

[0081] In this embodiment, S6 specifically includes: retrieving the corresponding control action based on the fault type output by the multi-label classification model, adjusting the control action in conjunction with the redundancy index of the node, determining the scheduling target parameters based on the fault type and the node's position in the topology, and generating a response instruction containing the control action and the scheduling target. The control action includes instructions such as cutting off power supply, reducing voltage operation, switching to a backup channel, remotely restarting the communication module, and issuing a manual inspection request. The scheduling target parameters are used to limit the operating target and execution conditions that the control action should achieve, including key parameters such as the target voltage value, current threshold, action duration, target switching node number, and backup path activation flag.

[0082] This invention, based on fault type tags and node redundancy indicators, matches and adjusts response instructions from a strategy library to generate instructions containing control actions and scheduling objectives, achieving an automatic transition from diagnostic results to execution control. Compared to traditional methods relying on manual judgment and fixed control instructions, this invention achieves closed-loop linkage between diagnosis and response, improving response efficiency and strategy adaptability, and enhancing the intelligence level of terminal-level automatic control.

[0083] An intelligent fault diagnosis system for power distribution terminal equipment based on the Internet of Things according to an embodiment of the present invention includes the following modules:

[0084] The operation data acquisition module is used to collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data;

[0085] The relation matrix construction module is used to generate behavioral coupling relation matrices and topological connection relation matrices based on runtime data.

[0086] The redundancy calculation module is used to calculate the number of redundant paths for each node based on the topology connection matrix.

[0087] The dynamic graph construction module is used to fuse structural edges and behavioral edges to generate a heterogeneous graph structure representing the state of the power distribution terminal.

[0088] The graph diffusion fusion module is used to perform local diffusion of structural edges and global diffusion of behavioral edges, and outputs the fusion state vector through a gating mechanism.

[0089] The multi-label classification module is used to input the fused state vector into the multi-label classification model to identify the fault type and probability score;

[0090] The response strategy generation module is used to generate control actions and scheduling target parameters based on fault type and redundancy index;

[0091] The closed-loop update module is used to execute response commands and provide feedback on control results, updating the parameters of the behavioral coupling relationship and the multi-label classification model.

[0092] This invention employs a modular design, dividing the entire diagnostic system into multiple functional modules, including data acquisition, modeling, diffusion, identification, strategy, and feedback. Each module has clearly defined functions and data interfaces, facilitating system integration and iterative upgrades. Compared to traditional centralized processing structures, this invention's system architecture offers greater scalability and engineering deployment value, making it suitable for intelligent diagnostic deployment scenarios involving large-scale power distribution terminal equipment.

[0093] Example 1:

[0094] To verify the feasibility of this invention in practice, it was applied to a high-voltage distribution area. In this area, the power distribution system covers over a hundred distribution terminals, supporting residential, commercial, and small industrial loads. This distribution area suffers from a complex topology, diverse equipment types, and frequent electrical disturbances. Traditional maintenance methods rely primarily on manual inspections and simple threshold alarms, which not only result in low diagnostic efficiency and frequent false alarms and missed alarms, but also fail to address the chain reaction of fault propagation caused by disturbances between terminals, severely restricting the stability and recovery speed of the power distribution system.

[0095] The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things proposed in this invention was deployed in the data center of this distribution area. The data acquisition module collects information such as voltage, current, active / reactive power, frequency, and status codes of all terminals at a 1-minute cycle, and uploads it after preliminary preprocessing by edge computing nodes. Subsequently, the system performs sliding window analysis on the operational data, calculates the power disturbance change rate and synchronization degree between terminals, generates a behavioral coupling relationship matrix, and automatically extracts physical connection relationships based on the wiring diagram to construct a topology connection relationship matrix and counts the number of redundant paths for each terminal.

[0096] The graph modeling module integrates the aforementioned structural and behavioral relationships to construct a dynamic graph structure. This structure is then input into a dual-scale graph diffusion inference network, performing local diffusion based on structural edges and global diffusion based on behavioral edges. A gating mechanism dynamically generates fusion weights based on the correlation, self-information, and mutual information density of each node's structural and behavioral state vectors, achieving differential dynamic control and thus obtaining highly expressive node state vectors. The multi-label classification module identifies fault types in the state vectors, including power outage faults, undervoltage anomalies, and reactive power anomalies, and outputs the probability of each fault type. Finally, the policy module combines fault type and node redundancy to generate response instructions, achieving precise control and resource scheduling, and recording the response results for model parameter self-updating.

[0097] Six typical terminals were selected from the actual operating data after system deployment for comparative analysis. As shown in the table, these terminals exhibit varying degrees of voltage fluctuation and power disturbance characteristics, with disturbance synchronization ranging from 0.33 to 0.91 and structural redundancy between 1 and 3. Before the deployment of the model of this invention, the average diagnostic accuracy of the system was 74.6%, and the response time was approximately 16 seconds. After the deployment of this invention, the overall diagnostic accuracy improved to 87.9%, the false alarm rate was significantly reduced, and the average response control time was shortened to less than 10 seconds. In particular, the improvement effect was more significant for nodes with high disturbance synchronization and complex path structures (such as terminals 103 and 106), fully verifying the effectiveness and robustness of this invention in handling fault diagnosis of power terminals with high coupling complexity and rapid state changes.

