A power distribution box centralized management system for carrier ad hoc network communication

Abstract

The application discloses a power distribution box centralized management and control system for carrier self-organizing network communication and relates to the technical field of power distribution box networking. The system comprises a centralized management platform and a plurality of intelligent power distribution box terminals. The centralized management platform is arranged in a monitoring center and is internally provided with an operating state knowledge graph construction module, an abnormal event semantic analysis module and an intelligent topology optimization decision module. The operating state knowledge graph construction module is used for constructing and dynamically updating an attribute graph model. The abnormal event semantic analysis module is used for automatically triggering a semantic analysis process when an abnormal event is detected. The intelligent topology optimization decision module is used for deciding whether to trigger topology reconstruction. Each intelligent power distribution box terminal is internally integrated with a carrier self-organizing network communication module and an edge computing processing unit. The application has the advantages that wiring-free networking is realized by relying on power line carrier communication, a dynamic network topology construction mechanism is designed, communication interruption self-recovery is realized, and the centralized management and control level of the power distribution box is improved.

Description

Technical Field

[0001] This invention relates to the field of distribution box networking technology, specifically to a centralized management and control system for distribution boxes using carrier self-organizing network communication. Background Technology

[0002] Currently, centralized management of distribution boxes mainly relies on three methods: independent operation, RS485 bus networking, or wireless public network communication. Independent operation cannot achieve remote monitoring, resulting in extremely low maintenance efficiency. While RS485 bus networking can achieve centralized management, it suffers from inherent drawbacks such as complex wiring, poor scalability, and the potential for localized communication paralysis due to single-point failures. Wireless public networks, such as 4G / 5G, offer flexible deployment, but their reliability is insufficient in signal blind spots such as basements and tunnels, and their long-term operating costs are high. In recent years, power line carrier (PLC) communication technology has been introduced into the field of distribution box management due to its advantages of reusing existing power lines and eliminating the need for wiring. However, existing PLC communication solutions are mostly static networks or rely on central nodes for route allocation, making it difficult to adapt to the dynamic changes in the power line channel environment that cause fluctuations in communication quality, leading to rigid network topology and limited communication reliability.

[0003] Furthermore, the existing technology for diagnosing communication faults in distribution boxes still relies on preset threshold rules, which cannot understand the complex causal chain behind the anomaly. For example, it cannot distinguish between communication interruption caused by carrier signal interference and communication interruption caused by aging of internal components in the distribution box. At the same time, it only outputs alarm information and cannot provide reasoning process and evidence chain, requiring maintenance personnel to manually investigate the cause, which is inefficient. Summary of the Invention

[0004] To solve the above-mentioned technical problems, a centralized management and control system for distribution boxes with carrier self-organizing network communication is provided. This technical solution solves at least one of the technical problems mentioned in the background section.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A centralized management and control system for distribution boxes using carrier ad hoc network communication includes a centralized management platform and multiple intelligent distribution box terminals. The centralized management platform is deployed in the monitoring center to issue control commands, receive and display the operating status data of all distribution boxes. It has built-in modules for building an operating status knowledge graph, analyzing abnormal events semantically, and making intelligent topology optimization decisions. The operational status knowledge graph construction module is used to construct and dynamically update an attribute graph model containing distribution box nodes, sensor nodes, communication link nodes, fault event nodes, and environmental factor nodes and their relationships, based on the electrical parameters, communication link quality indicators, equipment operation logs, and environmental data reported by each distribution box terminal. The abnormal event semantic analysis module is connected to the running status knowledge graph construction module, and is used to automatically trigger the semantic analysis process when an abnormal event is detected, and output the root cause judgment, credibility score and reasoning evidence chain. The intelligent topology optimization decision module is connected to the abnormal event semantic analysis module. It is used to combine the root cause judgment results output by the abnormal event semantic analysis module with the original link quality indicators to perform multi-dimensional weighted scoring and determine whether to trigger topology reconstruction. Each smart distribution box terminal integrates a carrier self-organizing network communication module and an edge computing processing unit; The carrier ad hoc network communication module is used to send and receive carrier signals through power lines, and has network topology awareness and route discovery functions, and monitors the communication link quality indicators of adjacent nodes in real time. The edge computing processing unit is used to collect local electrical parameters and, based on the link quality indicators fed back by the communication module, to form a self-organizing network according to the dynamic network topology construction mechanism.

