Ring main unit live detection system and control method fusing infrared sensing and phm diagnosis

Abstract

This invention relates to the field of ring main unit (RMU) testing technology, specifically to a live-line testing system and control method for RMUs integrating infrared sensing and PHM (Prognostics and Health Management) diagnostics. It includes: a RMU topology modeling unit; a multi-source synchronous acquisition unit; a dynamic attention PHM diagnostic unit with a built-in dynamic attention mechanism adapted to the RMU's power operation scenarios; and a fault location output unit. This invention integrates a Fourier steady-state heat conduction model into the dynamic attention mechanism, quantifying the heat conduction attenuation coefficient between nodes. This effectively distinguishes between temperature changes caused by internal heat conduction effects within the RMU and abnormal temperature rises caused by the fault source itself, avoiding misdiagnosis. Simultaneously, it dynamically adjusts the weight allocation of infrared temperature features and electrical features based on the real-time operating conditions of the RMU, adapting to the fault feature analysis needs under different operating scenarios such as high load and low load, thus improving the accuracy and scenario adaptability of fault diagnosis in live-line testing scenarios of distribution network RMUs.

Description

Technical Field

[0001] This invention relates to the field of ring main unit testing technology, and more specifically, to a live-line testing system and control method for ring main units that integrates infrared sensing and PHM diagnostics. Background Technology

[0002] As a key core device in the power distribution network, the operating status of ring main units directly affects the reliability and security of the power supply system. Live-line detection technology, capable of monitoring internal operating parameters and identifying potential faults in real time without interrupting power supply, is a core means of ensuring the safe and stable operation of ring main units. The combination of the non-contact temperature measurement advantages of infrared sensing technology with the data analysis capabilities of PHM (Prognostics and Health Management) diagnostic technology has become an important development direction in the field of live-line detection of ring main units.

[0003] In the existing technology, relevant patents have explored the field of ring main unit fault detection and early warning. For example, Chinese patent CN202210733741.5 discloses a fault detection method for power system insulated ring main units. It trains historical data using a radial basis function model optimized by a genetic algorithm, collects various parameters of the ring main unit, and compares them with statistical limits to achieve fault identification and troubleshooting, thereby improving the accuracy of fault judgment. Another example is Chinese patent CN202211359471.2, which discloses an online fault early warning method and system for ring main units. It receives fault information and operating data sent by monitoring devices, determines the early warning status through threshold judgment, and sends the results to a remote server and terminal to realize fault early warning and maintenance notification.

[0004] Despite the design advantages of the aforementioned technical solutions, they also suffer from the following technical defects: Firstly, the fault cause judgment lacks physical support and adaptability to operating conditions. Chinese patent CN202210733741.5 uses fixed logic analysis data based on model fitting and statistical limit comparison, while Chinese patent CN202211359471.2 relies on fixed thresholds to determine the warning status. Neither of these solutions considers the interference of heat transfer phenomena within the enclosed space of the ring main unit on the temperature monitoring signal, making it impossible to distinguish between temperature fluctuations caused by environmental heat conduction and abnormal temperature rises caused by the fault source itself, which can easily lead to misjudgments of faults. Furthermore, neither solution incorporates real-time load fluctuation adjustments within the ring main unit. The core analytical dimensions lack targeted adaptation to the differences in fault characteristics under different operating conditions such as high and low loads, resulting in limited accuracy and scenario adaptability of diagnostic results. Secondly, the depth of multi-source data fusion is insufficient, limiting the accuracy of fault location: Chinese patent CN202210733741.5 only performs model fitting analysis on the collected parameters, and Chinese patent CN202211359471.2 only performs simple data reception and threshold judgment. Neither of them establishes the correlation constraints between the physical structure of the ring main unit and the monitoring nodes, failing to achieve deep coupling of multi-source monitoring data. This results in insufficient accuracy in fault feature extraction, making it difficult to locate faults to the specific connection point level, affecting the targeting and efficiency of maintenance. In view of this, we propose a ring main unit live-line detection system and control method that integrates infrared sensing and PHM diagnosis. Summary of the Invention

[0005] The purpose of this invention is to provide a live-line detection system and control method for ring main units that integrates infrared sensing and PHM diagnosis, so as to solve the problems mentioned in the background art, such as the lack of physical law support for fault cause judgment, insufficient adaptability to operating conditions, insufficient depth of multi-source data fusion, and limited accuracy of fault location.

[0006] To address the aforementioned technical problems, one objective of this invention is to provide a live-line detection system for ring main units that integrates infrared sensing and PHM diagnostics, comprising:

[0007] The ring main unit topology modeling unit transforms the physical structure of the ring main unit into a topology graph model containing node physical location parameters and connection characteristic parameters. Each key electrical connection point is independently defined as a node, and the connection weights between nodes are determined based on physical distance and contact characteristics. A node mapping table is then generated and sent to the multi-source synchronous acquisition unit.

[0008] The multi-source synchronous acquisition unit receives the node mapping relationship table output by the ring main unit topology modeling unit, uses an infrared thermal imaging acquisition device to acquire thermal image data inside the ring main unit based on the node mapping relationship table, maps it to the corresponding nodes according to the structural partition, collects the electrical parameter characteristics of each node location, realizes node-level deep fusion of infrared spatial information and electrical time-series data, and sends the node-level deep fusion data of infrared spatial information and electrical time-series data and the real-time operating status data of the ring main unit to the dynamic attention PHM diagnostic unit.

[0009] The Dynamic Attention PHM diagnostic unit receives node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit, as well as real-time operating status data of the ring main unit. It has a built-in dynamic attention mechanism adapted to the power operation scenario of the ring main unit. The heat conduction physical model reflecting the internal heat conduction law of the ring main unit is integrated into the dynamic attention mechanism adapted to the power operation scenario of the ring main unit. Based on the heat conduction physical model, the heat conduction attenuation coefficient between nodes is quantified. Combined with the real-time operating status of the ring main unit, the weights of infrared temperature features and electrical features are automatically adjusted. Under high load conditions, the focus is on analyzing changes in electrical features, and under low load conditions, the focus is on identifying subtle gradients in infrared temperature distribution. Based on the heat conduction attenuation coefficient, the temperature changes of heat conduction effect and fault source are distinguished. The fault location analysis results are sent to the fault location output unit. At the same time, the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenario of the ring main unit are fed back to the multi-source synchronous acquisition unit.

[0010] The fault location output unit receives the fault location analysis results output by the dynamic attention PHM diagnostic unit, outputs specific connection point-level fault location information and maintenance scope, and simultaneously feeds back the fault node location information to the ring network cabinet topology modeling unit.

[0011] As a further improvement to this technical solution, the ring network cabinet topology modeling unit includes a topology structure parsing module, a node independent definition module, a connection weight calculation module, and a mapping table generation and sending module, wherein:

[0012] The topology parsing module obtains the physical structure parameters of the ring main unit and transforms the physical structure of the ring main unit into a topology graph model that includes node physical location parameters and connection characteristic parameters.

[0013] The node independent definition module filters the key electrical connection points of the ring main unit from the topology graph model and defines each key electrical connection point as a node independently;

[0014] The connection weight calculation module calculates and determines the connection weight between nodes based on the physical distance and contact characteristics between nodes;

[0015] The mapping table generation and sending module generates a node mapping relationship table based on node information and the connection weights between nodes, and sends the node mapping relationship table to the multi-source synchronous acquisition unit.

[0016] As a further improvement to this technical solution, the multi-source synchronous acquisition unit includes a data receiving module, an infrared data acquisition and mapping module, an electrical parameter acquisition module, a node-level data fusion module, and a data sending module, wherein:

[0017] The data receiving module receives and stores the node mapping relationship table output by the ring network cabinet topology modeling unit;

[0018] The infrared data acquisition and mapping module, based on the node mapping relationship table, acquires thermal image data inside the ring network cabinet through the infrared thermal image acquisition device, and maps the thermal image data to the corresponding nodes according to the structural partitions.

[0019] The electrical parameter acquisition module collects the electrical parameter characteristics of each node location and synchronously records the parameter acquisition timestamp;

[0020] The node-level data fusion module performs deep node-level fusion of infrared spatial information and electrical time-series data, associating the thermal image data, electrical parameter characteristics, and acquisition timestamps of the corresponding nodes.

[0021] The data transmission module synchronously sends the node-level deep fusion data of infrared spatial information and electrical timing data, as well as the real-time operating status data of the ring main unit, to the Dynamic Attention (PHM) diagnostic unit.

[0022] As a further improvement to this technical solution, the dynamic attention PHM diagnostic unit includes a data receiving and preprocessing module, an attention mechanism integration module, a weight adjustment and fault differentiation module, and a data interaction module, wherein:

[0023] The data receiving and preprocessing module receives node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit, as well as real-time operating status data of the ring main unit. It performs format standardization and outlier removal on the data to generate a standardized node dataset.

[0024] The attention mechanism integration module has a built-in dynamic attention mechanism adapted to the power operation scenario of the ring main unit, which integrates the heat conduction physical model reflecting the internal heat conduction law of the ring main unit into the dynamic attention mechanism adapted to the power operation scenario of the ring main unit.

