Power distribution box health state quantification method and system based on digital twinning

Abstract

This invention discloses a method and system for quantifying the health status of distribution boxes based on digital twins, belonging to the field of fault prediction and health management technology. Key technical points include: obtaining an initial digital twin model through hardware sensing and model building; updating the model using a neighbor discovery protocol; obtaining internal hotspot temperature estimates using a state observer based on pole placement; obtaining the posterior probability of faulty nodes using a dynamic Bayesian network; and obtaining a composite risk index based on risk factors and physical boundary constraints. This invention achieves accurate assessment and quantification of the health status of distribution boxes by implementing adaptive topology updates through a neighbor discovery protocol, internal temperature inversion through a state observer based on pole placement, and risk quantification by integrating physical boundary constraints through a dynamic Bayesian network. This improves the accuracy of fault prediction and operational efficiency.

Description

Technical Field

[0001] This invention relates to the field of fault prediction and health management technology, and more specifically to a method and system for quantifying the health status of a distribution box based on digital twins. Background Technology

[0002] As the terminal node of a low-voltage power distribution network, the distribution box plays a crucial role in power distribution, metering, and protection. Its operational reliability directly impacts user power supply safety and electricity experience. With the deepening of power grid digital transformation, digital twin technology has been gradually introduced into the field of power distribution equipment condition monitoring. By constructing a virtual mapping of physical entities, it enables real-time perception and assessment of equipment operating status. However, existing digital twin technologies are mainly geared towards large distribution cabinets or prefabricated substations, typically relying on pre-set fixed topology models, directly measured internal parameters, and purely data-driven fault prediction methods. When applied to distribution boxes, the diverse models and often customized internal configurations based on site requirements make it difficult for fixed topology models to match the dynamic changes in actual connection relationships, hindering the development of digital twin models. The discrepancy between the physical model and the actual physical entity leads to inaccurate quantitative assessment models. Furthermore, the constraints of non-intrusive monitoring prevent direct measurement of internal key hotspot temperatures, such as those of circuit breaker contacts, resulting in a lack of necessary input for the thermal circuit model and insufficient accuracy in indirect estimation of internal hotspot temperatures. This further impacts the accuracy of subsequent health status quantification. Moreover, distribution box faults are often caused by the coupling of multiple factors, including electrical, thermal, and mechanical issues. Historical fault data is sparse, making it difficult for purely data-driven methods to accurately characterize complex fault mechanisms. Purely mechanistic models, on the other hand, are difficult to adapt to actual field conditions due to the difficulty in obtaining parameters. Both factors contribute to significant discrepancies between the fault probability quantification results and the actual situation. These factors combined result in insufficient accuracy in the quantitative assessment of distribution box health status using existing technologies, highlighting the limitations of current technologies. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for quantifying the health status of distribution boxes based on digital twins. This method obtains the current effective topology map based on the connection attributes of components within the distribution box through a neighbor discovery protocol and updates it to obtain a corrected digital twin model. A lumped parameter thermal circuit model is constructed using the corrected digital twin model. Based on a state observer and a dynamic Bayesian network, and according to physical boundary constraints and hierarchical rules, real-time assessment and quantification of the distribution box's health status are achieved. This solves the problems of low assessment accuracy, inaccurate quantification, and poor adaptability in existing technologies due to their inability to adapt to the variability of distribution box topology, limitations of non-intrusive monitoring, and complex fault coupling. This improves the accuracy of fault prediction and health management of distribution boxes, as well as the efficiency of operation and maintenance.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] This invention provides a method for quantifying the health status of a distribution box based on digital twins, comprising:

[0006] Based on the physical entity and factory configuration of the distribution box, an initial digital twin model is obtained by deploying hardware sensors and constructing geometric models, electrical topology templates and thermal circuit model templates;

[0007] Based on the connection attributes of components in the distribution box, the current valid topology map is obtained through the neighbor discovery protocol, and the initial digital twin model is updated accordingly to obtain the corrected digital twin model.

[0008] Acquire the status data of the distribution box, construct a lumped parameter thermal circuit model based on the corrected digital twin model, and obtain the estimated internal hot spot temperature through a state observer based on the pole placement method;

[0009] Based on the internal hotspot temperature estimate and the causal logic of the distribution box fault, the posterior probability of the fault node is obtained through a dynamic Bayesian network. The composite risk index of the distribution box is obtained by defining risk factors and introducing physical boundary constraints for correction.

[0010] As a further improvement of the present invention, the physical entity and factory configuration based on the distribution box are compared with the initial digital twin model by deploying hardware sensors and constructing geometric models, electrical topology templates, and thermal circuit model templates, including:

[0011] Based on the physical entity of the distribution box, a geometric model matching the physical entity is obtained by simplification and feature extraction through 3D modeling.

[0012] Based on the factory configuration of the distribution box, an electrical topology template is obtained by defining component nodes and electrical connection edges using the topology node modeling method.

[0013] Based on the typical thermal characteristics of the distribution box, thermal nodes are divided by the lumped parameter method and the thermal resistance and thermal capacity between the thermal nodes are defined to obtain the thermal circuit model template.

[0014] The initial digital twin model of the distribution box is obtained based on the geometric model, electrical topology template, and thermal circuit model template.

[0015] As a further improvement of the present invention, the step of obtaining the current valid topology map based on the connection attributes of components in the distribution box through a neighbor discovery protocol, and updating the initial digital twin model accordingly to obtain a corrected digital twin model, includes:

[0016] Based on the high-speed power line carrier communication capability between the master control communication module and the slave node module in the distribution box, communication data is obtained through the interaction method of the master control communication module periodically broadcasting neighbor discovery frames and the slave node replying with response frames in random time slots;

[0017] Based on the communication data and the connection attributes of the components in the distribution box, a node adjacency matrix is ​​constructed through a neighbor discovery protocol, and a preliminary topology map is obtained by verifying the node connectivity using a connected component algorithm.

[0018] Based on the preliminary topology map and the electrical topology template, the topology difference is determined by calculating the graph editing distance, and abnormal data is removed to obtain the current valid topology map;

[0019] Based on the current valid topology diagram, the corrected digital twin model is obtained by adding or deleting nodes and correcting connection relationships in the electrical topology template and thermal circuit model template of the initial digital twin model.

[0020] As a further improvement of the present invention, a lumped parameter thermal circuit model is constructed based on the modified digital twin model, and the internal hot spot temperature estimate is obtained through a state observer based on the pole placement method, including:

[0021] Based on the modified digital twin model, the heat transfer relationship between nodes is derived through the heat flow conservation law and Kirchhoff's laws, resulting in the state-space equation of the lumped parameter thermal circuit model.

[0022] Based on the state-space equations, the observer gain matrix is ​​designed using the pole placement method to obtain the state observer.

[0023] Based on the status data of the distribution box, the internal hot spot temperature estimate is obtained by discretization recursive calculation through the status observer.

[0024] As a further improvement of the present invention, based on the modified digital twin model, the heat transfer relationship between nodes is derived through the heat flow conservation law and Kirchhoff's laws, resulting in the state-space equations of the lumped-parameter thermal circuit model, including:

[0025] Based on the current valid topology in the modified digital twin model, the order of the lumped parameter thermal path model is determined by identifying heat source nodes and combining the total number of nodes.

[0026] Based on the order of the lumped parameter thermal circuit model, a thermal network topology is constructed by connecting nodes sequentially with thermal resistance and thermal capacity.

[0027] Based on the aforementioned thermal network topology, heat flow balance equations for each node are established using the heat flow conservation law and Kirchhoff's laws. The state-space equations of the lumped parameter thermal circuit model are then derived by further deducing the heat transfer relationships between nodes.

[0028] As a further improvement of the present invention, the step of obtaining the estimated internal hotspot temperature value by performing discretization recursive calculation through the state observer based on the distribution box status data includes:

[0029] Based on the lumped parameter thermal circuit model, the state observer is discretized by setting a discretization period to obtain the state estimation recursive equation;

[0030] The status data of the distribution box is processed, and the real-time power of the heat source node is calculated according to Joule's law.

[0031] Based on the processed state data and real-time power, the state estimation recursive equation is substituted into the recursive calculation to obtain the preliminary temperature estimate of the internal hotspot.

[0032] Based on the deviation between the preliminary internal hot spot temperature estimate and the measured internal surface temperature of the distribution box, the internal hot spot temperature estimate is obtained by feedback correction of the input parameters of the state estimation recursive equation.

[0033] As a further improvement of the present invention, based on the estimated internal hotspot temperature and the causal logic of the distribution box fault, the posterior probability of the fault node is obtained through a dynamic Bayesian network. A composite risk index for the distribution box is obtained by defining risk factors and introducing physical boundary constraints for correction, including:

[0034] Based on the causal logic of the distribution box fault, a dynamic Bayesian network is obtained by defining evidence nodes, hidden state nodes, fault nodes and constructing causal relationships between nodes. The causal logic of the distribution box fault is pre-constructed by fusing the physical mechanism of the distribution box fault with expert experience according to the modified digital twin model.

[0035] Based on the estimated internal hotspot temperature and the real-time operating data of the distribution box, the posterior probability of the fault node is obtained through the membership function and the forward and backward algorithms.

[0036] Based on the factory configuration of the distribution box, the corrected probability of the fault node is obtained through physical boundary constraint correction.

[0037] Based on the corrected failure node probability, a composite risk index is obtained by defining risk factors and introducing coupling coefficients.

[0038] As a further improvement of the present invention, based on the estimated internal hotspot temperature and real-time operating data of the distribution box, the posterior probability of the fault node is obtained through a membership function and forward and backward algorithms, including:

[0039] Based on the estimated internal hotspot temperature and the real-time operating data of the distribution box, the state probability value is obtained through the membership function and used as the input of the dynamic Bayesian network.

[0040] Based on the dynamic Bayesian network, a Bayesian network inference model is obtained by setting the time window length and introducing the time transition probability of the Markov assumption.

[0041] Based on the evidence input, the forward probability is obtained by traversing the observation data within the time window using a forward algorithm, and the backward probability is obtained by traversing the observation data within the time window using a backward algorithm.

[0042] The posterior probability of the faulty node within the time window is obtained based on the forward and backward probabilities.

[0043] As a further improvement of the present invention, based on the factory configuration of the distribution box, the corrected probability of the fault node is obtained through physical boundary constraint correction, including:

[0044] The physical boundary constraints are determined based on the factory configuration of the distribution box.

[0045] Based on the estimated internal hot spot temperature and the real-time operating data of the distribution box, the physical boundary constraints are compared to determine the out-of-limit physical quantities and their corresponding fault nodes.

[0046] Based on the aforementioned physical quantities exceeding the limits and their corresponding fault nodes, the preliminary corrected fault node probabilities are obtained by setting the prior probability of the fault node to a fixed value and re-performing dynamic Bayesian network inference.