[0006] Preferably, the abnormal event semantic analysis module specifically includes: Call the chain path search unit: Based on the running state knowledge graph, use the breadth-first search algorithm to search all candidate root cause paths in reverse from the abnormal event node, and sort them according to the strength of causal association and time proximity; Large Model Semantic Analysis Unit: Extracts corresponding time-series data fragments, equipment configuration information, historical maintenance records, and environmental meteorological data for each candidate root cause path, constructs structured prompt words, inputs them into the large language model for reasoning, and outputs root cause judgment, credibility score, and reasoning evidence chain.

[0007] Preferably, based on the operational state knowledge graph, a breadth-first search algorithm is used to reverse-search all candidate root cause paths starting from the abnormal event node, and sorted according to the strength of causal association and temporal proximity. Specifically, this includes: Based on the detected abnormal event type, the corresponding abnormal event node is located in the running status knowledge graph. The abnormal event node is a communication interruption event node or a communication delay abnormal event node. Create a priority queue, add the abnormal event node as the starting node to the queue, and set the maximum search depth threshold and time window threshold; Take the current node out of the queue, traverse all incoming edges pointing to the current node in the running state knowledge graph, and obtain all predecessor nodes of the current node. The incoming edges include causal association edges and temporal association edges. For each predecessor node, if the path length does not exceed the maximum search depth threshold, the difference between the timestamp of the predecessor node and the timestamp of the abnormal event node does not exceed the time window threshold, and the causal association strength between the predecessor node and the current node is greater than the preset threshold α, then add it to the queue and record the complete path from the predecessor node to the abnormal event node. For all records, candidate root cause paths are sorted according to the strength of causal association as the first sorting key and time proximity as the second sorting key.

[0008] Preferably, the specific steps for sorting based on causal correlation strength as the first sorting key and temporal proximity as the second sorting key are as follows: First, sort them in descending order according to the weighted average of the causal association strength of each segment in all candidate root cause paths; Then, based on the weighted average of the temporal proximity of each segment in all candidate root cause paths, where temporal proximity is the reciprocal of the timestamp difference, they are sorted in descending order.

[0009] Preferably, the structured prompts include: role definition, anomaly information, sequentially ordered candidate root cause paths, analysis requirements, and thought chain guidance words.

[0010] Preferably, the large language model is any one of a commercial API model, a locally deployed open-source model, or a dedicated model fine-tuned for power system fault diagnosis tasks.

[0011] 7. The centralized management and control system for distribution boxes of carrier ad hoc network communication according to claim 6, characterized in that the output results of the intelligent topology optimization decision module and the subsequent topology adjustment effect are fed back to the operating status knowledge graph construction module for updating the causal relationships in the knowledge graph.

[0012] Preferably, the dynamic network topology construction mechanism specifically includes: Each smart distribution box terminal broadcasts beacon frames to discover neighboring nodes, forming the initial mesh network structure; Based on the initial mesh network structure, the link with the highest communication quality between each smart distribution box terminal and the centralized management platform is selected as the networking link of the smart distribution box terminal. The edge computing processing unit periodically evaluates the signal-to-noise ratio, bit error rate, and packet loss rate of each link connected to it. When the network link quality of the smart distribution box terminal is lower than the preset threshold, the smart distribution box terminal actively initiates a route rediscovery request to find an alternative network link and synchronously updates the routing table with adjacent nodes. When the centralized management platform detects an abnormal event, it analyzes whether a topology reconfiguration is triggered. If it is, it sends a network reconfiguration command to the smart distribution box terminals in the designated area. If it is not triggered, it does not respond.

