Search method and device for substation secondary full loop based on FBS algorithm

Abstract

The application relates to the technical field of power system intelligent operation and maintenance, in particular to a substation secondary full loop search method and device based on an FBS algorithm. The method comprises the following steps: obtaining a search source point and a target point according to initialization parameters; establishing a search level path and formulating a forward search strategy; determining a target loop path through reverse backtracking; constructing a digital twin model integrating physical, electrical and functional information and visualizing in a shadow mode; and combining real-time data and historical fault cases to predict potential risks. According to the embodiment, the search range is limited through forward detection, the path is accurately positioned through backward tracking, the algorithm complexity is significantly reduced, the search efficiency and loop positioning accuracy are improved, and safe simulation and risk pre-warning are supported.

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology of power systems, specifically to a search method and device for the secondary full circuit of a substation based on the FBS algorithm. Background Technology

[0002] With the deepening of smart substation construction, secondary systems widely adopt standardized digital models such as SDD (Substation Design Description) or SPD (Substation Physical Description) to describe equipment objects and cable connection relationships between terminals, forming a mesh topology defined by XML elements. This model expresses semantic information such as cable cores, terminal numbers, virtual terminal mappings, and functional labels in a structured manner, providing fundamental support for the digital modeling, automatic identification, and visualization of secondary circuits. Currently, mainstream circuit search methods still rely on Depth-First Search (DFS) and Breadth-First Search (BFS) algorithms, which search for connected paths from the source point to the target point in the SDD / SPD topology graph by traversing nodes and edges, possessing characteristics such as strong versatility and simple implementation.

[0003] In the process of realizing this invention, the inventors discovered that current methods have problems such as low search efficiency, high resource consumption, and inaccurate path positioning when dealing with large-scale, high-density, and multi-level substation secondary connection networks. Due to the lack of directional guidance, the algorithm needs to redundantly access a large number of irrelevant branches; it cannot terminate invalid explorations in advance based on the target point, resulting in the time complexity increasing exponentially with the network size; and it is difficult to support advanced application requirements such as rapid fault tracing, dynamic topology adaptation, and risk prediction. Summary of the Invention

[0004] One of the objectives of this invention is to provide a search method and apparatus for the secondary full circuit of a substation based on the FBS algorithm, which can solve the problems of low search efficiency, high resource consumption, and inaccurate positioning in the prior art in complex substation secondary network topologies.

[0005] To solve the above-mentioned technical problems, the embodiments of the present invention are implemented as follows: Firstly, a search method for the secondary full circuit of a substation based on the FBS algorithm is provided, including the following steps: The search source and target points are obtained based on the initialized search parameters, which include information on all equipment in the substation and their interconnection relationships. Establish a search hierarchy path based on the search source point and target point, and formulate a search strategy based on the search hierarchy path; Determine the target loop path of the target point by reverse-engineering the search hierarchy path; A loop topology diagram is created based on the loop path, and the loop topology diagram is visualized to obtain a visual model.

[0006] Secondly, an early warning device based on the FBS algorithm for substation secondary full-circuit search is provided, comprising the following modules: Acquisition module: Acquires the search source point and target point based on the initialized search parameters. The search parameters include information on all equipment in the substation and their interconnection relationships. Strategy formulation module: Establishes a search hierarchy path based on the search source point and target point, and formulates a search strategy based on the search hierarchy path; Loop path determination module: Determines the target loop path of the target point by reverse-engineering the search hierarchy path; Visualization module: Creates a loop topology map based on the loop path, and performs visualization processing on the loop topology map to obtain a visualization model.

[0007] As can be seen from the above technical solution of the present invention, the present invention provides a search method for the entire secondary circuit of a substation based on the FBS algorithm. The embodiment obtains equipment and connection relationship information between the source and target points through initialization, establishes a hierarchical path, and formulates a forward search strategy. It only advances the detection of connection relationships layer by layer without expanding the entire network topology, significantly reducing the search space. Then, it returns from the target point to the source point along the hierarchical path through reverse tracing, accurately locating the complete circuit path and avoiding blind traversal. Furthermore, it integrates physical, electrical, and functional information to construct a digital twin model, and achieves safety visualization and dynamic flow simulation in shadow mode. Finally, it combines real-time monitoring data with historical fault cases for comparative analysis to achieve risk trend prediction and structured early warning. The embodiment, employing a forward detection and backward tracking mechanism, solves the technical problems of low search efficiency, high resource consumption, and inaccurate positioning in complex secondary networks using traditional DFS / BFS algorithms, thereby improving circuit search response speed, path reconstruction accuracy, intuitive human-computer interaction, and forward-looking risk prevention. Attached Figure Description

[0008] Figure 1 The flowchart illustrates the search method for the secondary full circuit of a substation based on the FBS algorithm provided in the first embodiment of the present invention.