[0047] Based on the initially corrected failure node probability, a second fine-tuning is performed by combining observational evidence of non-exceeding physical quantities to obtain the corrected failure node probability; the non-exceeding physical quantities are the remaining physical quantities in the physical boundary constraints other than the exceeding physical quantities.

[0048] This invention provides a digital twin-based system for quantifying the health status of a distribution box, the digital twin-based system comprising:

[0049] Initial twin module: Based on the physical entity and factory configuration of the distribution box, an initial digital twin model is obtained by deploying hardware sensors and building geometric models, electrical topology templates and thermal circuit model templates;

[0050] Topology correction module: Used to obtain the current valid topology map based on the connection attributes of components in the distribution box through the neighbor discovery protocol, and update the initial digital twin model accordingly to obtain the corrected digital twin model;

[0051] Hotspot estimation module: used to acquire distribution box status data, construct lumped parameter thermal circuit model based on the modified digital twin model, and obtain internal hotspot temperature estimates through a state observer based on pole placement method;

[0052] Risk quantification module: Based on the estimated internal hotspot temperature and the causal logic of the distribution box failure, the module obtains the posterior probability of the failure node through a dynamic Bayesian network, and obtains the composite risk index of the distribution box by defining risk factors and introducing physical boundary constraints.

[0053] This invention constructs an initial digital twin model based on the physical entity and factory configuration of the distribution box, and uses a neighbor discovery protocol for topology adaptive updates. It inverts the internal hotspot temperature using a state observer designed with a lumped parameter thermal path model and pole placement method. Based on a dynamic Bayesian network, it quantifies the composite risk index by fusing fault causal logic and physical boundary constraints, and achieves health status quantification. This enables real-time assessment of the distribution box's health status and provides quantitative support for predictive maintenance. It solves the problems of existing digital twin technologies being unable to adapt to the topology variability, non-intrusive monitoring limitations, and complex fault coupling in distribution boxes, thus improving the accuracy, quantification precision, and scenario adaptability of fault prediction and health management. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the steps of the digital twin-based method for quantifying the health status of a distribution box according to the present invention.

[0055] Figure 2 Flowchart of steps for topology correction of a digital twin model;

[0056] Figure 3 Flowchart of steps to obtain an internal hotspot temperature estimate;

[0057] Figure 4 Flowchart of the steps for calculating the composite risk index;

[0058] Figure 5 This is a schematic diagram of the structure of the digital twin-based distribution box health status quantification system of the present invention. Detailed Implementation

[0059] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof.

[0060] Identical parts are indicated by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "up," and "down" used in the following description refer to directions in the accompanying drawings, while the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific part, respectively.

[0061] The term "and / or" in the following text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0062] like Figure 1 As shown, this invention provides a method for quantifying the health status of a distribution box based on digital twins, comprising:

[0063] Based on the physical entity and factory configuration of the distribution box, an initial digital twin model is obtained by deploying hardware sensors and constructing geometric models, electrical topology templates and thermal circuit model templates;

[0064] Based on the connection attributes of components in the distribution box, the current valid topology map is obtained through the neighbor discovery protocol, and the initial digital twin model is updated accordingly to obtain the corrected digital twin model.

[0065] Acquire the status data of the distribution box, construct a lumped parameter thermal circuit model based on the corrected digital twin model, and obtain the estimated internal hot spot temperature through a state observer based on the pole placement method;

[0066] Based on the internal hotspot temperature estimate and the causal logic of the distribution box failure, the posterior probability of the failure node is obtained through a dynamic Bayesian network. By defining risk factors and introducing physical boundary constraints for correction, the composite risk index of the distribution box is obtained.

[0067] The physical entity of the distribution box is the hardware of the low-voltage power distribution equipment to be monitored, including the box shell, circuit breakers, connectors, busbars, and internal electrical connection structures. The factory configuration of the distribution box is the inherent configuration information set by the manufacturer when the distribution box leaves the factory, including original configuration data such as component list, electrical connection drawings, rated operating parameters, material thermophysical parameters, and mechanical life parameters, which are obtained by retrieving the distribution box's factory technical data, equipment nameplate information, and the manufacturer's supporting database. The distribution box status data is real-time monitoring information reflecting the operating status of the distribution box, including the surface temperature of the shell collected by the surface temperature sensor, the current of each branch collected by the current sensor, the ambient temperature and humidity inside the box collected by the temperature and humidity sensor, and the door status collected by the door magnetic switch, which are obtained through sensor deployment and data acquisition.

[0068] This embodiment achieves digital mapping and dynamic updating of the physical entity of the distribution box by constructing a digital twin model. It uses a lumped-parameter thermal circuit model and a state observer based on the pole placement method to accurately estimate the internal hotspot temperature of the distribution box. It uses a dynamic Bayesian network combined with physical boundary constraints to achieve accurate inference and correction of fault node probabilities. By defining risk factors, calculating a composite risk index, and performing graded early warnings, it realizes real-time assessment and quantification of the distribution box's health status, improving the accuracy and timeliness of distribution box fault prediction and enabling predictive maintenance. This solves the problems of low assessment accuracy and poor adaptability of existing digital twin technologies due to their inability to adapt to the topological variability, non-intrusive monitoring limitations, and complex fault coupling of distribution boxes. It effectively reduces the failure rate and maintenance costs of distribution boxes, improves the accuracy of fault quantification and maintenance efficiency, and enhances the stability and reliability of the power distribution system.

[0069] Furthermore, this embodiment provides a step-by-step approach to obtain an initial digital twin model based on the physical entity and factory configuration of a distribution box, by deploying hardware sensors and constructing geometric models, electrical topology templates, and thermal circuit model templates. This includes:

[0070] Based on the physical entity of the distribution box, a geometric model matching the physical entity is obtained by simplification and feature extraction through 3D modeling.

[0071] Based on the factory configuration of the distribution box, an electrical topology template is obtained by defining component nodes and electrical connection edges using the topology node modeling method.

[0072] Based on the typical thermal characteristics of the distribution box, thermal nodes are divided by the lumped parameter method and the thermal resistance and thermal capacity between thermal nodes are defined to obtain the thermal circuit model template.

[0073] The initial digital twin model of the distribution box is obtained based on the geometric model, electrical topology template, and thermal circuit model template.

[0074] Among them, 3D modeling is a technical means of digitally modeling the physical entity of the distribution box based on 3D modeling software, used to transform the physical structure of the distribution box into a digital 3D model; structural simplification and feature extraction is a processing method for simplifying the complex structure of the distribution box physical entity, eliminating redundant structures, and extracting core structural features; matching the geometric model of the physical entity is a 3D digital model consistent with the core structural features of the distribution box physical entity; topology node modeling is a topology modeling method that abstracts electrical components in the distribution box as nodes and electrical connection relationships as edges, used to construct a topology model representing the electrical connection relationships of the distribution box; component nodes are used to... The electrical components in the distribution box, such as circuit breakers, connectors, and busbar connection points, are abstracted using topological node modeling. Electrical connection edges, representing the electrical connections between components within the distribution box, are abstracted using the same method. The electrical topology template is a topology model representing the factory-preset electrical connections of the distribution box. Typical thermal characteristics of the distribution box are the thermal properties of its materials and the heat transfer path characteristics of its internal structure, obtained by retrieving the factory material testing report and analyzing the heat transfer path in conjunction with the structural design drawings. The lumped parameter method decomposes the continuous thermal system into... A thermal analysis method is used to construct a thermal circuit model characterizing the heat transfer characteristics of a distribution box by discretizing the distribution box's thermal system into several thermal nodes with concentrated thermal resistance and heat capacity, and analyzing the heat transfer relationships between these nodes. Each thermal node is a thermal analysis unit with concentrated thermal resistance and heat capacity obtained by discretizing the distribution box's thermal system using the lumped parameter method. These nodes include internal heat source nodes, internal air nodes, inner wall nodes, outer shell surface nodes, and external environment nodes. The thermal resistance and heat capacity between the thermal nodes are thermal resistance parameters characterizing the heat transfer resistance between different thermal nodes and heat capacity parameters characterizing the heat storage capacity of each thermal node. Their initial values ​​are obtained by referencing the thermal conductivity of the distribution box material and calibrating them using simulations under typical operating conditions. The thermal circuit model template is a thermal model characterizing the heat transfer characteristics of a distribution box, constructed using the lumped parameter method based on the typical thermal characteristics of the distribution box. The initial digital twin model of the distribution box is a digital basic model of the distribution box constructed by integrating the geometric model, electrical topology template, thermal circuit model template, and hardware sensor configuration information of the matching physical entity. The hardware sensor configuration information is a collection of basic parameters and layout information of various sensing devices and communication modules deployed in the distribution box. It is obtained by recording, collecting, and organizing the equipment selection parameters, installation location coordinates, communication network relationships, and binding relationships between equipment unique identifiers and box IDs during the hardware sensor deployment process.

[0075] Specifically, firstly, based on the physical entity of the distribution box, computer-aided design software is used to read the external dimensions and internal layout drawings from the distribution box product manual. A digital 3D model is then constructed, including the box shell, the mounting bases for each circuit breaker, the busbar routing, and the locations of connectors. This digital 3D model is simplified by removing small chamfers, fillets, and non-critical decorative structures. Key geometric features such as the outer contour of the box shell, the coordinates of component installation locations, and the geometric paths of the busbars are extracted to obtain a geometric model matching the physical entity. Then, based on the distribution box's factory configuration, the component list and electrical connection drawings from the distribution box's factory technical documentation are retrieved. A topology node modeling method is used to connect the components listed in the component list... Each circuit breaker, connector, and busbar connection point is defined as a node in graph theory. Connection lines representing electrical connections in the electrical connection diagram are defined as edges connecting the nodes, constructing an initial electrical connection diagram. Each node in the initial electrical connection diagram is assigned factory attributes such as component type, rated current, and rated voltage. Each edge in the initial electrical connection diagram is assigned connection attributes such as connection type and conductor cross-sectional area, resulting in an electrical topology template. Then, based on the typical thermal characteristics of the distribution box, the thermal conductivity, specific heat capacity, and density of the box's outer shell material, as well as the thermal properties of the internal air, are consulted from the material handbook. The heat transfer path is analyzed according to heat transfer principles, and the heat transfer path is divided into internal heat source nodes using the lumped parameter method. Five types of thermal nodes are defined: internal heat source nodes, internal air nodes, internal wall nodes of the enclosure, external surface nodes of the enclosure, and external environment nodes. The thermal resistance between internal heat source nodes and internal air nodes, internal air nodes and internal wall nodes of the enclosure, internal wall nodes and external surface nodes of the enclosure, and external environment nodes of the enclosure are defined. The heat capacity of each thermal node is defined. The heat capacity of internal heat source nodes is calculated based on the material and volume of the corresponding component. The heat capacity of internal air nodes is calculated based on the volume of air inside the enclosure and the thermal properties of the air. The heat capacity of internal wall nodes and external surface nodes of the enclosure is calculated based on the material and volume of the corresponding region. This results in a thermal path model template. Finally, the geometric model, electrical topology template, thermal circuit model template, and sensor layout coordinate information and communication node address information recorded during hardware deployment are integrated to match the physical entity. The spatial position association between the component nodes in the geometric model and the electrical topology template is established, the mapping relationship between the component nodes in the electrical topology template and the internal heat source nodes in the thermal circuit model template is established, and the correspondence between the surface nodes of the enclosure in the thermal circuit model template and the installation position of the surface temperature sensor is established. A composite data structure containing geometric information, electrical connection information, heat transfer information, and hardware configuration information is generated. This composite data structure is stored on the cloud digital twin platform and associated with the corresponding box ID to obtain the initial digital twin model of the distribution box.