[0013] Preferably, based on the initial mesh network structure, the link with the highest communication quality between each smart distribution box terminal and the centralized management platform is selected. The specific network links for the smart distribution box terminals include: Based on the initial mesh network structure, all communication links between the intelligent distribution box terminal and the centralized management platform are determined. The number of hops and link stability of each connected communication link are determined, and a comprehensive evaluation is conducted based on the number of hops and link stability of the communication links to determine the uplink networking link of each smart distribution box terminal.

[0014] Preferably, when the distribution box terminal detects that the communication interruption between itself and the upper-level node exceeds a set time, it actively initiates a route rediscovery request. During the search for a backup network link, the distribution box terminal is promoted to a temporary coordinator and is responsible for the data aggregation and temporary storage of its lower-level subnets.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention integrates a carrier self-organizing network communication module and an edge computing processing unit into each intelligent distribution box terminal, using power line carrier communication as the communication medium to achieve cabling-free networking, reducing construction and renovation costs. Through a dynamic network topology construction mechanism, it can automatically initiate route rediscovery when link quality degrades and automatically promote temporary coordinators to achieve self-healing when the network is interrupted. Furthermore, it combines the global perspective of a centralized management platform to issue topology reconstruction commands, thereby improving communication reliability and network adaptability while maintaining low cost.

[0016] This invention introduces a runtime status knowledge graph construction module, an abnormal event semantic analysis module, and an intelligent topology optimization decision-making module into a centralized management platform. By constructing an attribute graph model containing multiple types of nodes and their relationships, and employing a breadth-first search-based reverse root cause path search algorithm combined with the deep semantic reasoning capabilities of a large language model, it achieves root cause localization and interpretability analysis of abnormal communication events. This allows topology reconfiguration decisions to no longer rely solely on a single link quality threshold, but instead integrate the semantic understanding results of causal correlation strength, temporal proximity, and device runtime status. This effectively avoids misjudgments and invalid topology adjustments, while outputting a complete chain of reasoning evidence and repair suggestions, greatly improving the troubleshooting efficiency of maintenance personnel and the overall intelligent management level of the system. Attached Figure Description

[0017] Figure 1This is a block diagram of the centralized control system for power distribution boxes based on carrier ad hoc network communication proposed in this invention. Figure 2 This is a flowchart of the dynamic network topology construction mechanism proposed in this invention; Figure 3 This is a flowchart of the method for searching and sorting candidate root cause paths proposed in this invention. Detailed Implementation

[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Example 1: Reference Figure 1 As shown, a centralized management and control system for distribution boxes using carrier ad hoc network communication includes a centralized management platform and multiple intelligent distribution box terminals: The centralized management platform is deployed in the monitoring center to issue control commands, receive and display the operating status data of all distribution boxes. It has built-in modules for building an operating status knowledge graph, semantic analysis of abnormal events, and intelligent topology optimization decision-making. The centralized management platform is implemented using a server cluster and establishes a communication connection with at least one intelligent distribution box terminal through the TCP / IP protocol. It is also equipped with a graphical human-machine interface for real-time display of network topology, equipment status, alarm information, and operation and maintenance reports. The operational status knowledge graph construction module is used to construct and dynamically update an attribute graph model containing distribution box nodes, sensor nodes, communication link nodes, fault event nodes, and environmental factor nodes and their relationships, based on the electrical parameters, communication link quality indicators, equipment operation logs, and environmental data reported by each distribution box terminal. The abnormal event semantic analysis module is connected to the runtime status knowledge graph construction module. When an abnormal event is detected, it automatically triggers the semantic analysis process and outputs root cause judgment, credibility score and reasoning evidence chain. The intelligent topology optimization decision module is connected to the abnormal event semantic analysis module. It is used to combine the root cause judgment results output by the abnormal event semantic analysis module with the original link quality indicators to perform multi-dimensional weighted scoring and determine whether to trigger topology reconstruction. Each smart distribution box terminal integrates a carrier self-organizing network communication module and an edge computing processing unit; The carrier ad hoc network communication module is used to send and receive carrier signals through power lines, and has network topology awareness and route discovery functions, and monitors the communication link quality indicators of adjacent nodes in real time. The edge computing processing unit is used to collect local electrical parameters and, based on the link quality indicators fed back by the communication module, to form a self-organizing network according to the dynamic network topology construction mechanism.