[0098] Table 1: Verification Data Table for Intelligent Fault Diagnosis of Distribution Terminals

[0099]

[0100] As can be seen from the "Verification Data Table for Intelligent Fault Diagnosis of Distribution Terminals," this invention demonstrates significant advantages in practical applications of distribution terminals. The table selects six typical terminal nodes, comparing their performance indicators across multiple dimensions before and after deploying the fault diagnosis system, covering disturbance behavior, structural characteristics, diagnostic effectiveness, and response efficiency.

[0101] First, considering the frequency of voltage fluctuations and the amplitude of power disturbances, terminal 103 exhibits the highest power disturbance value of 1.01kW and a voltage fluctuation frequency of 7 times / day, indicating that its operating state fluctuates drastically and it is a high-disturbance node. In contrast, terminal 102 exhibits relatively stable disturbance characteristics, with a power disturbance amplitude of only 0.22kW and the lowest fluctuation frequency of only 1 time / day. These differences are accurately quantified into different disturbance synchronization indices in behavioral coupling calculations. For example, the synchronization indices of terminals 101 and 104 are as high as 0.91 and 0.87, respectively, indicating that they have strong cooperative response relationships with other nodes in the network and are susceptible to disturbances from neighboring nodes.

[0102] Secondly, in terms of the number of structurally redundant paths, terminals 103 and 106 both have three paths, making them nodes with high fault tolerance. Terminal 102, however, only has a single path access, resulting in lower structural redundancy and limited reconfigurability after a failure. In the response strategy generation of this invention, this redundancy index is used to assist in the selection of optimal scheduling paths, thereby improving system recovery capabilities.

[0103] From a diagnostic performance perspective, before initial deployment, the fault diagnosis accuracy of each terminal was generally between 65% and 82%, with terminal 103 having the lowest accuracy at only 64.3%. After deploying this invention, the accuracy significantly improved, with terminal 103 rising to 84.0%, an increase of nearly 20 percentage points. This verifies the superior characterization capability of the fused dynamic graph and dual-scale diffusion model under complex perturbations. Other nodes, such as 101 and 104, also saw accuracy improvements exceeding 9%, with the overall average accuracy improvement reaching over 13%.

[0104] Regarding response efficiency, this invention reduces the average response time from approximately 16 seconds to less than 10 seconds after optimization. Taking terminal 105 as an example, the response time was 16.4 seconds before optimization and reduced to 10.2 seconds after optimization; terminal 102 requires only 8.7 seconds after optimization, indicating that the system still has good decision-making efficiency under a low-redundancy structure.

[0105] In summary, the table accurately reflects the robustness and high performance of the invention under various node conditions. It not only improves the accuracy and precision of fault identification but also significantly enhances response speed, making it suitable for intelligent diagnosis and adaptive control scenarios in large-scale power distribution terminals.

[0106] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent fault diagnosis of power distribution terminal equipment based on the Internet of Things, characterized in that, Includes the following steps: S1. Collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data; S2. Perform sliding window processing on the running data, extract the node power change trend, calculate the power disturbance synchronization degree between any pair of nodes, generate a behavior coupling relationship matrix, generate a topology connection relationship matrix based on the terminal physical connection information, and calculate the number of redundant connection paths to obtain the redundancy index of each node. S3. By fusing the topological connection matrix and the behavioral coupling matrix, a dynamic graph containing structural edges and behavioral edges is constructed. S4. Input the dynamic graph into the dual-scale graph diffusion inference network, and perform local diffusion based on structural edges and global diffusion based on behavioral edges respectively. Then, fuse the node state feature vectors output by the diffusion path through a gating mechanism. S5. Input the node state feature vector into a trainable multi-label classification model to identify the fault type and probability, and output the classification label. S6. Based on the fault type and redundancy index, generate response instructions that include control actions and scheduling objectives. S7. Execute response commands, record control results, update the parameters of the behavior coupling matrix and the multi-label classification model, and complete the diagnostic closed-loop optimization.

2. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 1, characterized in that, S2 specifically includes: S21. Active power time series for each node of the distribution terminal. Perform sliding window processing, assuming the window step size is... Calculate nodes within each time window power change rate ,in, Indicates the first in the window Each sampling time; S22, For any two distribution terminals, the nodes and Extract the power change rate sequence within the same time window. and Calculate power disturbance synchronization Define the behavioral coupling matrix The Item for : ； in, and Representing nodes respectively With nodes The average value of the power change rate sequence within the current time window, and the behavioral coupling matrix. Characterizing the consistency of perturbation trends among different nodes, when A value close to 1 indicates a highly synchronized power disturbance trend between the two nodes; S23. Based on the physical wiring relationships between nodes, extract the direct connection states between nodes and construct a topology connection matrix. , Let be a symmetric adjacency matrix, representing the structural connectivity elements of a power distribution network, defined as follows: ； S24, Based on the topological connection matrix A depth-first search algorithm is used to enumerate each node. All reachable paths, count the number of reachable redundant paths. And calculate the node redundancy index. ,in, This represents the maximum number of redundant paths among all nodes in the current power distribution network. Used for metric nodes The relative connectivity redundancy level indicates that a node has more alternative paths to carry out failover tasks.

3. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 2, characterized in that, In step S24, a depth-first search algorithm is used to enumerate each node. The process of identifying all reachable paths includes: starting with each distribution terminal as the starting node, constructing a connection graph based on the topology connection matrix, traversing all paths connected to the node using a depth-first search algorithm, limiting the path length to no more than the maximum allowed number of hops set in the system during the traversal, skipping any paths containing duplicate nodes or forming closed loops, and counting each path that satisfies the conditions of no repetition, no loops, and reachability to other nodes within a limited length as a valid path. The total number of valid paths that can be enumerated starting from the distribution terminal is the number of reachable paths for the node.

4. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 3, characterized in that, The process of constructing the dynamic graph in S3 includes: establishing structural edges between power distribution terminals based on the topological connection matrix as the connection basis of structural edges; establishing behavioral edges reflecting the power disturbance synchronization relationship between nodes based on the behavioral coupling matrix as the weight reference; and after labeling the structural edges and behavioral edges with attribute categories, integrating them into a heterogeneous graph structure. In the heterogeneous graph, each node represents a power distribution terminal, and the edge types are distinguished as physical connection edges and behavioral coupling edges. Among them, structural edges indicate that there is an actual wiring connection between nodes, and behavioral edges indicate that the power disturbance between nodes is synchronous.

5. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 4, characterized in that, S4 specifically includes: S41. Using the structural edges corresponding to the topological connection matrix as the local diffusion path input channel, perform local graph state propagation based on the set of adjacent nodes of each node to generate structural edge diffusion output. S42. Using the behavioral edges corresponding to the behavioral coupling relationship matrix as the input channel for the global diffusion path, the behavioral information is propagated between non-adjacent nodes based on the weights of the synchronization between nodes, generating the behavioral edge diffusion output. S43. Generate node state vectors for the structural edge diffusion output and the behavioral edge diffusion output, respectively, and denote them as structural state vectors. With behavioral state vector ,Will and In the input gating mechanism, through calculation and The correlation distribution, self-information, and mutual information feature density in the vector space are adaptively generated by the gating mechanism to form a fusion weight coefficient vector, which dynamically controls the fusion ratio of the structural edge diffusion output and the behavioral edge diffusion output. S44. Using the fusion weight coefficient vector output by the gating mechanism, the diffusion results of structural edges and behavioral edges are weighted and fused to generate the fused node state feature vector. .

6. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 5, characterized in that, S5 specifically includes: the process of inputting the node state feature vector into the multi-label classification model to identify the fault type in S5 includes: inputting the node state feature vector generated by the gating mechanism into the multi-label classification model. The input is fed into a trainable multi-label classification model, which includes an input layer, a feature compression layer, a label association modeling layer, and an output layer. In the feature compression layer... Dimensional compression is performed to obtain compressed feature vectors. In the label association modeling layer, a label co-occurrence matrix based on historical sample statistics is constructed. Multi-label feature mapping is then performed based on the label co-occurrence matrix to generate multi-label embedding vectors. In the output layer Execute the activation operation and output the fault type and probability score corresponding to each node. The fault types include power failure, overvoltage abnormality, undervoltage abnormality, current imbalance, reactive power abnormality, communication abnormality, and status code abnormality.

7. The intelligent fault diagnosis method for power distribution terminal equipment based on the Internet of Things according to claim 6, characterized in that, S6 specifically includes: retrieving the corresponding control action based on the fault type output by the multi-label classification model, adjusting the control action in combination with the redundancy index of the node, determining the scheduling target parameters based on the fault type and the position of the node in the topology, and generating a response instruction containing the control action and the scheduling target.

8. An intelligent fault diagnosis system for power distribution terminal equipment based on the Internet of Things (IoT), applied to the intelligent fault diagnosis method for power distribution terminal equipment based on the IoT as described in any one of claims 1 to 7, characterized in that, Includes the following modules: The operation data acquisition module is used to collect voltage, current, active power, reactive power, frequency, status code and sampling timestamp of the power distribution terminal to construct operation data; The relation matrix construction module is used to generate behavioral coupling relation matrices and topological connection relation matrices based on runtime data. The redundancy calculation module is used to calculate the number of redundant paths for each node based on the topology connection matrix. The dynamic graph construction module is used to fuse structural edges and behavioral edges to generate a heterogeneous graph structure representing the state of the power distribution terminal. The graph diffusion fusion module is used to perform local diffusion of structural edges and global diffusion of behavioral edges, and outputs the fusion state vector through a gating mechanism. The multi-label classification module is used to input the fused state vector into the multi-label classification model to identify the fault type and probability score; The response strategy generation module is used to generate control actions and scheduling target parameters based on fault type and redundancy index; The closed-loop update module is used to execute response commands and provide feedback on control results, updating the parameters of the behavioral coupling relationship and the multi-label classification model.

Images