[0076] For example, this embodiment assumes that the distribution box is a single-phase four-digit energy metering box. Its physical entity includes a box shell, four circuit breakers, and a busbar. The component list in the factory configuration includes four miniature circuit breakers with a rated current of 63A, and the electrical connection drawing shows that the four circuit breakers are connected in parallel to the busbar. First, based on this physical entity, computer-aided design software is used to create a 3D model according to the external dimensions and internal layout diagram in the product manual. A 3D model including the box shell, the mounting bases of the four circuit breakers, and the busbar is constructed. Non-critical features such as the chamfer of the mounting screw holes are removed from the model, and the outer contour dimensions of the box shell and the coordinates of the installation positions of the four circuit breakers (denoted as positions) are extracted. Based on the geometric path of the busbar, a geometric model matching the physical entity is obtained; then, based on the factory configuration of the distribution box, the component list and electrical connection drawings are retrieved, and the four circuit breakers are defined as nodes using the topology node modeling method. Define the busbar as a node ,Will and , and , and , and The electrical connections between them are defined as edges Construct an initial electrical connection diagram and assign nodes to By assigning factory-defined attributes such as a rated current of 63A and a rated voltage of 230V, an electrical topology template is obtained. Then, based on the typical thermal characteristics of a distribution box, the thermal conductivity of PC plastic is consulted. Specific heat capacity ,density thermal conductivity of air Specific heat capacity ,density The thermal nodes are delineated using the lumped parameter method: the four circuit breaker contacts are defined as internal heat source nodes. The air inside the chamber is defined as the internal air node. Define the inner wall of the box as an inner wall node. The sensor mounting positions on the inner wall of the enclosure are defined as nodes on the enclosure surface. Define the external environment as an environment node. Furthermore, through corresponding calculations, we obtain... and thermal resistance between for , and thermal resistance between for , and thermal resistance between for , and thermal resistance between for , and thermal resistance between for , and thermal resistance between for , and thermal resistance between for ;definition heat capacity for , heat capacity for , heat capacity for , heat capacity for , heat capacity for , heat capacity for , heat capacity for The thermal circuit model template is obtained; finally, the geometric model, electrical topology template, thermal circuit model template, and sensor layout coordinate information and communication node address information recorded during hardware deployment are integrated to establish the location of the circuit breaker in the geometric model. Nodes in the electrical topology template , , , Spatial association, establishing nodes in the electrical topology template , , , Internal heat source nodes in the thermal path model template The mapping relationship is established to create a correspondence between the outer shell surface node S and the installation position of the surface temperature sensor in the thermal circuit model template. A composite data structure is generated and stored in the cloud digital twin platform to obtain the initial digital twin model of the single-phase four-digit energy metering box. In this embodiment, the specific values ​​of thermal resistance and thermal capacity are merely examples. Those skilled in the art can obtain these values ​​through simulation or experimentation based on the actual distribution box's materials, dimensions, and structure. This embodiment does not impose any limitations on these values.

[0077] This embodiment constructs a geometric model matching the physical entity through 3D modeling, builds an electrical topology template based on the factory configuration through topology node modeling, and builds a thermal circuit model template based on typical thermal characteristics through lumped parameter method. By integrating the geometric model, electrical topology template, thermal circuit model template, and hardware configuration information, an initial digital twin model is generated. This achieves a complete mapping and integrated construction from the physical entity of the distribution box to the digital space, improves the matching degree between the initial digital twin model and the physical entity, and enhances the model's refinement. This lays a solid digital foundation for the subsequent dynamic updates of the digital twin model and the health status assessment of the distribution box, while also improving the efficiency and adaptability of model construction.

[0078] Furthermore, this embodiment provides a step of obtaining a current valid topology map based on the connection attributes of components within the distribution box through a neighbor discovery protocol, and updating the initial digital twin model accordingly to obtain a corrected digital twin model, including:

[0079] Based on the high-speed power line carrier communication capability between the master control communication module and the slave node module in the distribution box, communication data is obtained through the interaction method of the master control communication module periodically broadcasting neighbor discovery frames and the slave node replying with response frames in random time slots;

[0080] Based on communication data and the connection attributes of components in the distribution box, a node adjacency matrix is ​​constructed through a neighbor discovery protocol, and a preliminary topology map is obtained by verifying node connectivity using a connected component algorithm.

[0081] Based on the preliminary topology map and the electrical topology template, the topology difference is determined by calculating the distance between the graph and the editing distance. After removing abnormal data, the current valid topology map is obtained.

[0082] Based on the current valid topology diagram, the corrected digital twin model is obtained by adding or deleting nodes and correcting connection relationships in the electrical topology template and thermal circuit model template of the initial digital twin model.

[0083] The main control communication module is a core control unit deployed at the main control position in the distribution box, integrating high-speed power line carrier communication and edge computing capabilities. It is obtained by installing it inside the distribution box and storing a unique box ID. The slave node module is a miniature communication unit deployed at each circuit breaker, connector, and busbar connection point, utilizing existing auxiliary contacts of the components or adding RFID tags. It is obtained by installing and assigning a unique identifier. The high-speed power line carrier communication capability is the ability to transmit data using existing power lines as the communication medium. The interaction method of the main control communication module periodically broadcasting neighbor discovery frames and the slave nodes replying with response frames in random time slots is that the main control communication module sends a message containing the box ID and serial number at a preset period. The communication mechanism involves a neighbor discovery frame and a response frame from each slave node after a random backoff time, containing its own ID and a received signal strength indication. This mechanism is used to obtain the communication reachability between nodes. The communication data is a collection of response frames from all slave nodes collected by the main control communication module. The component connection attributes within the distribution box represent the electrical connections between components via wires and busbars, obtained from the electrical connection diagram in the distribution box's factory configuration. The neighbor discovery protocol is a node identification and adjacency detection protocol based on power line carrier communication, used to obtain communication reachability and signal strength information between nodes. The node adjacency matrix is ​​a matrix constructed based on the communication data, representing the communication reachability between nodes, where matrix elements... Elements represent whether there is a direct communication link between corresponding nodes; the connected component algorithm is an algorithm in graph theory used to identify the largest connected subgraph in a graph, used to verify whether all nodes belong to the same electrically connected network; node connectivity is the property used by the connected component algorithm to determine whether each node belongs to the same connected component; the preliminary topology graph is a graph structure built based on the node adjacency matrix, representing the communication adjacency relationships between nodes; graph edit distance is a metric for measuring the degree of difference between two graph structures, obtained by calculating the minimum number of node or edge insertion, deletion, and replacement operations required to transform one graph into another; topology difference is the graph edit distance value between the preliminary topology graph and the electrical topology template; outlier data is data present in the preliminary topology graph. The communication data corresponding to isolated nodes or edges with abnormal signal strength; the current valid topology graph is a graph structure that reflects the current actual component connection relationship of the distribution box after removing abnormal data; node addition and deletion refers to the operation of adding or deleting nodes in the electrical topology template of the digital twin model to match the node changes in the current valid topology graph; connection relationship correction refers to the operation of adding, deleting or adjusting the weights of edges in the electrical topology template of the digital twin model to match the connection relationship changes in the current valid topology graph; the corrected digital twin model is a digital twin model that is consistent with the current actual configuration of the distribution box after updating the electrical topology template and thermal circuit model template of the initial digital twin model based on the current valid topology graph.

[0084] Specifically, such as Figure 2As shown, the main control communication module first operates at a preset cycle. A neighbor discovery frame is broadcast, containing the master communication module's own bin ID and an incrementing sequence number. Upon receiving the neighbor discovery frame, each slave node module replies with a response frame within a random time slot. The response frame contains the slave node module's own ID, received signal strength indicator, and a list of detectable neighbor node IDs. The master communication module collects all slave node module response frames and constructs a node adjacency matrix based on its own ID and the list of neighbor node IDs in the response frames. A matrix element of 1 indicates a direct communication link between nodes, and 0 indicates a non-direct link. Then, a connected component algorithm is used to analyze the connectivity of the node adjacency matrix. Specifically, all nodes are initialized to an unvisited state, and a depth-first search traversal is performed starting from any unvisited node, marking all reachable nodes as... For each connected component, this process is repeated until all nodes are visited. If the number of connected components obtained after traversal is greater than 1, an isolated node is determined to exist, triggering a communication anomaly alarm. The set of nodes and edges corresponding to the largest connected subgraph obtained after traversal is used as the initial topology graph. Then, the graph editing distance between the initial topology graph and the electrical topology template is calculated. Specifically, the cost of node insertion, deletion, and replacement operations is defined as 1, and the cost of edge insertion, deletion, and replacement operations is also defined as 1. A graph editing distance algorithm (such as the A* search algorithm) is used to calculate the minimum operation cost required to convert the initial topology graph into the electrical topology template. The graph editing distance is used as a topology difference metric. In the graph editing distance algorithm, the single cost of node insertion, deletion, and replacement operations is defined as... The single-operation cost for simultaneous insertion, deletion, and replacement is defined as follows: The cost is determined based on the principle of equivalent importance of each component and electrical connection in terms of functional integrity in the factory configuration of the distribution box.