[0020] Reference Figure 2 As shown, in some preferred embodiments, the dynamic network topology construction mechanism is specifically as follows: Each smart distribution box terminal broadcasts beacon frames to discover neighboring nodes and form an initial mesh network structure. The beacon frame contains information such as the node ID, communication capability, and remaining power. The broadcast period is 1 second, and it enters a steady state after 30 seconds. Based on the initial mesh network structure, the link with the highest communication quality between each smart distribution box terminal and the centralized management platform is selected as the networking link for the smart distribution box terminal. The specific steps are as follows: Based on the initial mesh network structure, all communication links between the intelligent distribution box terminal and the centralized management platform are determined. The number of node hops and link stability of each connected communication link are determined, and a comprehensive evaluation is performed based on the number of node hops and link stability of the communication link to determine the uplink networking link of each smart distribution box terminal. The link stability is defined as the average of the inverse of the variance of the signal-to-noise ratio of the most recent 10 samples between adjacent nodes in the connected communication link. In some preferred embodiments, a weighted summation method is used for the comprehensive evaluation based on the node hop count and link stability of the communication link. The specific steps are as follows: The reciprocal of the node hop count of the communication link is used as the node hop count evaluation index. After normalizing the node hop count evaluation index and link stability, a weighted sum is performed to obtain a comprehensive evaluation score. The link with the highest comprehensive evaluation score is selected as the uplink networking link. The specific calculation formula is as follows: To comprehensively evaluate the score, The normalized value of the node hop count evaluation metric. This is the normalized value for link stability. As weight, In some preferred embodiments, =0.4, =0.6; In some preferred embodiments, the TOPSIS approach is used for comprehensive evaluation based on the node hop count and link stability of the communication link. The specific steps are as follows: Construct decision matrix X, , This represents the total number of candidate communication links. Let be the number of node hops for the i-th candidate communication link. Let be the link stability of the i-th candidate communication link; The original decision matrix is ​​transformed into a normalized matrix using vector normalization. The calculation formula is: ; Set the weight of the node hop count as The weight of link stability is ,and + = 1. In this preferred embodiment, Take 0.4, Take 0.6, weighted normalization matrix ,in ; Determine the ideal solution and negative ideal solution : ; ; Calculate the Euclidean distance from each candidate link to the positive and negative ideal solutions. The distance from the i-th link to the positive ideal solution is: The distance from the i-th link to the negative ideal solution is: Calculate the relative proximity of each candidate link. : The value range of is [0,1]. A larger value indicates that the link is closer to the ideal solution, meaning it has better overall performance. All candidate links are ranked according to their relative proximity. Sort the links from largest to smallest and select the links with the highest proximity as the uplink networking links for the smart distribution box terminal. By comprehensively evaluating the number of hops and link stability of the communication link, the optimal uplink communication link for each smart distribution box terminal can be objectively and quantitatively selected based on a comprehensive consideration of both factors. This avoids the one-sidedness of evaluation by a single indicator and further improves the rationality of the network topology and communication quality.