[0025] The weight adjustment and fault differentiation module quantifies the inter-node heat conduction attenuation coefficient based on a heat conduction physical model. The weighting coefficient of infrared temperature features is automatically adjusted based on the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient Under high load conditions, the focus is on analyzing changes in electrical characteristics; under low load conditions, the focus is on identifying subtle gradients in the infrared temperature distribution, based on the thermal conductivity attenuation coefficient. Distinguish between the heat conduction effect and the temperature change of the fault source to generate fault location analysis results;

[0026] The data interaction module sends the fault location analysis results to the fault location output unit, and at the same time feeds back the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenario of the ring main unit to the multi-source synchronous acquisition unit.

[0027] As a further improvement to this technical solution, the attention mechanism integration module includes a model selection and configuration submodule, a logic embedding submodule, and a mapping relationship construction submodule, wherein:

[0028] The model selection and configuration submodule adopts the Fourier steady-state heat conduction model as the physical model of heat conduction that reflects the heat conduction law inside the ring main unit, and presets the value range of the core parameters of the model to adapt to the heat conduction characteristics of the closed space of the ring main unit.

[0029] The logic embedding submodule embeds the node heat conduction calculation logic of the Fourier steady-state heat conduction model into the weight allocation layer of the dynamic attention mechanism adapted to the power operation scenario of the ring network cabinet, so that the calculation process of the heat conduction physical model is deeply coupled with the attention weight iteration process.

[0030] The mapping relationship construction submodule establishes a one-to-one mapping relationship between node heat conduction correlation degree and attention weight based on the node heat conduction calculation results.

[0031] As a further improvement to this technical solution, the weight adjustment and fault differentiation module uses the thermal conduction attenuation coefficient... The process of distinguishing between heat conduction effects and fault source temperature changes includes the following steps:

[0032] S33.1. Based on the Fourier steady-state heat conduction model and combined with the physical attribute parameters of the nodes in the ring main unit topology model, the node... With nodes Thermal conduction attenuation coefficient between ;

[0033] S33.2, Preset thermal conductivity attenuation coefficient threshold The thermal conductivity attenuation coefficient obtained from S33.1 With threshold By comparing the results, the cause of the node temperature change is determined, the interference of heat conduction effect on fault determination is shielded, and a judgment basis is provided for subsequent feature weight adjustment.

[0034] As a further improvement to this technical solution, the weight adjustment and fault differentiation module adjusts the infrared temperature feature weight coefficient in conjunction with the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient The process includes the following steps:

[0035] S33.3, Thermal conductivity attenuation coefficient obtained based on S33.1 Based on the ratio of real-time load power to rated load power of the ring main unit, the infrared temperature characteristic weighting coefficient is determined. With electrical characteristic weighting coefficient The allocation ratio, and the infrared temperature feature weighting coefficient With electrical characteristic weighting coefficient satisfy ;

[0036] S33.4. Based on the real-time load power of the ring main unit, three operating conditions are divided: high load, low load, and intermediate load. Based on the weight allocation ratio determined in S33.3, the infrared temperature characteristic weight coefficient is dynamically adjusted. With electrical characteristic weighting coefficient The values ​​are adapted to the feature analysis needs under different working conditions.

[0037] As a further improvement to this technical solution, the data interaction module includes a fault result structured processing submodule, an encrypted transmission submodule, a weight parameter organization submodule, and a data acquisition optimization instruction generation submodule, wherein:

[0038] The fault result structuring processing submodule receives the fault location analysis results generated by the weight adjustment and fault differentiation module, and extracts the fault node number, fault type, and corresponding heat conduction attenuation coefficient. Infrared temperature feature weighting coefficient and electrical characteristic weighting coefficient , Generate structured fault information according to a preset format;

[0039] The encrypted transmission submodule uses a dedicated encrypted communication protocol for power systems to send structured fault information to the fault location output unit;

[0040] The weight parameter sorting submodule summarizes the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenario of the ring main unit, removes invalid parameters and standardizes the format.

[0041] The data acquisition optimization instruction generation submodule generates data acquisition optimization instructions based on the processed parameters, according to the infrared temperature feature weighting coefficient. Electrical characteristic weighting coefficient and thermal conductivity attenuation coefficient Based on the numerical values, the infrared thermal imaging frame rate, electrical parameter acquisition accuracy, and regional data acquisition intensity of the multi-source synchronous acquisition unit were adjusted respectively.

[0042] As a further improvement to this technical solution, the fault location output unit includes a fault result receiving and parsing module, a fault information output module, and a feedback information processing module, wherein:

[0043] The fault result receiving and parsing module receives the fault location analysis results output by the Dynamic Attention PHM diagnostic unit, extracts the fault node number, physical coordinates and associated topology path information, and completes information parsing and organization.

[0044] Based on the parsed information, the fault information output module generates specific connection point-level fault location information and maintenance scope, and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format.

[0045] The feedback information processing module performs format standardization processing on the fault node location information and feeds back the standardized fault node location information to the ring main unit topology modeling unit, providing data support for updating the ring main unit topology model.

[0046] The second objective of this invention is to provide a control method for live-line detection of ring main units that integrates infrared sensing and PHM diagnostics. Based on the aforementioned live-line detection system for ring main units that integrates infrared sensing and PHM diagnostics, the method includes the following steps:

[0047] S1. Topology modeling and node mapping table generation: Obtain the physical structure parameters of the ring main unit, screen the key electrical connection points of the ring main unit and define them as nodes independently, determine the connection weights based on the physical distance and contact characteristics between nodes, generate the node mapping table and send it to the data acquisition stage.

[0048] S2. Multi-source data synchronous acquisition and node-level deep fusion: Receive and store the node mapping relationship table, acquire the internal thermal image data of the ring main unit through the infrared thermal imaging acquisition device based on the node mapping relationship table, map it to the corresponding node according to the structural partition, collect the electrical parameter characteristics of each node location and synchronously record the parameter acquisition timestamp, perform node-level deep fusion of infrared spatial information and electrical time series data, and synchronously send the node-level deep fusion data and the real-time operating status data of the ring main unit to the diagnostic stage.

[0049] S3. Dynamic Attention PHM Diagnosis and Feature Weight Adjustment Parameter Feedback: Receives node-level deep fusion data and real-time operating status data of the ring main unit. After format standardization and outlier removal, a standardized node dataset is generated. The Fourier steady-state heat conduction model is integrated into a dynamic attention mechanism adapted to the power operation scenario of the ring main unit. The heat conduction attenuation coefficient between nodes is quantified. The operating conditions are divided according to the real-time operating status of the ring main unit, and the weights of infrared temperature features and electrical features are dynamically adjusted. Under high load conditions, the focus is on analyzing changes in electrical features, while under low load conditions, the focus is on identifying subtle gradients in infrared temperature distribution. Based on the heat conduction attenuation coefficient, the temperature changes of heat conduction effect and fault source are distinguished. Fault location analysis results are generated and sent to the fault location output stage. At the same time, the feature weight adjustment parameters are summarized to generate data acquisition optimization instructions and fed back to the data acquisition stage.

[0050] S4. Fault Location Output and Fault Node Location Information Feedback: Receives fault location analysis results, extracts fault node number, physical coordinates and associated topology path information and parses and organizes them, generates specific connection point-level fault location information and maintenance scope and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format, and feeds back the fault node location information to the topology modeling stage after standardizing the format.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] 1. This invention integrates a Fourier steady-state heat conduction model into a dynamic attention mechanism adapted to the power operation scenarios of ring main units (RNBs). By quantifying the heat conduction attenuation coefficient between nodes, it can effectively distinguish between temperature changes caused by internal heat conduction effects within the RNB and abnormal temperature rises caused by the fault source itself, thus avoiding misdiagnosis. Simultaneously, by dynamically adjusting the weight allocation of infrared temperature features and electrical features based on the real-time operating conditions of the RNB, it adapts to the fault feature analysis needs under different operating scenarios such as high load and low load, improving the accuracy and scenario adaptability of fault diagnosis in live detection scenarios of distribution network RNBs.

[0053] 2. This invention constructs a ring main unit topology model that includes node physical location parameters and connection characteristic parameters, establishes node-level association constraints between key electrical connection points and monitoring data, and achieves deep fusion of infrared spatial information and electrical time-series data. This not only accurately extracts fault features but also locates faults to specific connection points, improving the accuracy of fault location and the targeted nature of maintenance work. At the same time, through a closed-loop design that feeds back feature weight adjustment parameters from the diagnostic stage to the acquisition stage and feeds back fault node location information from the output stage to the topology modeling stage, the data acquisition strategy and topology model are continuously optimized, further ensuring the long-term stability and reliability of the system. Attached Figure Description

[0054] Figure 1This is a schematic diagram of the system framework of the present invention;

[0055] Figure 2 This is a schematic diagram of the workflow of the Dynamic Attention PHM diagnostic unit in this invention;

[0056] Figure 3 This is a schematic diagram of the method steps of the present invention;

[0057] The meanings of the labels in the diagram are as follows:

[0058] 1. Ring network cabinet topology modeling unit; 11. Topology structure parsing module; 12. Independent node definition module; 13. Connection weight calculation module; 14. Mapping table generation and sending module;

[0059] 2. Multi-source synchronous acquisition unit; 21. Data receiving module; 22. Infrared data acquisition and mapping module; 23. Electrical parameter acquisition module; 24. Node-level data fusion module; 25. Data transmission module;

[0060] 3. Dynamic Attention PHM Diagnostic Unit; 31. Data Reception and Preprocessing Module; 32. Attention Mechanism Integration Module; 33. Weight Adjustment and Fault Differentiation Module; 34. Data Interaction Module;

[0061] 4. Fault location output unit; 41. Fault result receiving and parsing module; 42. Fault information output module; 43. Feedback information processing module. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0063] like Figures 1-2 As shown, this embodiment provides a live-line detection system for ring main units that integrates infrared sensing and PHM diagnostics, including:

[0064] Ring Main Unit Topology Modeling Unit 1: The Ring Main Unit Topology Modeling Unit 1 transforms the physical structure of the ring main unit into a topology graph model containing node physical location parameters and connection characteristic parameters. Each key electrical connection point is defined independently as a node, and the connection weight between nodes is determined based on physical distance and contact characteristics. A node mapping relationship table is generated and sent to the multi-source synchronous acquisition unit 2.