[0085] The difference between the image editing distance and the preset threshold If the image editing distance is greater than the preset difference threshold, a comparison will be made. If the on-site electrical configuration has been altered, abnormal data will be removed from the preliminary topology diagram: isolated nodes will be removed, and nodes with signal strength less than or equal to a preset signal strength threshold will be removed. The process involves: 1) connecting nodes to edges and removing single abnormal edges caused by temporary communication interference to obtain the current valid topology graph. 2) Based on the current valid topology graph, the electrical topology template in the initial digital twin model is modified by adding, deleting, and correcting nodes and connections: If a node in the current valid topology graph is not present in the electrical topology template, it is added to the template and assigned default factory attributes; if a node in the electrical topology template is not present in the current valid topology graph, it is removed; if the connecting edges between nodes in the current valid topology graph are inconsistent with those in the electrical topology template, the edge set in the electrical topology template is updated. 3) Simultaneously, based on the number and connection relationships of component nodes in the current valid topology graph, the number and connection relationships of heat source nodes in the thermal circuit model template are automatically adjusted to maintain a one-to-one correspondence between heat source nodes and component nodes, resulting in the corrected digital twin model. The specific rules for automatic adjustment include: adding and deleting heat source nodes, mapping each circuit breaker node or connector node in the current valid topology diagram to a corresponding heat source node; if there are new component nodes in the current valid topology diagram that are not included in the thermal circuit model template, then the same number of heat source nodes are added to the thermal circuit model template, and the newly added heat source nodes are assigned an initial default thermal resistance value. and heat capacity ,in Based on the rated current of the components Calculations show that The thermal resistance connection is estimated based on the component volume. For newly added heat source nodes, thermal resistance connections are established with adjacent thermal nodes (such as internal air nodes) based on their spatial position in the current valid topology. If an existing heat source node is deleted, all its connected thermal resistance connections are simultaneously deleted. The heat capacity parameter is recalculated based on the updated total number of heat source nodes, recalculating the heat capacity value of the internal air nodes. To ensure that the total heat capacity of the thermal circuit model matches the actual heat capacity of the distribution box, the calculation formula is as follows: ,in To adjust the heat capacity of the internal air nodes, To increase the number of heat source nodes, The incremental contribution of each heat source node to the heat capacity of the air inside the chamber, typically taken as a value. .

[0086] Among them, the preset difference threshold The method for determining the value is as follows: based on the initial total number of nodes in the electrical topology template. Determined when the initial total number of nodes is... When less than or equal to 10, the difference threshold Values When the initial total number of nodes When the difference threshold is greater than 10, The value is not less than Divide by The smallest integer quotient. Preset signal strength threshold. The typical value range is obtained by calibrating the receiver sensitivity of the main control communication module and the background noise level of power line carrier communication in the distribution box. When no on-site calibration is performed, the default value is... .

[0087] For example, this embodiment assumes that the distribution box is a single-phase four-digit energy metering box, and the factory-configured electrical topology template includes nodes. Four circuit breaker nodes and busbar nodes ,side This indicates that four circuit breakers are connected in parallel to the busbar. However, during actual installation, only three circuit breakers were installed due to user requirements. The corresponding actual component nodes are as follows: busbar node The electrical connection relationship is and , and , and Connect separately. The main control communication module communicates at a preset interval. A neighbor discovery frame is broadcast every second. The three slave node modules reply with response frames in random time slots. After collecting the response frames, the master control communication module constructs a node adjacency matrix. After verification using the connected component algorithm, all three nodes and the parent node belong to the same connected component, and there are no isolated nodes, thus obtaining a preliminary topology graph. This preliminary topology graph contains nodes and edge Further calculations of the preliminary topology graph With electrical topology template Graph edit distance between: Preliminary topology graph Contains 4 nodes and 3 edges, electrical topology template It contains 5 nodes and 4 edges; the nodes need to be deleted. and edge The image editing distance is calculated to be 2; assuming a preset difference threshold. At this point, the current image edit distance equals the preset difference threshold. The system determined that the on-site configuration had changed; after removing abnormal data, there were no isolated nodes or abnormal edges, thus obtaining the current valid topology graph. The currently valid topology graph Includes nodes and edge Based on the current valid topology graph Adding and deleting nodes in the electrical topology template of the initial digital twin model: removing nodes and corresponding edges , will node and Association, will and Correlate the data; simultaneously adjust the thermal path model template to include the four heat source nodes. Adjusted to three heat source nodes The thermal resistance connection relationship is adjusted accordingly to obtain the corrected digital twin model. When the preset difference threshold in the above assumptions... 2. Preset signal strength threshold When the signal strength of the three slave nodes is -50dBm, and the response signal strengths are -35dBm, -38dBm, and -42dBm respectively, all signal strengths are higher than the threshold, and no abnormal edges are removed. In this embodiment, the preset period, preset difference threshold, and preset signal strength threshold are merely examples. Those skilled in the art can set these according to the actual operating environment of the distribution box, communication quality requirements, and maintenance needs. This embodiment does not impose any limitations on these settings.

[0088] This embodiment acquires communication data through the neighbor discovery protocol between the master control communication module and the slave node module, verifies node connectivity using the connected component algorithm to generate a preliminary topology map, determines topology differences by graph editing distance and removes abnormal data to obtain the current valid topology map, and updates the electrical topology template and thermal circuit model template of the digital twin model by adding, deleting, and modifying node connections. This achieves automatic identification of the actual electrical topology inside the distribution box and dynamic updating of the digital twin model, improving the real-time matching degree between the digital twin model and the physical entity of the distribution box. It solves the problem of fixed topology and inability to adapt to changes in on-site configuration of traditional digital twin models, improves the consistency between the digital twin model and the physical entity, enhances the dynamic adaptability and scenario adaptability of the model, and provides an accurate model foundation for subsequent health status assessment.

[0089] Furthermore, this embodiment provides a step for constructing a lumped-parameter thermal circuit model based on a modified digital twin model, and obtaining internal hotspot temperature estimates through a state observer based on the pole placement method, including:

[0090] Based on the modified digital twin model, the heat transfer relationship between nodes is derived through the heat flow conservation law and Kirchhoff's laws, resulting in the state-space equations of the lumped-parameter thermal circuit model, specifically including:

[0091] Based on the current valid topology in the corrected digital twin model, the order of the lumped parameter thermal circuit model is determined by identifying heat source nodes and combining the total number of nodes. Based on the order of the lumped parameter thermal circuit model, a thermal network topology is constructed by connecting nodes sequentially with thermal resistance and thermal capacity. Based on the thermal network topology, the heat flow balance equations for each node are established by the heat flow conservation law and Kirchhoff's laws. The state space equation of the lumped parameter thermal circuit model is obtained by further deriving the heat transfer relationship between nodes.

[0092] Based on the state-space equations, the observer gain matrix is ​​designed using the pole placement method to obtain the state observer.

[0093] Based on the distribution box status data, the estimated internal hotspot temperature is obtained through discretization and recursive calculation using a status observer, specifically including:

[0094] Based on the lumped-parameter thermal circuit model, the state observer is discretized by setting a discretization period to obtain the state estimation recursive equation. The state data of the distribution box is processed, and the real-time power of the heat source node is calculated according to Joule's law. Based on the processed state data and real-time power, the state estimation recursive equation is substituted into the equation for successive recursive calculations to obtain the preliminary temperature estimate of the internal hot spot. Based on the deviation between the preliminary internal hot spot temperature estimate and the measured temperature of the inner surface of the distribution box, the input parameters of the state estimation recursive equation are corrected through feedback to obtain the final internal hot spot temperature estimate.

[0095] The lumped-parameter thermal circuit model is a thermal analysis model that discretizes the thermal system of the distribution box into several thermal nodes with concentrated thermal resistance and heat capacity. The heat transfer characteristics are characterized by the thermal resistance and heat capacity between nodes. This model is constructed based on a modified digital twin model combined with the law of conservation of heat flux and Kirchhoff's laws. Heat source nodes are the thermal nodes corresponding to the core electrical components that generate Joule heat within the distribution box, including circuit breaker contacts and busbar connection points, identified from the current valid topology diagram of the modified digital twin model. The total number of nodes is the total number of thermal nodes included in the lumped-parameter thermal circuit model, covering heat source nodes, internal air nodes, inner wall nodes of the box, outer surface nodes of the box, and external environment nodes. The order of the lumped-parameter thermal circuit model is also specified. The model dimension parameters, consistent with the total number of nodes in the thermal circuit model, are obtained by identifying heat source nodes in the corrected digital twin model and counting the total number of all thermal nodes. The thermal network topology is the heat transfer network structure formed by connecting the thermal nodes of the lumped parameter thermal circuit model through thermal resistance and thermal capacity, and is constructed based on the thermal characteristics and spatial relationships of each node. The law of conservation of heat flux is the fundamental law characterizing the conservation of energy in a thermal system, stating that the algebraic sum of heat flux entering a certain thermal node is zero, and is used to establish the heat flux balance equation for each thermal node. Kirchhoff's laws are thermal laws applicable to thermal circuit analysis, stating that the distribution of heat flux in thermal circuit nodes follows the same law as the circuit current, and are used to assist in deriving the heat transfer relationship between nodes. The heat flow balance equation is a differential equation established for each thermal node based on the law of conservation of heat flow and Kirchhoff's laws, characterizing the relationship between node temperature changes and heat flow, thermal resistance, and heat capacity. The inter-node heat transfer relationship is a quantitative relationship of heat transfer between thermal nodes through thermal resistance, which is further derived from the heat flow balance equation. The state-space equation of the lumped parameter thermal circuit model is a mathematical model that transforms the heat flow balance equation into a system of first-order linear differential equations, with the temperature of each node as the state variable, the heat source power and ambient temperature as the input variables, and the shell surface temperature as the output variable. The pole placement method is a control theory method for designing the gain matrix of the state observer. By selecting the location of the observer poles, the gain matrix is ​​determined, enabling the observer to estimate... The error convergence mechanism is used to design a state observer for a lumped-parameter thermal circuit model. The observer gain matrix is ​​a matrix parameter designed based on the pole placement method, used to introduce output feedback to correct the estimation error of the state observer, so that the poles of the observer meet the preset dynamic performance requirements. The state observer is a Romberg state observer designed based on the state space equation of the lumped-parameter thermal circuit model and the pole placement method, used to estimate the internal hot spot temperature that cannot be directly measured based on the observable distribution box state data. The discretized recursive operation is an operation method that transforms the continuous-time state observer into a discrete-time form and then performs iterative calculation by successively substituting real-time data, used to recursively estimate the internal hot spot temperature based on the real-time acquired distribution box state data.The discretization period is the time step set when discretizing the continuous-time state observer, determined based on the thermal time constant of the distribution box's thermal system. Discretization involves using numerical integration to transform the continuous-time state observer equations into discrete-time recursive equations, adapting to the discrete acquisition characteristics of real-time data. Joule's law, a physical law characterizing the heat generated by current flowing through a conductor (i.e., the square of the current multiplied by the conductor's resistance equals the thermal power), is used to calculate the real-time Joule thermal power of the distribution box's heat source nodes. The real-time power of each heat source node is the real-time thermal power generated by current flowing through each heat source node within the distribution box, calculated based on Joule's law combined with branch currents collected by current sensors and node contact resistances. The state estimation recursive equation is a discrete-time mathematical equation obtained after discretizing the state observer, used to substitute real-time data and successively derive the temperature estimate for each thermal node. The preliminary temperature of the internal hotspots... The estimated value is the internal hotspot temperature estimate obtained directly by substituting the processed distribution box status data and the real-time power of the heat source nodes into the state estimation recursive equation, without any deviation feedback correction. The measured value of the distribution box's inner surface temperature is the actual temperature data collected by surface temperature sensors deployed on the inner wall of the distribution box, obtained through real-time sensor acquisition and data transmission. The feedback correction uses the deviation between the initial estimated internal hotspot temperature and the measured value of the distribution box's inner surface temperature as a feedback quantity to adjust the correction method of the input parameters of the state estimation recursive equation, thereby reducing the error of the temperature estimate. The input parameters of the state estimation recursive equation are the data substituted into the equation, including the real-time power of the heat source nodes, ambient temperature, measured value of the outer casing surface temperature, and feedback correction quantity. The internal hotspot temperature estimate is the result obtained after feedback correction, and is the final output of the state observer.