[0021] The edge computing processing unit periodically evaluates the signal-to-noise ratio, bit error rate, and packet loss rate of each link connected to it. When the network link quality of the smart distribution box terminal is lower than the preset threshold, the smart distribution box terminal actively initiates a route rediscovery request to find an alternative network link and synchronously updates the routing table with adjacent nodes. Specifically, when the distribution box terminal detects that the communication interruption between itself and the upper-level node exceeds the set time, it actively initiates a route rediscovery request. During the search for a backup network link, the distribution box terminal is promoted to a temporary coordinator, responsible for the data aggregation and temporary storage of its lower-level subnets. Once the communication link is restored, it automatically returns the coordination rights, thereby realizing the self-healing of the network.

[0022] When the centralized management platform detects an abnormal event, it analyzes whether a topology reconfiguration is triggered. If it is triggered, it sends a network reconfiguration command to the smart distribution box terminal in the designated area. If it is not triggered, it does not respond. Abnormal events include, but are not limited to, communication interruption and communication delay anomalies.

[0023] Example 2: Based on Example 1, this example further proposes that the abnormal event semantic analysis module specifically includes: Call chain path search unit: Based on the running state knowledge graph, a breadth-first search algorithm is used to search all candidate root cause paths in reverse from the abnormal event node, and sort them according to the strength of causal association and temporal proximity; Large Model Semantic Analysis Unit: Extracts corresponding time-series data fragments, equipment configuration information, historical maintenance records, and environmental meteorological data for each candidate root cause path, constructs structured prompt words, inputs them into the large language model for reasoning, and outputs root cause judgment, credibility score, and reasoning evidence chain.

[0024] Specifically, refer to Figure 3 As shown, based on the runtime state knowledge graph, a breadth-first search algorithm is used to search backwards from the abnormal event node for all candidate root cause paths, and sorted by causal association strength and temporal proximity. Specifically, this includes: Based on the detected abnormal event type, the corresponding abnormal event node is located in the running status knowledge graph. The abnormal event node is either a communication interruption event node or a communication delay abnormal event node. Create a priority queue, add the abnormal event node as the starting node to the queue, and set the maximum search depth threshold and time window threshold; Take the current node out of the queue, traverse all incoming edges pointing to the current node in the running state knowledge graph, and get all the predecessor nodes of the current node. The incoming edges include causal relationship edges and temporal relationship edges. For each predecessor node, if the path length does not exceed the maximum search depth threshold, the difference between the timestamp of the predecessor node and the timestamp of the abnormal event node does not exceed the time window threshold, and the causal association strength between the predecessor node and the current node is greater than the preset threshold α, then add it to the queue and record the complete path from the predecessor node to the abnormal event node. For all records, the candidate root cause paths are sorted according to the strength of causal association as the first sorting key and the temporal proximity as the second sorting key. That is, they are first sorted in descending order according to the weighted average of the strength of causal association of each segment in the candidate root cause paths of all records. Then, based on the weighted average of the temporal proximity of each segment in all candidate root cause paths, where temporal proximity is the reciprocal of the timestamp difference, they are sorted in descending order.

[0025] In the reverse breadth-first search process, a priority queue is used instead of a regular queue to ensure that predecessor nodes with stronger causal relationships are traversed first, thus converging to the most likely root cause path as early as possible. Simultaneously, by setting maximum search depth and time window thresholds, the search space is effectively controlled, avoiding computational explosion due to excessively large graph size, and excluding irrelevant nodes that are too far apart in time, ensuring the timeliness and relevance of search results. Furthermore, a preset threshold for causal relationship strength is introduced for pruning, which eliminates weakly related or even unrelated redundant paths, further focusing on high-confidence candidate root cause paths. In the sorting stage, paths are first sorted in descending order by the weighted average of causal relationship strength, and then in descending order by the weighted average of temporal proximity. This ensures that when multiple paths have similar causal strength, paths that are closer in time to the occurrence of the anomalous event receive higher priority, thus ensuring that the candidate paths output to the large model semantic analysis unit possess both strong causal logic and temporal rationality. By combining reverse breadth-first search with multiple pruning and ranking strategies, this method achieves efficient and accurate extraction of candidate root cause paths from the runtime knowledge graph. Compared to full traversal or unconstrained search, this method reduces computational complexity while ensuring search completeness, enabling real-time preliminary screening of root cause paths even in large-scale distribution box networks. Simultaneously, the dual ranking mechanism based on causal correlation strength and temporal proximity ensures that the candidate paths received by the large model semantic analysis unit have the highest causal credibility and temporal proximity, thereby improving the accuracy and efficiency of large model root cause inference, avoiding interference from irrelevant paths in the inference process, and providing a high-quality data foundation for subsequent intelligent topology optimization decisions.