[0065] In this embodiment, the ring network cabinet topology modeling unit 1 includes a topology structure parsing module 11, a node independent definition module 12, a connection weight calculation module 13, and a mapping table generation and sending module 14, wherein:

[0066] The topology parsing module 11 obtains the physical structure parameters of the ring network cabinet and transforms the physical structure of the ring network cabinet into a topology graph model that includes node physical location parameters and connection characteristic parameters;

[0067] Specifically, the core function of the topology parsing module 11 is to obtain the physical structure parameters of the ring network cabinet and convert them into a topology graph model that includes node physical location parameters and connection characteristic parameters. The specific operation is as follows:

[0068] Physical structure parameter acquisition: The topology analysis module 11 reads the electrical schematic diagram and mechanical structure diagram of the ring main unit, extracting the cabinet dimensions, component layout, electrical connection paths, key connection point locations, and specifications. Simultaneously, it uses a laser rangefinder and a high-definition industrial camera to conduct on-site surveys, verifying the actual installation parameters of the commissioned ring main unit and correcting discrepancies between the drawings and the actual site conditions. Specific parameters acquired include: cabinet length, width, and height dimensions; the routing length and cross-sectional dimensions of busbars and branch lines; the installation coordinates of components such as circuit breakers; and the number, distribution, and material / structure type of connection points.

[0069] Topology graph model construction: The topology structure analysis module 11 uses graph theory to construct an undirected topology graph model. The vertices of the model correspond to potential nodes, and the edges correspond to the electrical connection relationships between nodes. The physical location parameters of the nodes (two-dimensional coordinates with the lower left corner of the cabinet as the origin) and connection characteristic parameters (connection type, conductor cross-section, insulation level) are marked in the model. The model is stored in XML format to ensure interactive compatibility with the multi-source synchronous acquisition unit 2.

[0070] The node independent definition module 12 filters the key electrical connection points of the ring main unit from the topology graph model and defines each key electrical connection point as a node independently;

[0071] Specifically, the core function of the node independent definition module 12 is to select key electrical connection points from the topology graph model of the topology parsing module 11 and define them independently as nodes. The specific operation is as follows:

[0072] Key electrical connection point screening: Based on fault statistics, the independent node definition module 12 screens three types of key electrical connection points: current transmission transfer points, active contacts, and high current dense connection areas, specifically including bus joints, circuit breaker contacts, disconnector contacts, cable terminal joints, and grounding switch connection points.

[0073] Independent node identification: The independent node definition module 12 uses a four-level coding rule of "cabinet number - area number - type number - serial number" to uniquely identify the selected connection points. Each node corresponds one-to-one with the vertex of the topology graph model and is associated with the storage of information such as node physical coordinates, material, and structural type.

[0074] The connection weight calculation module 13 calculates and determines the connection weight between nodes based on the physical distance and contact characteristics between nodes;

[0075] Specifically, the core function of the connection weight calculation module 13 is to quantify the correlation strength between heat conduction and electrical signal transmission between nodes based on the physical distance and contact characteristics between nodes, determine the connection weight between nodes, and provide basic coefficients for subsequently distinguishing heat conduction effects from fault sources. The specific implementation process is as follows:

[0076] Parameter extraction: The connection weight calculation module 13 extracts the physical distance between any two related nodes from the topology graph model constructed by the topology parsing module 11. (Unit: mm); Extract the contact resistance of the connection point between two nodes from the node information stored in the node independent definition module 12. (Unit: μΩ), Contact pressure (Unit: N) Contact area (Unit: mm2). The above parameter values ​​are all determined based on the ring main unit design manual and installation process standards. For example, the contact resistance of the copper busbar bolt connection ranges from 50μΩ to 200μΩ, and the contact pressure ranges from 300N to 800N.

[0077] Parameter dimension normalization: Due to the differences in the dimensions of each parameter, the connection weight calculation module 13 adopts a linear normalization method to map all parameters to the [0,1] interval, eliminating the influence of dimensions on the calculation results. Taking the normalization of physical distance as an example, it is negatively correlated with the connection weight (the closer the distance, the stronger the association between nodes). The normalization formula is:

[0078] ;

[0079] in, This represents the maximum physical distance between all nodes within the ring main unit. This represents the minimum physical distance between adjacent nodes. This is the normalized value for physical distance.

[0080] Meanwhile, the normalization logic of contact resistance is consistent with that of physical distance and is negatively correlated with connection weight. It is executed with reference to the normalization formula of physical distance mentioned above. Contact pressure and contact area are positively correlated with connection weight. After adjusting the molecular calculation logic with reference to the formula mentioned above, normalization processing is performed.

[0081] Inter-node connection weight calculation: The connection weight calculation module 13 calculates the final connection weight using a weighted summation method. Based on the operating conditions of the ring main unit, fault statistics, and on-site testing requirements, a weight allocation strategy is set for the normalized parameters corresponding to physical distance, contact resistance, contact pressure, and contact area. The weight coefficients of each parameter satisfy the constraint that the summation is 1. The specific calculation formula is as follows:

[0082] ;

[0083] In the formula, This represents the connection weight between nodes, with a value range of [0,1]. These are the weighting coefficients for the normalized parameters of physical distance, contact resistance, contact pressure, and contact area, respectively. ; The closer the value is to 1, the stronger the correlation between heat conduction and electrical signal transmission between the two nodes.

[0084] Calculation result storage: The connection weight calculation module 13 stores the identifier, physical distance, contact characteristic parameters, and finally calculated connection weight of each group of associated nodes. Associative storage is performed to provide complete data support for the node mapping relationship table to be constructed by the mapping table generation and sending module 14.

[0085] The mapping table generation and sending module 14 generates a node mapping relationship table based on node information and the connection weights between nodes, and sends the node mapping relationship table to the multi-source synchronous acquisition unit 2.

[0086] Specifically, the core function of the mapping table generation and sending module 14 is to generate a node mapping relationship table based on node information and connection weights, and send it to the multi-source synchronous acquisition unit 2. The specific operation is as follows:

[0087] Node mapping relationship table construction: The mapping table generation and sending module 14 extracts the node identifier, location and other information from the node independent definition module 12, as well as the weight data from the connection weight calculation module 13, to construct a two-dimensional node mapping relationship table. The core fields include source node identifier, target node identifier, physical distance, connection weight, connection characteristic parameters, and topology path remarks.

[0088] Mapping table transmission and verification: The mapping table generation and transmission module 14 transmits the mapping table via ModbusTCP or IEC61850 protocol, and performs CRC32 verification before transmission; if no feedback is received from the multi-source synchronous acquisition unit 2, a retransmission mechanism is triggered, with a retransmission interval of 1 second and a maximum of 3 retransmissions to ensure data transmission integrity.

[0089] Multi-source synchronous acquisition unit 2 receives the node mapping relationship table output by ring main unit topology modeling unit 1, uses infrared thermal imaging acquisition device to acquire thermal image data inside the ring main unit based on the node mapping relationship table and maps it to the corresponding nodes according to structural partitions, collects electrical parameter characteristics of each node location, realizes node-level deep fusion of infrared spatial information and electrical time-series data, and sends the node-level deep fusion data of infrared spatial information and electrical time-series data and the real-time operating status data of the ring main unit to dynamic attention PHM diagnostic unit 3;

[0090] In this embodiment, the multi-source synchronous acquisition unit 2 includes a data receiving module 21, an infrared data acquisition and mapping module 22, an electrical parameter acquisition module 23, a node-level data fusion module 24, and a data sending module 25, wherein:

[0091] Data receiving module 21 receives and stores the node mapping relationship table output by ring network cabinet topology modeling unit 1;

[0092] Specifically, the data receiving module 21 is used to receive and store the node mapping relationship table output by the ring network cabinet topology modeling unit 1 to ensure that the data is complete and usable. The specific implementation is as follows:

[0093] The data receiving module 21 establishes a communication link using the same industrial communication protocol (Modbus TCP protocol or IEC61850 protocol) as the mapping table generation and sending module 14, and receives the node mapping table and CRC32 checksum sent by the ring main unit topology modeling unit 1. The data receiving module 21 first compares the received data with the checksum. After confirming that there are no transmission errors, it stores the node mapping table in XML format in a local cache unit and simultaneously establishes an index directory, marking the association between node identifiers and their corresponding physical locations and connection characteristics, facilitating quick lookup and retrieval by the infrared data acquisition mapping module 22 and the electrical parameter acquisition module 23. If the checksum comparison is inconsistent, the data receiving module 21 sends a reception failure signal to the ring main unit topology modeling unit 1, triggering a retransmission mechanism until the received data passes verification.

[0094] Based on the node mapping relationship table, the infrared data acquisition and mapping module 22 acquires the internal thermal image data of the ring network cabinet through the infrared thermal image acquisition device, and maps the thermal image data to the corresponding nodes according to the structural partition.