[0096] Specifically, such as Figure 3 As shown, firstly, based on the current valid topology in the corrected digital twin model, the heat source nodes corresponding to all circuit breaker contacts and busbar connection points are identified. Combined with statistics on internal air nodes, inner wall nodes of the enclosure, outer surface nodes of the enclosure, and external environment nodes, the total number of thermal nodes is obtained. ,Will The order of the lumped-parameter thermal path model is determined. Based on the spatial location and heat transfer characteristics of each thermal node, the heat source node is sequentially connected to the internal air node, the internal air node to the inner wall node of the enclosure, the inner wall node to the outer surface node of the enclosure, and the outer surface node to the external environment node via thermal resistance. The heat capacity of each thermal node is connected in parallel to the environmental reference point, thus constructing a thermal network topology. Based on this thermal network topology, a heat flow balance differential equation is established for each thermal node according to the law of conservation of heat flow and Kirchhoff's laws. The heat flow balance equations for all nodes are then rearranged and transformed into a vector of the temperature change rate of each thermal node. The derivatives of the state variables and the temperature vectors of each thermal node are given. A vector consisting of system state variables, heat source power, and ambient temperature. Input quantity, temperature vector of the nodes on the surface of the enclosure. The state-space equations for the lumped-parameter thermal circuit model of the output quantity are, i.e. , ,in The system matrix is ​​derived from the thermal resistance matrix. and heat capacity matrix It is determined that the dimension is consistent with the order of the thermal circuit model, has no dimension, and characterizes the inherent heat transfer properties of the thermal system. It will be used uniformly thereafter. Refers to the system matrix; matrix The input matrix represents the degree of influence of the input quantities on the system state; For the output matrix, for Temperature vectors at each thermal node at any given time; for The temperature vector of the nodes on the surface of the outer shell of the enclosure is determined at each instant; then, based on this state-space equation, the observer gain matrix of the state observer is designed using the pole placement method. The pole placement method, taking into account the large thermal inertia of the distribution box, sets the observer poles to be higher than the system matrix. eigenvalue fast Double the negative real part, to ensure The eigenvalues ​​are all negative real parts and the poles take values ​​in the range of 1 / 2. Order of magnitude, where the system matrix The eigenvalues ​​are obtained by solving the characteristic equation. get, To and Identity matrices of the same dimension For the system matrix eigenvalues; the designed gain matrix Substituting into the Luneburg state observer equation, we get Complete the construction of the state observer, where Let be the estimated temperature change rate vector for each thermal node. This represents the estimated temperature vector for each thermal node. The output estimation error is the difference between the measured and estimated values ​​of the outer casing surface temperature; then, the discretization period is set according to the thermal time constant of the distribution box's thermal system. The state observer is discretized using the first-order forward Euler method, and a heat flux attenuation coefficient is introduced. The heat flux accumulation effect in the discretization recursive process is corrected, and the heat flux attenuation coefficient is calculated. Let be a dimensionless constant, whose physical meaning is to characterize the proportion of natural dissipation in the heat flow transfer process per unit time step, and its calculation formula is: ,in It is a natural constant. This represents the minimum value of all thermal time constants in the lumped-parameter thermal circuit model. , and The first The method involves considering the thermal resistance and heat capacity corresponding to each thermal node. A correction term for the heat flux attenuation coefficient is added, resulting in the recursive equation for state estimation. ,in for The estimated temperature vectors of each thermal node at time t. It is an identity matrix with dimensions matching the order of the thermal path model and has no units. for The input vector consists of the real-time power of each heat source node and the external ambient temperature at any given time.

[0097] Simultaneously, the real-time collected distribution box status data undergoes filtering, deduplication, and outlier removal data processing, based on Joule's law. Calculate the real-time power of each heat source node, where Indicates the first Each heat source node is at Real-time Joule thermal power at any given moment This refers to the branch current after data processing. The initial value for the contact resistance of the heat source nodes is taken from the factory design value. Finally, the processed distribution box status data and the calculated real-time power of the heat source nodes are substituted into the state estimation recursive equation, and successive recursive calculations are performed to obtain the temperature estimate of each thermal node. The internal hot spot temperature is extracted as the preliminary internal hot spot temperature estimate, and the temperature deviation between this preliminary estimate and the measured temperature of the inner surface of the distribution box is calculated. Temperature deviation As input parameters to the recursive equation for state estimation corrected by feedback quantity and The corrected input parameters are then substituted back into the recursive equation for iterative calculation. The specific correction method is as follows: the estimated temperature of the nodal surface of the outer shell of the enclosure is calculated from the preliminary temperature estimate of the internal hot spot. Measured value of the inner surface temperature of the distribution box Deviation between Input the ambient temperature value for the next recursive time step. Revised to ,in This is the proportional correction factor, and its value range is... The default value is This yields the internal hotspot temperature estimate after accuracy correction.

[0098] For example, this embodiment assumes that the internal hot spot temperature to be estimated is that of a single-phase three-digit energy metering box after topology correction. The current effective topology of its corrected digital twin model includes three circuit breaker contact heat source nodes. Combining the internal air, the inner wall of the box, the outer surface of the box, and the external environment nodes, the total number of thermal nodes is counted. That is, the order of the lumped parameter thermal circuit model is 7, and the discretization period is set. The observer poles take values ​​of Initial value of contact resistance at heat source node First, based on the corrected digital twin model of the distribution box, the heat source nodes corresponding to the three circuit breaker contacts were identified. Combined with internal air nodes Internal wall nodes of the box Surface nodes of the enclosure External environment nodes The thermal circuit model was determined to be of order 7, and the thermal network topology was constructed according to the heat transfer relationship. respectively with Connected via a thermal resistance of 0.5K / W. and Connected via a thermal resistance of 0.8K / W. and Connected via a thermal resistance of 0.6K / W. and The thermal network is connected with a thermal resistance of 1.0 K / W, and the heat capacity of each node is connected in parallel. Based on this thermal network topology, a heat flow balance equation is established for each node, and the state-space equation of the 7th order lumped parameter thermal circuit model is obtained. The system matrix is... The input matrix is ​​obtained from the calculation of each thermal resistance and thermal capacity. for 3D matrix, output matrix for A 3D matrix is ​​then used; subsequently, based on the state-space equations, the pole placement method is employed to design the observer gain matrix. Set the observer poles to ,make The eigenvalues ​​are all Substituting these values ​​into the equation, the state observer equation is obtained. Then, a discretization period of 10 seconds is set, and the observer is discretized using a first-order forward Euler method with a heat flux attenuation coefficient, yielding the state estimation recursive equation. The collected distribution box state data is filtered to remove abnormal current and temperature jumps. The real-time power of the three heat source nodes is calculated according to Joule's law. If the collected branch currents are 10A, 12A, and 8A, then the calculated power is... Finally, the real-time power, processed surface temperature, and ambient temperature are substituted into the recursive equation to obtain preliminary estimates of the internal hot spot temperatures: 45℃, 48℃, and 42℃. The measured internal surface temperature is 38℃. The temperature deviations between the preliminary estimates and the measured values ​​are calculated to be 7℃, 10℃, and 4℃, respectively. This temperature deviation is used as feedback to correct the input power and surface temperature parameters of the recursive equation. The corrected internal hot spot temperature estimates are then calculated again to be 43℃, 45℃, and 41℃, thus completing the accurate estimation of the internal hot spot temperature. In this embodiment, the thermal circuit model order, discretization period, observer pole values, contact resistance, and other values ​​are merely examples. Those skilled in the art can set these values ​​according to the actual number of components, thermal characteristics, and operating environment of the distribution box. This embodiment does not impose any limitations on these settings.

[0099] This embodiment identifies thermal nodes and constructs a thermal network topology based on a modified digital twin model. It derives the state-space equations of a lumped-parameter thermal circuit model by combining the heat flow conservation law and Kirchhoff's laws. A state observer adapted to the thermal characteristics of the distribution box is designed using the pole placement method. After discretizing the observer, recursive calculations are performed using real-time data and Joule's law, and feedback corrections are used to obtain estimated internal hotspot temperatures. This achieves accurate estimation of internal hotspot temperatures in the distribution box, which cannot be directly measured, overcoming the limitations of non-intrusive monitoring and improving the accuracy and real-time performance of temperature estimation. It provides crucial temperature monitoring data for subsequent identification of faulty nodes and health status assessment of the distribution box, while also improving the model's adaptability to the actual operating state of the distribution box and the reliability of the estimation results.

[0100] Furthermore, this embodiment provides a step-by-step approach to obtain a composite risk index for the distribution box based on internal hotspot temperature estimates combined with causal logic of distribution box faults, obtaining the posterior probability of the faulty node through a dynamic Bayesian network, and defining risk factors and introducing physical boundary constraints for correction. The steps include:

[0101] Based on the causal logic of distribution box failure, a dynamic Bayesian network is obtained by defining evidence nodes, hidden state nodes, and fault nodes and constructing causal relationships between nodes. The causal logic of distribution box failure is pre-constructed by fusing the physical mechanism of distribution box failure with expert experience based on the modified digital twin model.

[0102] Based on the estimated internal hotspot temperature and real-time operating data of the distribution box, the posterior probability of the fault node is obtained through membership functions and forward and backward algorithms, specifically including:

[0103] Based on the internal hotspot temperature estimate and the real-time operation data of the distribution box, the state probability value is obtained through the membership function and used as the input of the dynamic Bayesian network. Based on the dynamic Bayesian network, the Bayesian network inference model is obtained by setting the time window length and introducing the time transition probability of the Markov assumption. Based on the evidence input, the forward probability is obtained by traversing the observation data within the time window through the forward algorithm, and the backward probability is obtained by backtracking the observation data within the time window through the backward algorithm. The posterior probability of the fault node within the time window is obtained based on the forward probability and the backward probability.

[0104] Based on the factory configuration of the distribution box, the corrected probability of fault nodes is obtained through physical boundary constraint correction, specifically including:

[0105] The physical boundary constraints are determined based on the factory configuration of the distribution box. Based on the estimated internal hotspot temperature and real-time operating data of the distribution box, the physical quantities exceeding limits and their corresponding fault nodes are determined by comparing them with the physical boundary constraints. Based on the exceeding physical quantities and their corresponding fault nodes, the prior probabilities of the fault nodes are set to fixed values, and dynamic Bayesian network inference is performed again to obtain a preliminarily corrected probability of the fault nodes. When multiple physical quantities exceed limits simultaneously, the fault nodes that need to be set to fixed values ​​are determined according to the following priority rules: First, the highest priority is the fault node corresponding to the internal hotspot temperature exceeding the limit, which is directly related to safety; second, the second highest priority is the fault node corresponding to the branch current exceeding the limit; finally, the lowest priority is the fault node corresponding to the number of operations exceeding the limit. If multiple physical quantities exceed limits within the same priority, the prior probabilities of all corresponding fault nodes are set to fixed values. And it is updated simultaneously during dynamic Bayesian network inference. If multiple out-of-limit physical quantities point to the same fault node, then only the prior probability of that fault node is set to 0. once;

[0106] Based on the initial corrected failure node probability, the corrected failure node probability is obtained by a second fine-tuning by combining observational evidence of physical quantities that do not exceed the limit.