[0026] In this embodiment, the large language model is any one of the following: a commercial API model such as GPT-4, a locally deployed open source model such as Llama, or a dedicated model fine-tuned for power system fault diagnosis tasks. Structured prompts include: role definition, anomaly information, candidate root cause paths ordered in sequence, analysis requirements, and thought process guidance words, among which: Role definition is used to define the expert role played by the large language model. For example, you are a senior power system fault diagnosis expert who is good at analyzing the root causes of anomalies in the distribution box carrier communication network. The abnormal information is extracted from the time-series data segments, equipment configuration information, historical maintenance records and environmental meteorological data for each candidate root cause path; The candidate root cause path is a sorted list of candidate root cause paths. Each path contains fields such as node sequence, causal relationship strength of each segment, and timestamp difference. Analysis requirements: Indicate the output results and output format of the model. In this embodiment, the output format is required to be JSON. The output results are to determine whether each candidate path is the true root cause, and output the reasoning process and credibility score. Thought chain guidance: Instructs the large language model to reason step by step according to the steps of data observation, causal inference, elimination of interference, and conclusion confirmation.

[0027] The output of the intelligent topology optimization decision module and the subsequent topology adjustment effects are fed back to the runtime knowledge graph construction module to update the causal relationships in the knowledge graph.

[0028] The structured prompts are designed with full consideration of the professional requirements of power system fault diagnosis: the role definition clarifies the expert identity of the model, making its output more domain-specific; the anomaly information includes multi-dimensional context such as time series data, equipment configuration, historical maintenance records, and environmental meteorology, providing rich reasoning material for the large model; the sequentially ordered candidate root cause paths ensure that the large model prioritizes the paths with the highest causal strength and closest time, improving reasoning efficiency; the analysis requirements constrain the output in JSON format, facilitating automated parsing and integration by downstream modules; the thought chain guides proceed step by step according to the steps of "data observation → causal inference → interference elimination → conclusion confirmation," effectively suppressing the illusion problem common in open-domain reasoning of large models, ensuring the logical rigor and traceability of the output results. In addition, the output results and topology adjustment effects of the intelligent topology optimization decision module are fed back to the operational state knowledge graph construction module to update the causal relationships between relevant nodes in the graph. For example, if a topology adjustment successfully solves the communication interruption problem, the strength value of the corresponding causal edge will increase, thus forming a closed loop of continuous iterative optimization, enabling the system's fault diagnosis capability to continuously improve with the accumulation of operational data.

[0029] By introducing a large language model and combining it with carefully designed structured prompts, a leapfrog upgrade from numerical-driven to semantic-driven approaches has been achieved. The large model can comprehensively utilize multi-source heterogeneous information for deep causal reasoning, not only outputting root cause judgments and credibility scores but also providing a complete chain of reasoning evidence. This allows maintenance personnel to intuitively understand the ins and outs of fault occurrences, reducing the cognitive burden and time cost of manual troubleshooting. Multiple model selection options give the system flexible adaptability in different scenarios, leveraging the powerful reasoning capabilities of cloud-based large models while also employing local deployment solutions to ensure security in data-sensitive scenarios. Simultaneously, the introduction of a feedback mechanism allows the knowledge graph to learn from each topology adjustment, continuously correcting and strengthening causal relationships, achieving system self-evolution. Over long-term operation, the accuracy of fault diagnosis and the rationality of topology optimization will continuously improve, enhancing the system's intelligence level and long-term service value.