[0095] Specifically, the infrared data acquisition and mapping module 22, based on the node mapping relationship table stored in the data receiving module 21, completes the acquisition of thermal image data inside the ring network cabinet and accurate node mapping, as implemented below:

[0096] The infrared data acquisition and mapping module 22 establishes a signal connection with the infrared thermal imaging acquisition device. The infrared thermal imaging acquisition device adopts a non-contact installation method and is fixed in an unobstructed position inside the ring network cabinet to ensure coverage of all key node areas. The acquisition parameters are set based on the operating temperature range of the ring network cabinet (temperature range -20℃~150℃, acquisition frame rate 1~5 frames / second to meet real-time monitoring requirements).

[0097] During the data acquisition process, the infrared data acquisition and mapping module 22 calls the node mapping relationship table stored in the data receiving module 21 to extract the physical location coordinates and structural partition information of each node. It then splits the thermal image output by the infrared thermal imaging acquisition device according to structural partitions. Through a coordinate matching algorithm, it maps the temperature data (taking the average temperature of the region as the node's thermal image data) of the corresponding node's physical location area in the thermal image to the corresponding node. Simultaneously, it records the thermal image data acquisition timestamp (synchronized with the system clock, with millisecond-level accuracy). After mapping, the infrared data acquisition and mapping module 22 stores the node identifier, corresponding thermal image temperature data, and acquisition timestamp in association, providing infrared data support for the node-level data fusion module 24.

[0098] The electrical parameter acquisition module 23 acquires the electrical parameter characteristics of each node location and synchronously records the parameter acquisition timestamp;

[0099] Specifically, the electrical parameter acquisition module 23 is used to collect the electrical parameter characteristics of each node location and synchronously record the parameter acquisition timestamp to ensure that the electrical data and infrared data are time-aligned. The specific implementation is as follows:

[0100] Electrical parameter acquisition module 23 is configured with corresponding acquisition elements according to the node type:

[0101] For high-current nodes such as bus joints and circuit breaker contacts, Rogowski coil current sensors are configured.

[0102] For voltage monitoring nodes, voltage divider voltage sensors are configured, and all data acquisition components adopt a passive design to avoid affecting the normal operation of the ring main unit.

[0103] The parameters collected include voltage (kV), current (A), and power factor at the node. The acquisition frequency is kept consistent with the acquisition frame rate of the infrared data acquisition and mapping module 22 to ensure data synchronization.

[0104] During the data acquisition process, the electrical parameter acquisition module 23 calls the node mapping table of the data receiving module 21, and acquires the electrical parameters of the corresponding positions one by one according to the node identifier. It synchronously records the acquisition timestamp with the same accuracy as the infrared data. At the same time, it performs preliminary filtering on the acquired electrical parameters (using the moving average filtering method to eliminate instantaneous interference signals), and stores the processed electrical parameters, node identifier, and acquisition timestamp together, and sends them to the node-level data fusion module 24.

[0105] The node-level data fusion module 24 performs deep node-level fusion of infrared spatial information and electrical time-series data, and associates the thermal image data, electrical parameter characteristics and acquisition timestamps of the corresponding nodes.

[0106] Specifically, the core function of the node-level data fusion module 24 is to perform deep node-level fusion of infrared spatial information and electrical time-series data, realizing the association and binding of multi-source data on the same node. The specific implementation is as follows:

[0107] Data fusion reception and preprocessing: The node-level data fusion module 24 establishes communication links with the infrared data acquisition and mapping module 22 and the electrical parameter acquisition module 23, respectively, and synchronously receives two sets of data: one is the combined data of "node identifier-infrared temperature value-acquisition timestamp" sent by the infrared data acquisition and mapping module 22; the other is the combined data of "node identifier-voltage / current / power factor-acquisition timestamp" sent by the electrical parameter acquisition module 23 (with moving average filtering already completed). After receiving, the node-level data fusion module 24 first performs format verification on the two sets of data, discarding data with missing fields or incorrect identifiers, and retaining valid data with compliant formats.

[0108] Precise timestamp alignment: The node-level data fusion module 24 uses the system's unified clock as a reference (accuracy to the millisecond level) and employs a "time window matching + linear interpolation" algorithm to align the timestamps of two sets of data. A time deviation threshold of ±10 milliseconds is set. For infrared and electrical data from the same node, time matching is performed: if the timestamp difference is within the threshold, the corresponding data is directly associated; if the difference exceeds the threshold but is within a reasonable acquisition interval (≤ the time interval corresponding to the acquisition frame rate), the missing data is supplemented using linear interpolation to ensure time synchronization; if the difference exceeds a reasonable acquisition interval, the data set is marked as "time out of sync and invalid," stored separately in the abnormal data cache, and not involved in subsequent fusion. Simultaneously, the out-of-sync node identifier and timestamp information are recorded for subsequent troubleshooting of acquisition synchronization issues.

[0109] Node-level data association and fusion: Using the unique node identifier as the core index, the node-level data fusion module 24 constructs an independent structured data entry for each valid node. Core fields include: unique node identifier, infrared temperature value (°C), voltage value (kV), current value (A), power factor, infrared data acquisition timestamp, electrical data acquisition timestamp, fusion completion timestamp, and data validity markers (valid / invalid due to time synchronization failure / invalid due to missing fields). During the fusion process, a strict "one-to-one" association rule is followed, meaning one node identifier corresponds to one set of infrared data and one set of electrical data, ensuring accurate binding of multi-source data for a single node and avoiding cross-node data confusion.

[0110] Data integrity verification and storage: After fusion, the node-level data fusion module 24 performs an integrity check on the fused data of each node to confirm that there are no missing core fields or data conflicts (such as multiple sets of infrared / electrical data appearing on the same node at the same time). The fused data that passes the verification is stored in a local temporary database in JSON format, categorized by node identifier, and indexed for quick retrieval by the data sending module 25; the data that fails the verification is marked with the reason for the anomaly (such as missing fields or data conflicts) and stored in the anomaly cache along with invalid data that is out of sync with time. It is retained for 72 hours for later troubleshooting, and automatically cleaned up after the expiration period to release storage space.

[0111] The data transmission module 25 synchronously sends the node-level deep fusion data of infrared spatial information and electrical timing data, as well as the real-time operating status data of the ring main unit, to the dynamic attention PHM diagnostic unit 3.

[0112] Specifically, the data sending module 25 is used to synchronously send node-level deep fusion data and ring network cabinet real-time operating status data to the Dynamic Attention PHM diagnostic unit 3, and the specific implementation is as follows:

[0113] The data sending module 25 first integrates the node-level comprehensive dataset output by the node-level data fusion module 24, and at the same time collects the real-time operating status data of the ring network cabinet (including cabinet ventilation status, data acquisition device working status, power supply circuit on / off status, which are obtained by summarizing the feedback signals from each module).

[0114] The data transmission module 25 adopts the same industrial communication protocol as the front end, encapsulates the fused data and operating status data into standardized data frames, and adds a CRC32 check code to ensure complete transmission.

[0115] During transmission, the data transmission module 25 monitors the transmission status in real time. If it does not receive a reception feedback signal from the Dynamic Attention PHM diagnostic unit 3, it automatically triggers a retransmission mechanism (retransmission interval of 1 second, maximum retransmission 3 times). If retransmission fails, it marks the fault status and stores the corresponding data, which will be retransmitted after communication is restored, ensuring that the Dynamic Attention PHM diagnostic unit 3 can receive the detection data completely.

[0116] The Dynamic Attention PHM Diagnostic Unit 3 receives node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit 2, as well as real-time operating status data of the ring main unit. It incorporates a dynamic attention mechanism adapted to the power operation scenarios of the ring main unit. This mechanism integrates a heat conduction physical model reflecting the internal heat conduction patterns of the ring main unit, quantifies the heat conduction attenuation coefficient between nodes based on the heat conduction physical model, and automatically adjusts the weights of infrared temperature features and electrical features based on the real-time operating status of the ring main unit. Under high load conditions, it focuses on analyzing changes in electrical features; under low load conditions, it focuses on identifying subtle gradients in the infrared temperature distribution. Based on the heat conduction attenuation coefficient, it distinguishes between the heat conduction effect and the temperature changes of the fault source, sending the fault location analysis results to the fault location output unit 4. Simultaneously, it feeds back the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenarios of the ring main unit to the multi-source synchronous acquisition unit 2. The Dynamic Attention PHM Diagnostic Unit 3 includes a data receiving and preprocessing module 31, an attention mechanism integration module 32, a weight adjustment and fault differentiation module 33, and a data interaction module 34, wherein:

[0117] In this embodiment, the data receiving and preprocessing module 31 receives the node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit 2, as well as the real-time operating status data of the ring main unit, and performs format standardization and outlier removal processing on the data to generate a standardized node dataset.

[0118] Specifically, the core function of the data receiving and preprocessing module 31 is to receive the data output from the multi-source synchronous acquisition unit 2 and perform standardization processing to provide a high-quality dataset for subsequent diagnosis. The specific implementation process is as follows:

[0119] Data reception: The data reception preprocessing module 31 adopts the same industrial communication protocol (Modbus TCP / IEC 61850) as the data transmission module 25 of the multi-source synchronous acquisition unit 2, and receives two types of data: one is node-level deep fusion data of infrared spatial information and electrical timing data (including node identifiers and infrared temperature values). ,Voltage Current Power factor Collection timestamp ); secondly, real-time operating status data of the ring main unit (including cabinet ventilation status). Operating status of the data acquisition device Power supply circuit on / off status Real-time load power Upon receipt, the integrity of the data is confirmed by comparing the CRC32 checksum. If the verification is successful, the data enters the preprocessing process.