[0107] Based on the corrected failure node probability, a composite risk index is obtained by defining risk factors and introducing coupling coefficients.

[0108] The fault causal logic of the distribution box is a causal association rule system for fault generation and propagation, which integrates the physical mechanism of distribution box faults with the experience of domain experts and combines the component connection relationships and heat transfer characteristics in the modified digital twin model. Evidence nodes are directly observable input nodes in the dynamic Bayesian network, corresponding to physical quantities that can be collected or estimated during the operation of the distribution box, including estimated internal hotspot temperatures, branch currents, internal humidity, and number of operations. Hidden state nodes are intermediate nodes in the dynamic Bayesian network that are not directly observable but characterize the health status of components, including contact wear, insulation aging, and connection loosening. Fault nodes are output nodes in the dynamic Bayesian network that characterize the final fault mode, including short circuits, overload tripping, and leakage. Electrical issues, poor contact, etc.; causal relationships between nodes are established based on the fault causal logic, representing the directional relationships between evidence nodes, hidden state nodes, and fault nodes, characterizing the impact of physical quantity anomalies on the health state and the driving relationship of health state deterioration on fault occurrence; dynamic Bayesian network is a time-series probabilistic inference model that expands a static Bayesian network along the time axis and introduces time transition probabilities; membership function is a fuzzy mathematical function that maps continuous observation data to discrete state probabilities, used to transform continuous data such as internal hotspot temperature estimates and branch currents into discrete state probability distributions; state probability values ​​are the discrete state probability distribution results obtained after processing by the membership function, serving as input evidence for the dynamic Bayesian network; the time of introducing the Markov assumption. The transition probability is the probability of the same node state at adjacent time points, based on the Markov assumption. That is, the current node state depends only on the state at the previous time point, and is used to characterize the evolution of healthy and faulty states over time. The Bayesian network inference model is a time-series inference model constructed based on a dynamic Bayesian network combined with the time window length and time transition probability. The evidence input is a set of state probability values, including discrete state probabilities transformed from the observation data at each time point within the time window. The forward algorithm is a recursive algorithm that calculates the node state probability distribution time-by-time from the start to the end of the time window. The observation data within the time window consists of continuously collected distribution box state data and transformed state probability values ​​within the set time window length. The forward probability is calculated using the forward algorithm. The calculated probability distribution of node states at each time point represents the probability estimate of the current node state based on historical observation data; the backward algorithm is a recursive algorithm that backtracks from the end time of the time window to the start time to calculate the probability distribution of node states; the backward probability is the probability distribution of node states at each time point calculated by the backward algorithm, representing the probability correction of the current node state based on future observation data; the posterior probability of the fault node is the probability value of the fault node occurring within the time window obtained by combining the forward probability and the backward probability, eliminating the noise influence of single-point observation data; the physical boundary constraints are the insurmountable threshold conditions of physical quantities determined by the factory configuration of the distribution box, including the upper limit of material temperature resistance, the rated current of the circuit breaker, and the upper limit of mechanical life operation times, etc.Physical boundary constraint correction is a correction method that adjusts the posterior probability of fault nodes based on physical boundary constraints. It is used to correct probability estimation bias caused by data noise or inference bias to ensure that the results conform to physical laws. Exceeding physical quantities and their corresponding fault nodes are physical quantities in real-time data that exceed physical boundary constraints, as well as fault nodes directly associated with the anomalies of these physical quantities. The fault node probability after preliminary correction is the fault node probability obtained by re-inferring through dynamic Bayesian network after assigning the prior probability of the fault node corresponding to the exceeding physical quantity. The observational evidence of the physical quantities not exceeding limits is the state probability value obtained by processing the real-time data of the physical quantities not exceeding limits through a membership function; the secondary fine-tuning is an optimization process that combines the observational evidence of the physical quantities not exceeding limits to adjust the weights of the initially corrected fault node probability, used to balance the influence of the physical quantities exceeding limits and normal physical quantities to improve the rationality of the probability estimation; the corrected fault node probability is the final fault node probability obtained after physical boundary constraint correction and secondary fine-tuning, possessing both physical rationality and inference accuracy; the risk factor is defined as a composite parameter used to quantify the risk level of the fault node, including the corrected fault node probability, fault severity, and propagation path length, used to comprehensively characterize the probability of occurrence, severity, and scope of impact of the fault; the coupling coefficient is introduced as an adjustment parameter to enhance the weight of the propagation path length on the risk impact, used to balance the influence ratio of the fault node probability, fault severity, and propagation path length; the composite risk index is a normalized index obtained by weighting the risk factor and coupling coefficient, used to comprehensively assess the overall health risk level of the distribution box.

[0109] Specifically, such as Figure 4 As shown, firstly, based on the modified digital twin model, the physical mechanism of the distribution box fault (such as overheating leading to insulation aging, overload leading to tripping, etc.) and the experience of domain experts are integrated to pre-construct the causal logic of the distribution box fault. Based on this logic, three types of nodes are defined: evidence nodes (whose internal hotspot temperature estimates are...). Branch current Humidity inside the box Number of operations Hidden state nodes (contact wear) Insulation aging Loose connection ), fault node (short circuit) Overload tripping Leakage ), constructing causal relationships between nodes, such as → → , → , → → Unfold the static Bayesian network along the time axis and set the state self-transition probability to be... The state transition probability is State transition probability With state transition probability The determination is based on the mean time between failures (MTBF) of the components in the factory configuration of the distribution box. ,in The time slice step is the dynamic Bayesian network, and MTBF (Mean Time Between Failures) is the average time between failures of the component, which is used to characterize the probability that the component maintains its original healthy state between adjacent time slices. This represents the probability that the state of a component remains unchanged between adjacent time steps; This represents the probability that the component state transitions between adjacent time steps. A dynamic Bayesian network is then obtained. The initial assignment of the conditional probability table for this network uses a hybrid approach of physical thresholds and expert experience; for example, the probability of insulation aging is set to 0.8 when the internal hotspot temperature exceeds 70℃, and to 0.1 under normal temperature. Then, the time window length is set to... A Bayesian network inference model is constructed by introducing the time transition probability based on the Markov assumption. The estimated internal hotspot temperature and real-time operating data of the distribution box are input into the membership function to obtain the state probability value of each discrete state, which is then used as evidence input. The forward probability is calculated time-by-time through the forward algorithm. ,in For the front Observe data in real time. for Time node status, The model parameters for the forward algorithm are used to calculate the backward probability through reverse backtracking using the backward algorithm. In the formula, Indicates the backward probability; Represents conditional probability; From Time reaches the end of the time window A sequence of observation data composed of observation data at different times; Indicates in The node state at time t is ; For the model parameters of the backward algorithm, this formula represents the expression given the model parameters. and The status of the time node is Under the conditions, from arrive Observing data sequences at all times The probability of the faulty node is obtained by combining the forward and backward probabilities. .

[0110] Then, based on the factory configuration of the distribution box, determine the physical boundary constraints: upper limit of material temperature resistance. Circuit breaker rated current Mechanical lifespan and number of operation cycles Estimate the internal hotspot temperature Branch current Number of operations Compared with the constraints, the physical quantities exceeding the limits and their corresponding fault nodes are identified. The prior probability of the fault node is set to 1.0, and it is re-substituted into the dynamic Bayesian network inference to obtain the preliminary corrected probability of the fault node. Then, combined with the observational evidence of the physical quantities not exceeding the limits, the weight coefficients of each piece of evidence are determined using the analytic hierarchy process (AHP). The preliminary corrected probability is then fine-tuned a second time to obtain the corrected probability of the fault node. The specific fusion formula is as follows: ,in This represents the initially corrected probability of faulty nodes. The posterior probability of a faulty node, obtained solely from inference via a dynamic Bayesian network based on physical quantities that do not exceed limits, is fused with weights. Determined based on the degree of exceeding the limits of the physical quantity: when the physical quantity exceeds its physical boundary constraints... When above, ;otherwise Finally, the risk factor is defined based on the adjusted probability of failure nodes. Severity of the fault Length of propagation path Among them, the severity of the fault include Corresponding minor faults, Corresponding medium faults and For the corresponding severe fault, the propagation path length is the average number of edges from the faulty node to the final node corresponding to the fault event, introducing a coupling coefficient. Construct a formula for calculating the composite risk index. ,in This is the node weight, initially set to 1, and can be optimized using historical data. Given the set of all faulty nodes, the risk factors of each faulty node are substituted into the formula to calculate the composite risk index normalized to the [0,1] interval.

[0111] Wherein, coupling coefficient This is a dimensionless constant used to amplify the contribution of fault propagation path length to system risk; its physical meaning is to characterize the intensity of the cumulative risk effect when a fault propagates along a causal chain. Coupling coefficient. The range of values ​​is The voltage level of the distribution box is determined based on its rating: when the rated voltage of the distribution box is not greater than... hour, When the rated voltage is greater than hour, Node weight The optimization algorithm is as follows: Based on a historical fault case database of distribution boxes of the same model, the maximum likelihood estimation method is used to fit and optimize the weight coefficients of each fault node in the composite risk index; in the initial stage without historical data accumulation, the weights of all fault nodes are... All initialized to .

[0112] For example, this embodiment assumes that the device to be evaluated is a single-phase three-digit energy meter box, and the dynamic Bayesian network corresponding to its corrected digital twin model contains 3 evidence nodes (internal hotspot temperature estimates). Branch current Humidity inside the box Number of operations ), 3 hidden state nodes (contact wear) Insulation aging Loose connection ), 3 fault nodes (short circuit) Overload tripping Leakage Set the time window length. That is, including the current time and the previous two time points, the coupling coefficient Node weight Physical boundary constraints: Upper limit of material temperature resistance Circuit breaker rated current Mechanical lifespan and number of operation cycles Next, pre-build the fault cause-effect logic: → → , → , → → Construct a dynamic Bayesian network and set the state self-transition probability to be... State transition probability Conditional probability table Insulation aging The probability is 0.8. Overload trip The probability is 0.7; then, observation data at three time points within the time window are collected, converted into state probability values ​​through the membership function, the time window length is set to 3, and a Bayesian network inference model is constructed. The forward probability at each time point is calculated using the forward algorithm, and the backward probability is obtained by backtracking using the backward algorithm. The preliminary posterior probability of the fault node is obtained by fusion. Then, comparing the physical boundary constraints, at time 3... Greater than the material's upper temperature resistance limit This is an out-of-limit physical quantity, and the corresponding fault node is a short circuit. ,Will The prior probability is set to 1.0, and the preliminary revised probability is obtained through re-reasoning. Based on observational evidence of physical quantities not exceeding limits, the weighting coefficients were determined using the analytic hierarchy process (AHP). The corresponding weight is 0.3. The corresponding weight is 0.2, and the corrected probability is obtained after the second fine-tuning. Finally, the severity of the fault and the propagation path length were defined and substituted into the composite risk index formula to obtain a composite risk index of 0.457. In this embodiment, the values ​​for time window length, coupling coefficient, physical boundary threshold, fault severity, and propagation path length are merely examples. Those skilled in the art can set these values ​​according to the actual factory configuration of the distribution box, the fault hazard level, and maintenance requirements. This embodiment does not impose any limitations on these settings.