[0030] In summary, the advantages of this invention are as follows: By integrating a carrier self-organizing network communication module and an edge computing processing unit into each intelligent distribution box terminal, and utilizing power line carrier communication as the communication medium to achieve cabling-free networking, construction and renovation costs are reduced. Through a dynamic network topology construction mechanism, it can automatically initiate route rediscovery when link quality degrades and automatically promote temporary coordinators to achieve self-healing when the network is interrupted. Furthermore, by combining the global perspective of the centralized management platform to issue topology reconstruction commands, it improves communication reliability and network adaptability while maintaining low cost. Additionally, by introducing a runtime status knowledge graph construction module and an abnormal event language into the centralized management platform... The semantic analysis module and intelligent topology optimization decision module construct an attribute graph model containing multiple types of nodes and their relationships, and adopt a reverse root cause path search algorithm based on breadth-first search. Combined with the deep semantic reasoning capabilities of a large language model, they achieve root cause localization and interpretability analysis of abnormal communication events. This makes topology reconstruction decisions no longer rely solely on a single link quality threshold, but integrates semantic understanding results of causal correlation strength, temporal proximity, and device operating status, effectively avoiding misjudgments and invalid topology adjustments. At the same time, they output a complete chain of reasoning evidence and repair suggestions, greatly improving the troubleshooting efficiency of maintenance personnel and the overall intelligent management level of the system.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A centralized management and control system for distribution boxes using carrier ad hoc network communication, characterized in that, Includes a centralized management platform and multiple smart distribution box terminals: The centralized management platform is deployed in the monitoring center to issue control commands, receive and display the operating status data of all distribution boxes. It has built-in modules for building an operating status knowledge graph, analyzing abnormal events semantically, and making intelligent topology optimization decisions. The operational status knowledge graph construction module is used to construct and dynamically update an attribute graph model containing distribution box nodes, sensor nodes, communication link nodes, fault event nodes, and environmental factor nodes and their relationships, based on the electrical parameters, communication link quality indicators, equipment operation logs, and environmental data reported by each distribution box terminal. The abnormal event semantic analysis module is connected to the running status knowledge graph construction module, and is used to automatically trigger the semantic analysis process when an abnormal event is detected, and output the root cause judgment, credibility score and reasoning evidence chain. The intelligent topology optimization decision module is connected to the abnormal event semantic analysis module. It is used to combine the root cause judgment results output by the abnormal event semantic analysis module with the original link quality indicators to perform multi-dimensional weighted scoring and determine whether to trigger topology reconstruction. Each smart distribution box terminal integrates a carrier self-organizing network communication module and an edge computing processing unit; The carrier ad hoc network communication module is used to send and receive carrier signals through power lines, and has network topology awareness and route discovery functions, and monitors the communication link quality indicators of adjacent nodes in real time. The edge computing processing unit is used to collect local electrical parameters and, based on the link quality indicators fed back by the communication module, to form a self-organizing network according to the dynamic network topology construction mechanism.

2. The centralized control system for distribution boxes using carrier ad hoc network communication according to claim 1, characterized in that, The abnormal event semantic analysis module specifically includes: Call the chain path search unit: Based on the running state knowledge graph, use the breadth-first search algorithm to search all candidate root cause paths in reverse from the abnormal event node, and sort them according to the strength of causal association and time proximity; Large Model Semantic Analysis Unit: Extracts corresponding time-series data fragments, equipment configuration information, historical maintenance records, and environmental meteorological data for each candidate root cause path, constructs structured prompt words, inputs them into the large language model for reasoning, and outputs root cause judgment, credibility score, and reasoning evidence chain.