[0120] Format standardization processing: The data receiving preprocessing module 31 converts the fused data into a unified structured format, defining the standardized dataset fields as follows: .in, This indicates the unique identifier of the node (consistent with the node coding rules of ring network cabinet topology modeling unit 1).

[0121] Outlier removal: using "3" The combined algorithm of "criteria + logical consistency check" removes abnormal data:

[0122] Numerical outlier removal: for Calculate the sample mean using continuous parameters. and standard deviation Remove excess Data within a range;

[0123] Logical anomaly removal: Validation of " hour , " Logical rules such as "marking corresponding data as invalid" are used to eliminate logically conflicting data.

[0124] Standardized node dataset generation: The processed valid data is categorized and stored according to node identifier IDs to generate a standardized node dataset. ( (Total number of nodes), providing input data for the attention mechanism integration module 32.

[0125] In this embodiment, the attention mechanism integration module 32 incorporates a dynamic attention mechanism adapted to the power operation scenario of the ring main unit, integrating a physical model reflecting the internal heat conduction law of the ring main unit into the dynamic attention mechanism adapted to the power operation scenario of the ring main unit; the attention mechanism integration module 32 includes a model selection and configuration submodule, a logic embedding submodule, and a mapping relationship construction submodule, wherein:

[0126] The model selection and configuration submodule adopts the Fourier steady-state heat conduction model as the physical model of heat conduction that reflects the heat conduction law inside the ring main unit. The core parameter value range of the model is preset to adapt to the heat conduction characteristics of the closed space of the ring main unit.

[0127] Specifically, the core function of the model selection and configuration submodule is to determine the heat conduction physical model and parameter range suitable for ring main unit scenarios, which is implemented as follows:

[0128] Model selection criteria: The ring main unit is a closed metal cabinet. The heat conduction of the key internal nodes (copper busbar and aluminum busbar connection points) is mainly by solid conduction, without forced convection. This meets the applicable conditions of the Fourier steady-state heat conduction model. Therefore, this model was selected as the physical model for heat conduction.

[0129] The core formula of the model: The formula for calculating heat flow in the Fourier steady-state heat conduction model is as follows:

[0130] ;

[0131] in:

[0132] Represents a node With nodes Heat flow between them (unit: W);

[0133] Represents a node With nodes Thermal conductivity of the conductors connecting them (unit: W / ( The determination depends on the material of the connection point (e.g., copper). =386W / ( Aluminum material =209W / ( Copper-aluminum transition parts =298W / ( ));

[0134] Represents a node With nodes The contact area (unit: m²) is taken from the node mapping relationship table of ring network cabinet topology modeling unit 1;

[0135] Represents a node Infrared temperature values ​​(unit: °C) are taken from the standardized node dataset. ;

[0136] Represents a node Infrared temperature values ​​(unit: °C) are taken from the standardized node dataset. ;

[0137] Represents a node With nodes The physical distance (unit: m) is taken from the node mapping relationship table of ring network cabinet topology modeling unit 1 (converted from mm to m).

[0138] Model parameter adaptation:

[0139] Based on the actual operating environment of the ring main unit, the preset parameter value range is as follows:

[0140] thermal conductivity 200~400W / ( (Covered with copper, aluminum, and transition materials);

[0141] Contact area : (Corresponding to the actual structural dimensions of the connection points);

[0142] physical distance : 0.01~0.5m (corresponding to the spacing range of internal nodes of the ring main unit).

[0143] The logic embedding submodule embeds the node heat conduction calculation logic of the Fourier steady-state heat conduction model into the weight allocation layer of the dynamic attention mechanism adapted to the power operation scenario of the ring network cabinet, so that the calculation process of the heat conduction physical model is deeply coupled with the attention weight iteration process.

[0144] Specifically, the logic embedding submodule is used to embed the heat conduction calculation logic into the weight allocation layer of the dynamic attention mechanism, achieving deep coupling between the physical model and the attention mechanism. The specific implementation is as follows:

[0145] Dynamic attention mechanism infrastructure:

[0146] An improved architecture based on a self-attention mechanism is adopted, and the basic attention weight calculation formula is as follows:

[0147] ;

[0148] in:

[0149] These are the query matrix, key matrix, and value matrix (composed of a standardized node dataset). (obtained through linear transformation)

[0150] The weight matrix is ​​a linear transformation matrix;

[0151] for The dimensions of the matrix are used for scaling to avoid gradient vanishing.

[0152] Thermal conduction logic embedding method:

[0153] Introduce a thermal conduction attenuation coefficient in the attention weight allocation layer. (Derived from the Fourier steady-state heat conduction model), the formula for calculating attention weights is modified as follows:

[0154] ;

[0155] in:

[0156] Thermal conductivity attenuation coefficient The resulting diagonal matrix directly links the attention weight of each node pair to the correlation with heat conduction.

[0157] Embedded logic: The higher the correlation of heat conduction ( The closer to 1), the greater the attention weight of the corresponding node pair, ensuring that the diagnostic process prioritizes the relationship between the fault source node and the affected node.

[0158] Coupling verification: After embedding, the weight iteration process of the attention mechanism is performed synchronously with the heat conduction calculation process. Each weight update must be based on the latest... Value adjustment achieves a dual driving force of "data characteristics + physical laws".

[0159] The mapping relationship construction submodule establishes a one-to-one mapping relationship between node heat conduction correlation degree and attention weight based on the node heat conduction calculation results.

[0160] Specifically, the core function of the mapping relationship construction submodule is to establish a one-to-one mapping between node heat conduction correlation and attention weights, ensuring that the physical model results are effectively transformed into the basis for adjusting the attention mechanism. The specific implementation is as follows:

[0161] Definition of nodal heat conduction correlation: heat flow based on the Fourier steady-state heat conduction model Define nodes With nodes thermal conductivity correlation :

[0162] ;

[0163] in:

[0164] The maximum heat flux (in W) of all node pairs within the ring main unit is given by the standardized node dataset. All Calculated;

[0165] The closer the value is to 1, the more likely it is to be a node. With nodes The stronger the correlation with thermal conduction.

[0166] One-to-one mapping formula:

[0167] Establish With attention weight Linear mapping relationship:

[0168] ;

[0169] in:

[0170] For nodes With nodes Attention weights (dimensionless). ;

[0171] The mapping coefficient (with a value of 1.0 to ensure linear and distortion-free mapping);

[0172] This is the offset coefficient (value 0.0, to ensure...). hour );

[0173] Mapping constraints: (For each node) All associated nodes The sum of the weights is 1, which meets the weight normalization requirement of the attention mechanism.

[0174] Mapping update: Whenever the normalized node dataset is updated Update (new data collection), synchronous recalculation , and update This ensures that the mapping relationship adapts to changes in the operating status of the ring main unit in real time.

[0175] In this embodiment, the weight adjustment and fault differentiation module 33 quantifies the inter-node heat conduction attenuation coefficient based on the heat conduction physical model. The weighting coefficient of infrared temperature features is automatically adjusted based on the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient Under high load conditions, the focus is on analyzing changes in electrical characteristics; under low load conditions, the focus is on identifying subtle gradients in the infrared temperature distribution, based on the thermal conductivity attenuation coefficient. Distinguish between heat conduction effects and temperature changes at the fault source to generate fault location analysis results; where:

[0176] The weight adjustment and fault differentiation module 33 uses the thermal conduction attenuation coefficient. The process of distinguishing between heat conduction effects and fault source temperature changes includes the following steps:

[0177] S33.1. Based on the Fourier steady-state heat conduction model and combined with the physical attribute parameters of the nodes in the ring main unit topology model, the node... With nodes Thermal conduction attenuation coefficient between ;

[0178] Specifically, based on the Fourier steady-state heat conduction model and the physical property parameters of the nodes, the nodes are quantified. With nodes thermal conductivity attenuation coefficient The specific implementation is as follows:

[0179] Definition of attenuation coefficient: thermal conductivity attenuation coefficient Indicates heat flow from the node Passed to node The degree of attenuation during the process, with a value range of [0,1]. The closer to 1, the smaller the decay; the closer to 0, the greater the decay.

[0180] The quantification formula is: ;

[0181] in:

[0182] Thermal conductivity of the connecting conductor (unit: W / ( ), and in the Fourier steady-state heat conduction model Consistent;

[0183] Represents a node With nodes The contact area (unit: m²) is taken from the node mapping relationship table of ring network cabinet topology modeling unit 1;

[0184] Represents the thermal conductivity resistance coefficient (unit: ( ( / W), determined according to the type of connection point structure (e.g., bolted connection). ( ) / W, plug-in connection ( ) / W);

[0185] Represents a node With nodes The physical distance (unit: m) is taken from the node mapping relationship table of ring network cabinet topology modeling unit 1.

[0186] Based on the above, this implementation provides the following computational logic example: Given nodes With nodes For copper busbar bolt connection, , , , ,but: This result indicates that the node exhibits minimal attenuation of heat conduction and high heat transfer efficiency.

[0187] S33.2, Preset thermal conductivity attenuation coefficient threshold The thermal conductivity attenuation coefficient obtained from S33.1 With threshold By comparing the results, the cause of the node temperature change is determined, the interference of heat conduction effect on fault determination is shielded, and a judgment basis is provided for subsequent feature weight adjustment.