[0113] This embodiment constructs a dynamic Bayesian network by integrating the physical mechanism of the fault with expert experience. It uses forward and backward algorithms to achieve fusion inference of time-series data to obtain the posterior probability of the fault node. Physical boundary constraint correction and secondary fine-tuning ensure the physical rationality and accuracy of the probability estimation. By defining risk factors and coupling coefficients to construct a composite risk index, it achieves a quantitative assessment of the overall health risk. This realizes the accurate quantification and comprehensive assessment of the fault risk of the distribution box, improves the scientificity and reliability of the health status assessment, and solves the assessment bias problem caused by relying on only a single data or ignoring physical constraints in traditional assessment methods. It provides accurate risk quantification basis for subsequent graded early warning and enhances the guiding value of the assessment results for operation and maintenance decisions.

[0114] Furthermore, this embodiment provides a step for classifying early warning levels based on a composite risk index and preset grading rules, generating and publishing early warning information including fault causes and suggested measures, including:

[0115] Based on the composite risk index, the early warning level is divided according to the preset classification rules and the corresponding response rules are obtained.

[0116] Based on the inference results of the composite risk index and dynamic Bayesian network, the core fault node is located by tracing the fault causal chain, and preliminary early warning information including the fault cause and suggested measures is obtained.

[0117] Based on the initial early warning information, standardized early warning information is obtained by supplementing and improving it through the correlation-corrected digital twin model;

[0118] Based on the warning level and warning information, the corresponding information push channel is selected through a tiered release mechanism to obtain the warning information release results.

[0119] The pre-defined grading rules are based on the numerical range of the composite risk index to classify warning levels, combined with quantitative rules pre-set by the power distribution system's operation and maintenance needs, the severity of faults, and industry safety standards. Warning levels are grading indicators based on the composite risk index to characterize the health risk level of distribution boxes, including four levels: green for healthy states, yellow for states requiring attention, orange for abnormal states, and red for severe states. Response rules are the operation and maintenance procedures and action requirements set for each warning level, including fault investigation time limits, operation and maintenance priorities, and whether shutdown for maintenance is necessary. Core fault nodes are those that contribute the most to the composite risk index and are the main causes of fault occurrence. By tracing the fault causal chain of a dynamic Bayesian network, the node with the highest posterior probability is selected. The initial warning information is generated based on the composite risk index and the core fault node, containing basic warning elements, including equipment identification, warning level, core fault node, preliminary fault cause, and general recommended measures. The standardized warning information is formed by associating the initial warning information with the corrected digital twin model, supplementing it with fault location visualization information, fault causal chain, customized recommended measures, etc., and includes equipment full identification, warning time, composite risk index, warning level, core fault node, detailed fault cause, three-dimensional coordinates of fault location, customized handling suggestions, expected consequences, etc. The hierarchical release mechanism is a mechanism that selects different information push channels and release priorities according to the severity of the warning level, including three release modes: regular release, key release, and emergency release. The information push channels are the communication channels used to transmit warning information, including cloud monitoring platform pop-ups, mobile APP notifications for maintenance personnel, SMS reminders, emergency voice calls, on-site sound and light alarms, etc. The warning information release result is the feedback result after the warning information is successfully transmitted through the selected push channel, including release time, recipients, read status, processing progress tracking information, etc.

[0120] Specifically, firstly, a preset grading rule is set based on the value range of the composite risk index, including: composite risk index A green alert and a composite risk index are indicated when the risk level is greater than or equal to 0 and less than the preset first risk threshold. A yellow alert and a composite risk index are triggered when the risk level is greater than or equal to a preset first risk threshold and less than a preset second risk threshold. When the risk level is greater than or equal to the preset second risk threshold and less than the preset third risk threshold, an orange alert and a composite risk index are triggered. A red alert is triggered when the risk level is greater than or equal to a preset third risk threshold and less than or equal to 1. The preset third risk threshold is greater than both the preset second and first risk thresholds, determined by combining power distribution system operation and maintenance experience, fault hazard level statistics, industry safety standards, and verification through backtesting of historical fault data. Response rules are configured for each alert level, including: green alerts require no active maintenance, only continuous monitoring; yellow alerts require routine inspections within 72 hours; orange alerts require special investigations within 24 hours; and red alerts require immediate shutdown for maintenance (highest priority). The specific methods for determining the first, second, and third risk thresholds include: firstly, collecting historical fault and normal operation sample data of equipment of the same model as the distribution box, and calculating the composite risk index of each sample. Secondly, plot the cumulative distribution function curve of the fault samples, and set the cumulative distribution function value to... , and Corresponding The values ​​were determined as the baseline values ​​for the first, second, and third risk thresholds, respectively; finally, the operation and maintenance experts adjusted the baseline values ​​according to industry security standards, ensuring that they did not exceed [a certain threshold value]. After fine-tuning, it serves as the final preset risk threshold.

[0121] Then, the specific values ​​of the composite risk index and the inference results of the dynamic Bayesian network are extracted. The core fault node is located using a fault causal chain tracing algorithm, which adds risk contribution weight calculation using the formula... Calculate the risk contribution weight of each failed node, where This represents the corrected probability of the faulty node. The severity of the fault. The coupling coefficient is... To determine the propagation path length, select the top-weighted items. A fault node is designated as the core fault node. Detailed fault causes are obtained through fault causal chain analysis, and general suggested measures are generated by matching with a pre-set expert rule base, forming preliminary early warning information. This preliminary early warning information is then linked to a revised digital twin model. The model extracts the 3D installation coordinates of the components corresponding to the core fault node, the highlighted fault location markers in the geometric model, and the associated node information from the thermal circuit model and electrical topology template, supplementing the preliminary early warning information. Simultaneously, optimized suggested measures are implemented based on the specific configuration of the distribution box, forming standardized early warning information containing text descriptions and 3D visualizations. Finally, a tiered release mechanism is activated based on the warning level: green warnings are only stored and recorded on the cloud monitoring platform and are not actively pushed; yellow warnings are released via cloud monitoring platform pop-ups and mobile app notifications; orange warnings are released via cloud monitoring platform pop-ups, mobile app notifications, and SMS alerts; and red warnings are simultaneously released via cloud monitoring platform pop-ups, mobile app notifications, SMS alerts, emergency voice calls, and on-site audible and visual alarms. Feedback results from each channel are recorded to form a warning information release results ledger, tracking the progress of maintenance and repair.

[0122] For example, this embodiment assumes that the warning information to be issued is for a single-phase three-digit energy meter box, and its composite risk index is... The posterior probabilities of faulty nodes obtained from dynamic Bayesian network inference are: short circuit (0.92), overload tripping (0.13), and leakage current (0.09). Assume the preset first risk threshold, second risk threshold, and third risk threshold are 0.2, 0.5, and 0.8, respectively. First, based on the preset classification rules and the composite risk index... If the risk level is greater than the preset first risk threshold of 0.2 and less than the preset second risk threshold of 0.5, a yellow alert is issued, and the corresponding response rule is to complete a routine inspection within 72 hours. Then, a fault causal chain tracing algorithm is used to calculate the risk contribution weight of each fault node, assuming a coupling coefficient... The weight corresponding to the short circuit is The corresponding weight for overload tripping is: The corresponding weights for leakage current are: Filter out the top with the highest weight Several fault nodes (short circuit, leakage, overload trip) are used as core fault nodes. Based on the fault causal chain analysis, the preliminary fault cause is determined to be excessive internal hotspot temperature leading to insulation aging and a short circuit risk. A general recommended measure is generated by matching the expert rule base to check the circuit load and tighten the wiring terminals, forming a preliminary warning information including device ID, warning level, core fault node, preliminary cause, and general recommendations. Then, the corrected digital twin model is associated with it, the three-dimensional coordinates of the circuit breaker contacts corresponding to the short circuit are extracted, and this location is highlighted in the geometric model. A detailed fault causal chain is then added: internal hotspot... Temperature 125℃ (5℃ over-temperature) → Insulation aging → Short circuit risk. Optimization recommendations include: "Check if the circuit load is overloaded; if overloaded, shunt the current; disassemble the corresponding circuit breaker and tighten the contact terminals; clean the oxide layer on the contact surface and apply conductive grease; complete a re-inspection within 72 hours," forming a standardized early warning message. Finally, based on the yellow warning, a tiered release mechanism is activated, notifying maintenance personnel via a pop-up window on the cloud monitoring platform and their mobile app, recording the release time, and ensuring the maintenance team receives the message. Read confirmation is received after 15 minutes, forming the early warning message release result. In this embodiment, the preset risk threshold, the number of core fault nodes to be screened, and the 72-hour response time limit are merely examples. Those skilled in the art can set these values ​​according to the importance of the power distribution system, the configuration of maintenance resources, and the severity of the fault; this embodiment does not impose such limitations.

[0123] This embodiment achieves accurate classification of composite risk indices and matching of response rules through preset classification rules, locates core fault nodes and generates preliminary early warning information through a fault causal chain tracing algorithm, standardizes and visualizes early warning information through a correlation-corrected digital twin model, and ensures effective information transmission by selecting appropriate push channels through a hierarchical release mechanism. It realizes accurate classification, accurate positioning, standardized presentation and efficient release of early warning information, improves the pertinence and timeliness of operation and maintenance decisions, solves the problem of low operation and maintenance efficiency caused by vague early warning information and single release channels in traditional systems, provides clear and operable execution basis for predictive maintenance, and enhances the safe operation guarantee capability of the power distribution system.

[0124] Furthermore, such as Figure 5 As shown in the figure, this application provides a digital twin-based system for quantifying the health status of a distribution box. The system includes an initial twin module, a topology correction module, a hotspot estimation module, and a risk quantification module.