3. The centralized control system for distribution boxes using carrier ad hoc network communication according to claim 2, characterized in that, Based on the aforementioned operational state knowledge graph, a breadth-first search algorithm is used to reverse-search all candidate root cause paths starting from the abnormal event nodes, and sorted according to causal correlation strength and temporal proximity. Specifically, this includes: Based on the detected abnormal event type, the corresponding abnormal event node is located in the running status knowledge graph. The abnormal event node is a communication interruption event node or a communication delay abnormal event node. Create a priority queue, add the abnormal event node as the starting node to the queue, and set the maximum search depth threshold and time window threshold; Take the current node out of the queue, traverse all incoming edges pointing to the current node in the running state knowledge graph, and obtain all predecessor nodes of the current node. The incoming edges include causal association edges and temporal association edges. For each predecessor node, if the path length does not exceed the maximum search depth threshold, the difference between the timestamp of the predecessor node and the timestamp of the abnormal event node does not exceed the time window threshold, and the causal association strength between the predecessor node and the current node is greater than the preset threshold α, then add it to the queue and record the complete path from the predecessor node to the abnormal event node. For all records, candidate root cause paths are sorted according to the strength of causal association as the first sorting key and time proximity as the second sorting key.

4. The centralized control system for distribution boxes using carrier ad hoc network communication according to claim 3, characterized in that, The specific steps for sorting based on the strength of causal association as the first sorting key and time proximity as the second sorting key are as follows: First, sort them in descending order according to the weighted average of the causal association strength of each segment in all candidate root cause paths; Then, based on the weighted average of the temporal proximity of each segment in all candidate root cause paths, where temporal proximity is the reciprocal of the timestamp difference, they are sorted in descending order.

5. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 4, characterized in that, The structured prompts include: role definition, anomaly information, candidate root cause paths ordered in sequence, analysis requirements, and thought chain guidance words.

6. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 5, characterized in that, The large language model can be any one of a commercial API model, a locally deployed open-source model, or a dedicated model fine-tuned for power system fault diagnosis tasks.

7. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 6, characterized in that, The output of the intelligent topology optimization decision module and the subsequent topology adjustment effect are fed back to the runtime knowledge graph construction module to update the causal relationships in the knowledge graph.

8. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 1, characterized in that, The dynamic network topology construction mechanism is specifically as follows: Each smart distribution box terminal broadcasts beacon frames to discover neighboring nodes, forming the initial mesh network structure; Based on the initial mesh network structure, the link with the highest communication quality between each smart distribution box terminal and the centralized management platform is selected as the networking link of the smart distribution box terminal. The edge computing processing unit periodically evaluates the signal-to-noise ratio, bit error rate, and packet loss rate of each link connected to it. When the network link quality of the smart distribution box terminal is lower than the preset threshold, the smart distribution box terminal actively initiates a route rediscovery request to find an alternative network link and synchronously updates the routing table with adjacent nodes. When the centralized management platform detects an abnormal event, it analyzes whether a topology reconfiguration is triggered. If it is, it sends a network reconfiguration command to the smart distribution box terminals in the designated area. If it is not triggered, it does not respond.

9. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 8, characterized in that, Based on the initial mesh network structure, the link with the highest communication quality between each smart distribution box terminal and the centralized management platform is selected. The specific network links for the smart distribution box terminals include: Based on the initial mesh network structure, all communication links between the intelligent distribution box terminal and the centralized management platform are determined. The number of hops and link stability of each connected communication link are determined, and a comprehensive evaluation is conducted based on the number of hops and link stability of the communication links to determine the uplink networking link of each smart distribution box terminal.

10. A centralized control system for distribution boxes using carrier ad hoc network communication according to claim 9, characterized in that, When the distribution box terminal detects that its communication interruption with the upper-level node exceeds the set time, it actively initiates a route rediscovery request. During the search for a backup network link, the distribution box terminal is promoted to a temporary coordinator, responsible for the data aggregation and temporary storage of its lower-level subnets.

Images