[0188] Specifically, by setting a threshold and The comparison distinguishes between heat conduction effects and temperature changes at the fault source, specifically implemented as follows:

[0189] threshold Method for determining:

[0190] Based on statistical analysis of heat transfer data during normal operation of ring main units, data from 100 similar ring main units under rated load were selected to calculate the heat transfer data for all node pairs. The value is taken as the 95th percentile. ,Right now: ;in For normal operation of the node With nodes The thermal conductivity attenuation coefficient, statistically, The value range is 0.85-0.90 (suitable for most ring main units in normal heat conduction scenarios).

[0191] Comparison logic and judgment rules:

[0192] like : Decision node Temperature changes are mainly caused by nodes The heat conduction caused the temperature rise (not a fault cause) and it is marked as "conductive temperature rise". The fault determination trigger for this node pair is blocked.

[0193] like : Decision node Temperature changes and nodes The thermal conductivity correlation is relatively weak. If the temperature exceeds the normal range (normal operating temperature range ℃), it is marked as "suspected fault-type temperature rise" and enters the subsequent feature weight adjustment process.

[0194] Special case: If node Its own temperature is outside the normal range, and there are multiple [issues]. Nodes satisfy And if the temperature rises synchronously, then the node is determined. As a potential source of failure, the node For the nodes that transmit influence.

[0195] Result storage: Store the results of each node pair The comparison results and temperature rise type labels are associated and stored to provide a basis for the weight adjustment of S33.3.

[0196] The weight adjustment and fault differentiation module 33 adjusts the infrared temperature feature weight coefficient in conjunction with the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient The process includes the following steps:

[0197] S33.3, Thermal conductivity attenuation coefficient obtained based on S33.1 Based on the ratio of real-time load power to rated load power of the ring main unit, the infrared temperature characteristic weighting coefficient is determined. With electrical characteristic weighting coefficient The allocation ratio, and the infrared temperature feature weighting coefficient With electrical characteristic weighting coefficient satisfy ;

[0198] Specifically, based on Determine the ratio to load power. and The allocation ratio ( The specific implementation is as follows:

[0199] Load power ratio definition: Calculates the real-time load power of the ring main unit. With rated load power ratio :

[0200] in:

[0201] This is the load power ratio (dimensionless). ;

[0202] The rated load power of the ring main unit (unit: kW) is taken from the nameplate parameters of the ring main unit.

[0203] The formula for weight allocation ratio is:

[0204] Combination Statistical characteristics (all node pairs) The mean kˉ), defined The calculation formula is as follows:

[0205] ;

[0206] in:

[0207] The infrared temperature characteristic weighting coefficient (dimensionless). ;

[0208] This is the mean (dimensionless) thermal conductivity attenuation coefficient for all nodes. ( (Total number of nodes)

[0209] Derivation logic: The larger the value, the stronger the thermal conduction interference. The smaller the value (reduces the weight of temperature features and avoids misjudgment); The larger (the higher the load). The smaller the value (reducing the weight of temperature characteristics and emphasizing electrical characteristics), the better it meets the requirements for adapting to different operating conditions.

[0210] Weight constraint verification: by make sure This satisfies the feature weight normalization requirement.

[0211] S33.4. Based on the real-time load power of the ring main unit, three operating conditions are divided: high load, low load, and intermediate load. Based on the weight allocation ratio determined in S33.3, the infrared temperature characteristic weight coefficient is dynamically adjusted. With electrical characteristic weighting coefficient The values ​​are adapted to the feature analysis needs under different working conditions.

[0212] Specifically, three operating conditions are classified, and the proportional formula based on S33.3 is dynamically adjusted. and The specific implementation is as follows:

[0213] Operating condition classification standards:

[0214] Based on load power ratio The operating conditions are divided as follows:

[0215] Low load conditions: (Real-time load ≤ 30% of rated load);

[0216] Intermediate load condition: (Real-time load 30%~70% of rated load);

[0217] High load conditions: Real-time load ≥ 70% of rated load.

[0218] In addition, this implementation also provides the following weight adjustment examples under different operating conditions (in conjunction with...) Formula): Assuming a ring main unit , (Normal heat conduction level):

[0219] Low load conditions ( ): Under low load, electrical parameters fluctuate little, but thermal conduction interference is relatively prominent. The focus is on infrared temperature characteristics (higher ωT) to identify subtle temperature gradients.

[0220] Intermediate load condition ( ): It balances temperature and electrical characteristics, with a balanced weighting distribution.

[0221] High load conditions ): Under high load, fluctuations in electrical parameters (current, voltage) are highly correlated with faults, with a focus on electrical characteristics. (Higher), analyze changes in electrical parameters.

[0222] Weight update mechanism: Whenever the multi-source synchronous acquisition unit 2 outputs a new weight update... Data, synchronous calculation , ,renew and This ensures that the weights adapt to changes in operating conditions in real time.

[0223] Fault location analysis results generation: based on the adjusted and The comprehensive value of fault characteristics of the computing node :

[0224] ;

[0225] in:

[0226] For nodes The comprehensive value of fault characteristics (dimensionless);

[0227] , , These represent the average temperature, average current, and average voltage of all nodes, respectively.

[0228] , , These are the standard deviations of temperature, current, and voltage at all nodes, respectively.

[0229] Judgment rule: If ( The fault threshold is taken as the normal operating threshold. If the 99th percentile (approximately 3.0) is used to determine the node... For the faulty node, generate fault location analysis results (including the faulty node). Comprehensive value of fault characteristics , , , (Temperature rise type).

[0230] In this embodiment, the data interaction module 34 sends the fault location analysis results to the fault location output unit 4, and simultaneously feeds back the feature weight adjustment parameters of the dynamic attention mechanism adapted to the ring main unit power operation scenario to the multi-source synchronous acquisition unit 2. The data interaction module 34 includes a fault result structured processing submodule, an encrypted transmission submodule, a weight parameter organization submodule, and an acquisition optimization instruction generation submodule, wherein:

[0231] The fault result structured processing submodule receives the fault location analysis results generated by the weight adjustment and fault differentiation module 33, and extracts the fault node number, fault type, and corresponding heat conduction attenuation coefficient. Infrared temperature feature weighting coefficient and electrical characteristic weighting coefficient , Generate structured fault information according to a preset format;

[0232] Specifically, the steps for generating structured fault information are as follows:

[0233] Fault Information Extraction: Receive the fault location analysis results generated by the weight adjustment and fault differentiation module 33, and extract the core information: fault node number. Fault type (based on Composition determination: temperature-dominant / electricity-dominant / hybrid), thermal conductivity attenuation coefficient (faulty node and associated nodes) ), , Fault occurrence timestamp .

[0234] Structured format definition: Generate structured fault information :

[0235] ;

[0236] in Fault type (1=temperature-dominant, 2=electrically-dominant, 3=mixed). This indicates that the fault is valid.

[0237] The encrypted transmission submodule uses a dedicated encrypted communication protocol for power systems to send structured fault information to the fault location output unit 4;

[0238] Specifically, the IEC62351 encryption communication protocol (specifically for power systems, ensuring data transmission security) is used to transmit structured fault information. Encapsulated as an encrypted data frame:

[0239] Data frame format: Frame header (2 bytes) + (32 bytes) + encryption checksum (4 bytes) + frame trailer (2 bytes);

[0240] Transmission mechanism: The transmission adopts the "instant transmission + confirmation retransmission" mode. After transmission, it waits for the reception confirmation signal from the fault location output unit 4. If no confirmation is received, it retransmits once every 2 seconds, up to a maximum of 5 times, to ensure reliable transmission of fault information.

[0241] The weight parameter organization submodule summarizes the feature weight adjustment parameters (infrared temperature feature weight coefficient) for the dynamic attention mechanism adapted to the power operation scenario of ring main unit. Electrical characteristic weighting coefficient Thermal conductivity attenuation coefficient (Work condition determination results), invalid parameters were removed and the format was standardized;

[0242] Specifically, the steps for parameter aggregation, filtering, and standardization are as follows:

[0243] Parameter Summary: Summary of core parameters output by the weight adjustment and fault differentiation module 33: All node pairs Working condition determination results (1 = low load, 2 = medium load, 3 = high load).

[0244] Parameter filtering and standardization: Eliminating invalid parameters (such as...) (Outliers exceeding the range [0,1]) are converted to JSON format, with field names consistent with the standardized node dataset, ensuring that the multi-source synchronous acquisition unit 2 can be parsed.

[0245] The data acquisition optimization instruction generation submodule generates data acquisition optimization instructions based on the processed parameters, according to the infrared temperature feature weighting coefficient. Electrical characteristic weighting coefficient and thermal conductivity attenuation coefficient Based on the numerical conditions, the infrared thermal imaging frame rate, electrical parameter acquisition accuracy, and regional data acquisition intensity of the multi-source synchronous acquisition unit 2 are adjusted respectively.

[0246] Specifically, based on the processed parameters, data acquisition optimization instructions are generated to adjust the acquisition strategy of multi-source synchronous acquisition unit 2. The specific adjustment rules are as follows:

[0247] Infrared thermal imaging frame rate adjustment:

[0248] Set the base frame rate Frames per second, the adjustment formula is as follows:

[0249] ;

[0250] The higher the temperature (the more important the temperature feature), the higher the acquisition frame rate, up to a maximum of 5 frames / second;

[0251] Adjustment of electrical parameter acquisition accuracy:

[0252] Set basic accuracy (Relative error), the adjustment formula is as follows:

[0253] ;

[0254] The higher the value (the more important the electrical characteristics), the higher the acquisition accuracy, with a minimum of 0.3%.