[0125] Initial twin module: Based on the physical entity and factory configuration of the distribution box, an initial digital twin model is obtained by deploying hardware sensors and building geometric models, electrical topology templates and thermal circuit model templates;

[0126] Topology Correction Module: Based on the connection attributes of components in the distribution box, it obtains the current valid topology map through the neighbor discovery protocol, and updates the initial digital twin model accordingly to obtain the corrected digital twin model;

[0127] Hotspot estimation module: used to acquire distribution box status data, construct lumped parameter thermal circuit model based on the modified digital twin model, and obtain internal hotspot temperature estimates through a state observer based on the pole placement method;

[0128] Risk quantification module: Based on the estimated internal hotspot temperature and the causal logic of the distribution box failure, the module obtains the posterior probability of the failure node through a dynamic Bayesian network, and obtains the composite risk index of the distribution box by defining risk factors and introducing physical boundary constraints.

[0129] The initial twin module, topology correction module, hotspot estimation module, and risk quantification module are all located in the server. The server receives data transmitted from the acquisition devices and performs further analysis. The acquisition devices include surface temperature sensors, Hall current sensors, digital temperature and humidity sensors, magnetic door switches, master control communication modules, and slave node modules. The initial digital twin module is used to complete the initial construction of the digital model from the physical entity of the distribution box, and outputs the initial digital twin model to the topology correction module. The topology correction module is used to dynamically update the digital twin model based on the initial digital twin model and the communication data obtained by the master control communication module and the slave node module through the neighbor discovery protocol. It outputs the corrected digital twin model and synchronizes it to the hotspot estimation module and the risk quantification module, providing a unified model foundation for temperature estimation and health status quantification. The hotspot estimation module is used to receive the corrected digital twin model output by the topology correction module and the distribution box status data collected by the surface temperature sensor, Hall current sensor, and digital temperature and humidity sensor. It obtains the internal hotspot temperature estimate by constructing a lumped parameter thermal circuit model and calculating the internal hotspot temperature using the state observer, and transmits the estimate to the risk quantification module in real time. The risk quantification module is used to receive the internal hotspot temperature estimate output by the hotspot estimation module, combine the fault causal logic of the distribution box and dynamic Bayesian network inference to obtain the fault node probability, and calculate the composite risk index. The modules are connected in sequence and communicate with each other to realize the closed-loop processing of the entire process of the distribution box from digital modeling, dynamic topology updating, hotspot temperature estimation to health risk quantification.

[0130] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0132] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0133] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for quantifying the health status of a distribution box based on digital twins, characterized in that, include: Based on the physical entity and factory configuration of the distribution box, an initial digital twin model is obtained by deploying hardware sensors and constructing geometric models, electrical topology templates and thermal circuit model templates; Based on the connection attributes of components in the distribution box, the current valid topology map is obtained through the neighbor discovery protocol, and the initial digital twin model is updated accordingly to obtain the corrected digital twin model. Acquire the status data of the distribution box, construct a lumped parameter thermal circuit model based on the corrected digital twin model, and obtain the estimated internal hot spot temperature through a state observer based on the pole placement method; Based on the internal hotspot temperature estimate and the causal logic of the distribution box fault, the posterior probability of the fault node is obtained through a dynamic Bayesian network. The composite risk index of the distribution box is obtained by defining risk factors and introducing physical boundary constraints for correction.

2. The method for quantifying the health status of a distribution box based on digital twins according to claim 1, characterized in that, The physical entity and factory configuration based on the distribution box are used to obtain an initial digital twin model by deploying hardware sensors and constructing geometric models, electrical topology templates, and thermal circuit model templates, including: Based on the physical entity of the distribution box, a geometric model matching the physical entity is obtained by simplification and feature extraction through 3D modeling. Based on the factory configuration of the distribution box, an electrical topology template is obtained by defining component nodes and electrical connection edges using the topology node modeling method. Based on the typical thermal characteristics of the distribution box, thermal nodes are divided by the lumped parameter method and the thermal resistance and thermal capacity between the thermal nodes are defined to obtain the thermal circuit model template. The initial digital twin model of the distribution box is obtained based on the geometric model, electrical topology template, and thermal circuit model template.

3. The method for quantifying the health status of a distribution box based on digital twins according to claim 1, characterized in that, The process of obtaining the current valid topology map based on the connection attributes of components within the distribution box through a neighbor discovery protocol, and updating the initial digital twin model accordingly to obtain a revised digital twin model, includes: Based on the high-speed power line carrier communication capability between the master control communication module and the slave node module in the distribution box, communication data is obtained through the interaction method of the master control communication module periodically broadcasting neighbor discovery frames and the slave node replying with response frames in random time slots; Based on the communication data and the connection attributes of the components in the distribution box, a node adjacency matrix is ​​constructed through a neighbor discovery protocol, and a preliminary topology map is obtained by verifying the node connectivity using a connected component algorithm. Based on the preliminary topology map and the electrical topology template, the topology difference is determined by calculating the graph editing distance, and abnormal data is removed to obtain the current valid topology map; Based on the current valid topology diagram, the corrected digital twin model is obtained by adding or deleting nodes and correcting connection relationships in the electrical topology template and thermal circuit model template of the initial digital twin model.

4. The method for quantifying the health status of a distribution box based on digital twins according to claim 1, characterized in that, A lumped-parameter thermal circuit model is constructed based on the modified digital twin model. The estimated internal hotspot temperatures are obtained using a state observer based on the pole placement method, including: Based on the modified digital twin model, the heat transfer relationship between nodes is derived through the heat flow conservation law and Kirchhoff's laws, resulting in the state-space equation of the lumped parameter thermal circuit model. Based on the state-space equations, the observer gain matrix is ​​designed using the pole placement method to obtain the state observer. Based on the status data of the distribution box, the internal hot spot temperature estimate is obtained by discretization recursive calculation through the status observer.

5. The method for quantifying the health status of a distribution box based on digital twins according to claim 4, characterized in that, Based on the modified digital twin model, the heat transfer relationship between nodes is derived using the heat flow conservation law and Kirchhoff's laws, resulting in the state-space equations of the lumped-parameter thermal circuit model, including: Based on the current valid topology in the modified digital twin model, the order of the lumped parameter thermal path model is determined by identifying heat source nodes and combining the total number of nodes. Based on the order of the lumped parameter thermal circuit model, a thermal network topology is constructed by connecting nodes sequentially with thermal resistance and thermal capacity. Based on the aforementioned thermal network topology, heat flow balance equations for each node are established using the heat flow conservation law and Kirchhoff's laws. The state-space equations of the lumped parameter thermal circuit model are then derived by further deducing the heat transfer relationships between nodes.

6. The method for quantifying the health status of a distribution box based on digital twins according to claim 4, characterized in that, The process of obtaining the internal hotspot temperature estimate by discretizing and recursively calculating using the state observer based on the distribution box status data includes: Based on the lumped parameter thermal circuit model, the state observer is discretized by setting a discretization period to obtain the state estimation recursive equation; The status data of the distribution box is processed, and the real-time power of the heat source node is calculated according to Joule's law. Based on the processed state data and real-time power, the state estimation recursive equation is substituted into the recursive calculation to obtain the preliminary temperature estimate of the internal hotspot. Based on the deviation between the preliminary internal hot spot temperature estimate and the measured internal surface temperature of the distribution box, the internal hot spot temperature estimate is obtained by feedback correction of the input parameters of the state estimation recursive equation.

7. The method for quantifying the health status of a distribution box based on digital twins according to claim 1, characterized in that, Based on the estimated internal hotspot temperature and the causal logic of the distribution box fault, the posterior probability of the faulty node is obtained through a dynamic Bayesian network. A composite risk index for the distribution box is obtained by defining risk factors and introducing physical boundary constraints for correction, including: Based on the causal logic of the distribution box fault, a dynamic Bayesian network is obtained by defining evidence nodes, hidden state nodes, fault nodes and constructing causal relationships between nodes. The causal logic of the distribution box fault is pre-constructed by fusing the physical mechanism of the distribution box fault with expert experience according to the modified digital twin model. Based on the estimated internal hotspot temperature and the real-time operating data of the distribution box, the posterior probability of the fault node is obtained through the membership function and the forward and backward algorithms. Based on the factory configuration of the distribution box, the corrected probability of the fault node is obtained through physical boundary constraint correction. Based on the corrected failure node probability, a composite risk index is obtained by defining risk factors and introducing coupling coefficients.

8. The method for quantifying the health status of a distribution box based on digital twins according to claim 7, characterized in that, Based on the estimated internal hotspot temperature and real-time operating data of the distribution box, the posterior probability of the fault node is obtained through membership functions and forward and backward algorithms, including: Based on the estimated internal hotspot temperature and the real-time operating data of the distribution box, the state probability value is obtained through the membership function and used as the input of the dynamic Bayesian network. Based on the dynamic Bayesian network, a Bayesian network inference model is obtained by setting the time window length and introducing the time transition probability of the Markov assumption. Based on the evidence input, the forward probability is obtained by traversing the observation data within the time window using a forward algorithm, and the backward probability is obtained by traversing the observation data within the time window using a backward algorithm. The posterior probability of the faulty node within the time window is obtained based on the forward and backward probabilities.

9. The method for quantifying the health status of a distribution box based on digital twins according to claim 7, characterized in that, Based on the factory configuration of the distribution box, the corrected probability of fault nodes is obtained through physical boundary constraint correction, including: The physical boundary constraints are determined based on the factory configuration of the distribution box. Based on the estimated internal hot spot temperature and the real-time operating data of the distribution box, the physical boundary constraints are compared to determine the out-of-limit physical quantities and their corresponding fault nodes. Based on the aforementioned physical quantities exceeding the limits and their corresponding fault nodes, the preliminary corrected fault node probabilities are obtained by setting the prior probability of the fault node to a fixed value and re-performing dynamic Bayesian network inference. Based on the initially corrected failure node probability, a second fine-tuning is performed by combining observational evidence of non-exceeding physical quantities to obtain the corrected failure node probability; the non-exceeding physical quantities are the remaining physical quantities in the physical boundary constraints other than the exceeding physical quantities.

10. A digital twin-based distribution box health status quantification system, used to implement the digital twin-based distribution box health status quantification method according to any one of claims 1-9, characterized in that, The system includes: The initial twin module is used to obtain the initial digital twin model based on the physical entity and factory configuration of the distribution box by deploying hardware sensors and building geometric models, electrical topology templates and thermal circuit model templates; The topology correction module is used to obtain the current valid topology map based on the connection attributes of the components in the distribution box through the neighbor discovery protocol, and update the initial digital twin model accordingly to obtain the corrected digital twin model. The hot spot estimation module is used to acquire the status data of the distribution box, construct a lumped parameter thermal circuit model based on the modified digital twin model, and obtain the estimated value of the internal hot spot temperature through a state observer based on the pole placement method. The risk quantification module is used to obtain the posterior probability of the fault node through a dynamic Bayesian network based on the internal hot spot temperature estimate and the causal logic of the distribution box fault, and to obtain the composite risk index of the distribution box by defining risk factors and introducing physical boundary constraints for correction.

Images