[0255] Regional data collection intensity adjustment: For nodes in their respective areas, increase the data acquisition intensity (increase the acquisition frequency by 50%), focusing on monitoring areas related to heat conduction; for faulty nodes... In the area in question, the data collection frequency has been increased by 100%, enabling fault tracking and monitoring.

[0256] Command transmission: Encapsulate the adjusted frame rate, precision, and intensity parameters into acquisition optimization commands. The data is sent to the data receiving module 21 of the multi-source synchronous acquisition unit 2 via the ModbusTCP protocol, triggering an update of the acquisition strategy.

[0257] Fault location output unit 4 receives the fault location analysis results output by dynamic attention PHM diagnostic unit 3, outputs specific connection point-level fault location information and maintenance scope, and simultaneously feeds back the fault node location information to ring network cabinet topology modeling unit 1.

[0258] In this embodiment, the fault location output unit 4 includes a fault result receiving and parsing module 41, a fault information output module 42, and a feedback information processing module 43, wherein:

[0259] The fault result receiving and parsing module 41 receives the fault location analysis results output by the dynamic attention PHM diagnostic unit 3, extracts the fault node number, physical coordinates and associated topology path information, and completes information parsing and processing.

[0260] Specifically, the fault location analysis results output by the Dynamic Attention PHM diagnostic unit 3 are received via a power system standard encrypted communication protocol (such as IEC62351), and CRC32 data integrity verification is performed simultaneously to ensure that there are no errors or tampering during transmission. After successful verification, core data such as fault node number, physical coordinates, associated topology path, fault type, and fault occurrence timestamp are accurately extracted and organized into a structured dataset according to the logic of "fault node - associated information - diagnostic parameters". During the data organization process, the node coding rules of the ring network cabinet topology modeling unit 1 are strictly followed to ensure that the fault node number and the node mapping relationship of the topology model are consistent, and invalid and logically conflicting data are eliminated.

[0261] Based on the parsed information, the fault information output module 42 generates specific connection point-level fault location information and maintenance scope, and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format.

[0262] Specifically, based on the parsed structured data, precise fault location information at the connection point level is generated, clearly defining the connection type (e.g., busbar joint, circuit breaker contact), installation area (e.g., incoming line area within the cabinet), and physical location description (including specific coordinates and relative position description) of the fault node. The maintenance scope is defined by combining the associated topology path, clearly marking the core fault node and related nodes affected by the fault propagation, providing maintenance personnel with a clear direction for troubleshooting. Simultaneously, it adapts to two types of maintenance scenarios for outputting information: outputting visualized fault information (including highlighted fault locations) and audible and visual alarm signals to the local terminal of the ring network cabinet (e.g., embedded touchscreen); and pushing standardized text reports (e.g., JSON format) to remote maintenance ports, ensuring that both on-site and remote maintenance personnel can quickly obtain effective information.

[0263] The feedback information processing module 43 performs format standardization processing on the fault node location information and feeds back the standardized fault node location information to the ring network cabinet topology modeling unit 1, providing data support for updating the ring network cabinet topology model.

[0264] Specifically, core location information such as fault node number, physical coordinates, and fault type is extracted and standardized according to the model data format (e.g., XML) of ring main unit topology modeling unit 1 to ensure that the data fields are consistent with the node mapping table of ring main unit topology modeling unit 1. The standardized information is fed back to ring main unit topology modeling unit 1 via an adapted industrial communication protocol (e.g., ModbusTCP, IEC60870-5-101). Data verification is performed before transmission; if no confirmation signal is received, a limited retransmission mechanism is initiated. The feedback data will be directly used for fault status marking and dynamic updating of the topology model, providing historical fault data support for subsequent node connection optimization and parameter calculation, ensuring continuous improvement in system diagnostic accuracy.

[0265] like Figure 3 As shown, this embodiment also provides a control method for live-line detection of ring main units that integrates infrared sensing and PHM diagnostics. Based on the above-mentioned live-line detection system for ring main units that integrates infrared sensing and PHM diagnostics, the method includes the following steps:

[0266] S1. Topology modeling and node mapping table generation: Obtain the physical structure parameters of the ring main unit, screen the key electrical connection points of the ring main unit and define them as nodes independently, determine the connection weights based on the physical distance and contact characteristics between nodes, generate the node mapping table and send it to the data acquisition stage.

[0267] S2. Multi-source data synchronous acquisition and node-level deep fusion: Receive and store the node mapping relationship table, acquire the internal thermal image data of the ring main unit through the infrared thermal imaging acquisition device based on the node mapping relationship table, map it to the corresponding node according to the structural partition, collect the electrical parameter characteristics of each node location and synchronously record the parameter acquisition timestamp, perform node-level deep fusion of infrared spatial information and electrical time series data, and synchronously send the node-level deep fusion data and the real-time operating status data of the ring main unit to the diagnostic stage.

[0268] S3. Dynamic Attention PHM Diagnosis and Feature Weight Adjustment Parameter Feedback: Receives node-level deep fusion data and real-time operating status data of the ring main unit. After format standardization and outlier removal, a standardized node dataset is generated. The Fourier steady-state heat conduction model is integrated into a dynamic attention mechanism adapted to the power operation scenario of the ring main unit. The heat conduction attenuation coefficient between nodes is quantified. The operating conditions are divided according to the real-time operating status of the ring main unit, and the weights of infrared temperature features and electrical features are dynamically adjusted. Under high load conditions, the focus is on analyzing changes in electrical features, while under low load conditions, the focus is on identifying subtle gradients in infrared temperature distribution. Based on the heat conduction attenuation coefficient, the temperature changes of heat conduction effect and fault source are distinguished. Fault location analysis results are generated and sent to the fault location output stage. At the same time, the feature weight adjustment parameters are summarized to generate data acquisition optimization instructions and fed back to the data acquisition stage.

[0269] S4. Fault Location Output and Fault Node Location Information Feedback: Receives fault location analysis results, extracts fault node number, physical coordinates and associated topology path information and parses and organizes them, generates specific connection point-level fault location information and maintenance scope and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format, and feeds back the fault node location information to the topology modeling stage after standardizing the format.

[0270] Those skilled in the art will understand that the process of implementing all or part of the steps of the above embodiments can be carried out by hardware or by a program instructing the relevant hardware.

[0271] 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 preferred examples and are not intended to limit 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 present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A live-line detection system for ring main units integrating infrared sensing and PHM diagnostics, characterized in that, include: The ring network cabinet topology modeling unit (1) transforms the physical structure of the ring network cabinet into a topology model containing node physical location parameters and connection characteristic parameters. Each key electrical connection point is defined as a node, and the connection weight between nodes is determined based on physical distance and contact characteristics. A node mapping relationship table is generated and sent to the multi-source synchronous acquisition unit (2). Multi-source synchronous acquisition unit (2) receives the node mapping relationship table output by the ring network cabinet topology modeling unit (1), uses an infrared thermal imaging acquisition device to acquire thermal image data inside the ring network cabinet based on the node mapping relationship table and maps it to the corresponding nodes according to the structural partition, collects the electrical parameter characteristics of each node position, realizes node-level deep fusion of infrared spatial information and electrical time-series data, and sends the node-level deep fusion data of infrared spatial information and electrical time-series data and the real-time operating status data of the ring network cabinet to the dynamic attention PHM diagnostic unit (3). The Dynamic Attention PHM Diagnostic Unit (3) receives the node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit (2) and the real-time operating status data of the ring main unit. It has a built-in dynamic attention mechanism adapted to the power operation scenario of the ring main unit. It integrates the heat conduction physical model reflecting the internal heat conduction law of the ring main unit into the dynamic attention mechanism adapted to the power operation scenario of the ring main unit. Based on the heat conduction physical model, it quantifies the heat conduction attenuation coefficient between nodes. It automatically adjusts the weight of infrared temperature characteristics and electrical characteristics in combination with the real-time operating status of the ring main unit. Under high load conditions, it focuses on analyzing changes in electrical characteristics. Under low load conditions, it focuses on identifying subtle gradients in infrared temperature distribution. Based on the heat conduction attenuation coefficient, it distinguishes the temperature changes of heat conduction effect and fault source. It sends the fault location analysis results to the fault location output unit (4) and feeds back the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenario of the ring main unit to the multi-source synchronous acquisition unit (2). The fault location output unit (4) receives the fault location analysis results output by the dynamic attention PHM diagnostic unit (3), outputs the specific connection point level fault location information and maintenance range, and feeds back the fault node location information to the ring network cabinet topology modeling unit (1).

2. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 1, characterized in that, The ring network cabinet topology modeling unit (1) includes a topology structure parsing module (11), a node independent definition module (12), a connection weight calculation module (13), and a mapping table generation and sending module (14), wherein: The topology parsing module (11) obtains the physical structure parameters of the ring network cabinet and transforms the physical structure of the ring network cabinet into a topology graph model containing node physical location parameters and connection characteristic parameters. The node independent definition module (12) filters the key electrical connection points of the ring network cabinet from the topology model and defines each key electrical connection point as a node independently. The connection weight calculation module (13) calculates and determines the connection weight between nodes based on the physical distance and contact characteristics between nodes; The mapping table generation and sending module (14) generates a node mapping relationship table based on node information and the connection weight between nodes, and sends the node mapping relationship table to the multi-source synchronous acquisition unit (2).

3. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 2, characterized in that, The multi-source synchronous acquisition unit (2) includes a data receiving module (21), an infrared data acquisition and mapping module (22), an electrical parameter acquisition module (23), a node-level data fusion module (24), and a data sending module (25), wherein: The data receiving module (21) receives and stores the node mapping relationship table output by the ring network cabinet topology modeling unit (1); The infrared data acquisition and mapping module (22) acquires thermal image data inside the ring network cabinet through the infrared thermal image acquisition device based on the node mapping relationship table, and maps the thermal image data to the corresponding nodes according to the structural partition. The electrical parameter acquisition module (23) acquires the electrical parameter characteristics of each node location and synchronously records the parameter acquisition timestamp; The node-level data fusion module (24) performs deep node-level fusion of infrared spatial information and electrical time-series data, and associates the thermal image data, electrical parameter characteristics and acquisition timestamps of the corresponding nodes. The data transmission module (25) synchronously sends the node-level deep fusion data of infrared spatial information and electrical timing data, as well as the real-time operating status data of the ring main unit, to the dynamic attention PHM diagnostic unit (3).

4. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 3, characterized in that, The dynamic attention PHM diagnostic unit (3) includes a data receiving and preprocessing module (31), an attention mechanism integration module (32), a weight adjustment and fault differentiation module (33), and a data interaction module (34), wherein: The data receiving and preprocessing module (31) receives the node-level deep fusion data of infrared spatial information and electrical timing data output by the multi-source synchronous acquisition unit (2) and the real-time operating status data of the ring network cabinet, performs format standardization and outlier removal on the data, and generates a standardized node dataset. The attention mechanism integration module (32) has a built-in dynamic attention mechanism adapted to the power operation scenario of the ring main unit, which integrates the heat conduction physical model reflecting the internal heat conduction law of the ring main unit into the dynamic attention mechanism adapted to the power operation scenario of the ring main unit. The weight adjustment and fault differentiation module (33) quantifies the inter-node heat conduction attenuation coefficient based on the heat conduction physical model. The weighting coefficient of infrared temperature features is automatically adjusted based on the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient Under high load conditions, the focus is on analyzing changes in electrical characteristics; under low load conditions, the focus is on identifying subtle gradients in the infrared temperature distribution, based on the thermal conductivity attenuation coefficient. Distinguish between the heat conduction effect and the temperature change of the fault source to generate fault location analysis results; The data interaction module (34) sends the fault location analysis results to the fault location output unit (4), and at the same time feeds back the feature weight adjustment parameters of the dynamic attention mechanism adapted to the power operation scenario of the ring network cabinet to the multi-source synchronous acquisition unit (2).

5. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 4, characterized in that, The attention mechanism integration module (32) includes a model selection and configuration submodule, a logic embedding submodule, and a mapping relationship construction submodule, wherein: The model selection and configuration submodule adopts the Fourier steady-state heat conduction model as the physical model of heat conduction that reflects the heat conduction law inside the ring main unit, and presets the value range of the core parameters of the model to adapt to the heat conduction characteristics of the closed space of the ring main unit. The logic embedding submodule embeds the node heat conduction calculation logic of the Fourier steady-state heat conduction model into the weight allocation layer of the dynamic attention mechanism adapted to the power operation scenario of the ring network cabinet, so that the calculation process of the heat conduction physical model is deeply coupled with the attention weight iteration process. The mapping relationship construction submodule establishes a one-to-one mapping relationship between node heat conduction correlation degree and attention weight based on the node heat conduction calculation results.

6. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 5, characterized in that, The weight adjustment and fault differentiation module (33) uses the thermal conduction attenuation coefficient The process of distinguishing between heat conduction effects and fault source temperature changes includes the following steps: S33.

1. Based on the Fourier steady-state heat conduction model and combined with the physical attribute parameters of the nodes in the ring main unit topology model, the node... With nodes Thermal conduction attenuation coefficient between ; S33.2, Preset thermal conductivity attenuation coefficient threshold The thermal conductivity attenuation coefficient obtained from S33.1 With threshold By comparing the results, the cause of the node temperature change is determined, the interference of heat conduction effect on fault determination is shielded, and a judgment basis is provided for subsequent feature weight adjustment.

7. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 6, characterized in that, The weight adjustment and fault differentiation module (33) adjusts the infrared temperature feature weight coefficient in conjunction with the real-time operating status of the ring main unit. With electrical characteristic weighting coefficient The process includes the following steps: S33.3, Thermal conductivity attenuation coefficient obtained based on S33.1 Based on the ratio of real-time load power to rated load power of the ring main unit, the infrared temperature characteristic weighting coefficient is determined. With electrical characteristic weighting coefficient The allocation ratio, and the infrared temperature feature weighting coefficient With electrical characteristic weighting coefficient satisfy ; S33.

4. Based on the real-time load power of the ring main unit, three operating conditions are divided: high load, low load, and intermediate load. Based on the weight allocation ratio determined in S33.3, the infrared temperature characteristic weight coefficient is dynamically adjusted. With electrical characteristic weighting coefficient The values ​​are adapted to the feature analysis needs under different working conditions.

8. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 7, characterized in that, The data interaction module (34) includes a fault result structured processing submodule, an encrypted transmission submodule, a weight parameter organization submodule, and a data acquisition optimization instruction generation submodule, wherein: The fault result structured processing submodule receives the fault location analysis results generated by the weight adjustment and fault differentiation module (33), and extracts the fault node number, fault type, and corresponding heat conduction attenuation coefficient. Infrared temperature feature weighting coefficient and electrical characteristic weighting coefficient , Generate structured fault information according to a preset format; The encrypted transmission submodule uses a dedicated encrypted communication protocol for power systems to send structured fault information to the fault location output unit (4). The weight parameter processing submodule summarizes the feature weight adjustment parameters (infrared temperature feature weight coefficient) of the dynamic attention mechanism adapted to the power operation scenario of the ring main unit. Electrical characteristic weighting coefficient Thermal conductivity attenuation coefficient (Work condition determination results), invalid parameters were removed and the format was standardized; The data acquisition optimization instruction generation submodule generates data acquisition optimization instructions based on the processed parameters, according to the infrared temperature feature weighting coefficient. Electrical characteristic weighting coefficient and thermal conductivity attenuation coefficient Based on the numerical conditions, the infrared thermal imaging acquisition frame rate, electrical parameter acquisition accuracy, and regional data acquisition intensity of the multi-source synchronous acquisition unit (2) are adjusted respectively.

9. The ring main unit live-line detection system integrating infrared sensing and PHM diagnostics according to claim 8, characterized in that, The fault location output unit (4) includes a fault result receiving and parsing module (41), a fault information output module (42), and a feedback information processing module (43), wherein: The fault result receiving and parsing module (41) receives the fault location analysis results output by the dynamic attention PHM diagnostic unit (3), extracts the fault node number, physical coordinates and associated topology path information, and completes information parsing and organization; The fault information output module (42) generates specific connection point-level fault location information and maintenance scope based on the parsed information, and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format. The feedback information processing module (43) performs format standardization processing on the fault node location information and feeds back the standardized fault node location information to the ring network cabinet topology modeling unit (1) to provide data support for updating the ring network cabinet topology model.

10. A control method for live-line detection of ring main units integrating infrared sensing and PHM diagnostics, based on the live-line detection system for ring main units integrating infrared sensing and PHM diagnostics as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Topology modeling and node mapping table generation: Obtain the physical structure parameters of the ring main unit, screen the key electrical connection points of the ring main unit and define them as nodes independently, determine the connection weights based on the physical distance and contact characteristics between nodes, generate the node mapping table and send it to the data acquisition stage. S2. Multi-source data synchronous acquisition and node-level deep fusion: Receive and store the node mapping relationship table, acquire the internal thermal image data of the ring main unit through the infrared thermal imaging acquisition device based on the node mapping relationship table, map it to the corresponding node according to the structural partition, collect the electrical parameter characteristics of each node location and synchronously record the parameter acquisition timestamp, perform node-level deep fusion of infrared spatial information and electrical time series data, and synchronously send the node-level deep fusion data and the real-time operating status data of the ring main unit to the diagnostic stage. S3. Dynamic Attention PHM Diagnosis and Feature Weight Adjustment Parameter Feedback: Receives node-level deep fusion data and real-time operating status data of the ring main unit. After format standardization and outlier removal, a standardized node dataset is generated. The Fourier steady-state heat conduction model is integrated into a dynamic attention mechanism adapted to the power operation scenario of the ring main unit. The heat conduction attenuation coefficient between nodes is quantified. The operating conditions are divided according to the real-time operating status of the ring main unit, and the weights of infrared temperature features and electrical features are dynamically adjusted. Under high load conditions, the focus is on analyzing changes in electrical features, while under low load conditions, the focus is on identifying subtle gradients in infrared temperature distribution. Based on the heat conduction attenuation coefficient, the temperature changes of heat conduction effect and fault source are distinguished. Fault location analysis results are generated and sent to the fault location output stage. At the same time, the feature weight adjustment parameters are summarized to generate data acquisition optimization instructions and fed back to the data acquisition stage. S4. Fault Location Output and Fault Node Location Information Feedback: Receives fault location analysis results, extracts fault node number, physical coordinates and associated topology path information and parses and organizes them, generates specific connection point-level fault location information and maintenance scope and outputs it to the local terminal or remote maintenance port of the ring network cabinet in a preset format, and feeds back the fault node location information to the topology modeling stage after standardizing the format.

Images