[0009] Figure 2 This is a flowchart illustrating the process of determining the loop path of the target point in reverse based on the search hierarchy path in an embodiment of the present invention.

[0010] Figure 3 This is a flowchart illustrating the creation of a shadow mode isolated from the running system in an embodiment of the present invention.

[0011] Figure 4 This is a loop topology diagram of one embodiment of the present invention.

[0012] Figure 5 The flowchart illustrates the search method for the secondary full circuit of a substation based on the FBS algorithm provided in the second embodiment of the present invention.

[0013] Figure 6 This is a schematic diagram of an early warning device for substation secondary full-circuit search based on the FBS algorithm in the first embodiment of the present invention.

[0014] Figure 7 This is a schematic diagram of an early warning device for substation secondary full-circuit search based on the FBS algorithm in the second embodiment of the present invention. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit the invention.

[0016] Substation electrical secondary systems are characterized by a large number of devices, highly networked connections, strong dynamics in terminal-level topology, and tight coupling of circuit functions. In operation and maintenance, technical upgrades and acceptance, and fault handling scenarios, it is necessary to quickly and accurately identify the complete secondary circuit starting from a certain signal source (such as protection action output or remote signaling input) and ending at the target actuator (such as circuit breaker trip coil or alarm indicator). However, existing pathfinding methods based on Depth-First Search (DFS) or Breadth-First Search (BFS) have significant drawbacks when dealing with actual substation secondary digital models containing hundreds of IEDs, thousands of cable cores, and tens of thousands of connection nodes. Firstly, the algorithm's time complexity increases exponentially, with a single full-loop search taking tens of seconds or even minutes, failing to meet the "second-level response" requirements in the field. Secondly, blind traversal leads to the inclusion of numerous irrelevant nodes in the calculation, causing a surge in memory consumption and easily triggering system resource bottlenecks. Thirdly, only a linear path list is output, lacking a joint expression of the loop's physical layout, electrical constraints, and functional logic, making it difficult to support human-machine collaborative judgment. Fourthly, the lack of a closed-loop analysis mechanism with real-time operational data prevents the correlation and deduction of topology structure with equipment health status and historical fault modes, causing risk warnings to lag behind the actual deterioration process. These problems collectively result in low efficiency in secondary loop location, weak basis for handling, and invisible safety margins, severely restricting the improvement of the autonomous and controllable operation and maintenance capabilities of intelligent substation secondary systems.

[0017] Reference Appendix Figure 1 As shown, this embodiment provides a search method for the secondary full circuit of a substation based on the FBS algorithm, including the following steps: Step S001: Obtain the search source point and target point according to the initialized search parameters. The search parameters include information on all equipment in the substation and their interconnection relationships.

[0018] Understandably, the initial search parameters refer to the set of static and dynamic metadata of the substation secondary system that is pre-loaded and structured before starting the FBS (Forward-probing Backward-tracing Search) algorithm. This set contains at least two types of core data: the first type is equipment information, including but not limited to equipment type (such as protection devices, measurement and control devices, intelligent terminals, merging units), globally unique identifier (GUCI), cabinet number, installation location coordinates, terminal block number, and physical attributes of each terminal (such as terminal number, wiring method, rated current / voltage); the second type is interconnection relationship information, derived from SDD (Substation Design Description) or SPD (Substation Protection Description) standard files, defined in XML format, specifically covering the cable body, including three types of entities: cable, cable core, and connection node, and their topological relationships. The connection node serves as a logical convergence point, recording the terminal IDs and direction attributes (in / out) of all terminals connected to it, forming the mesh skeleton of the secondary circuit. Both the search source point and the target point are selected from this parameter set. The source point is typically a fault trigger point, such as the "trip output GOOSE signal" terminal of a line protection device; the target point is typically a controlled execution point, such as the "trip coil TQ" terminal of the corresponding circuit breaker operating box. This acquisition process is achieved by parsing the DOM tree structure of the SDD / SPD file and using XPath query statements to locate the device object corresponding to the specified GUCI and its terminal connection relationship, ensuring the semantic uniqueness and topological traceability of the source and target points. As an optional implementation, the source and target points can also be dynamically specified by manually inputting the GUCI through the human-machine interface or by scanning a QR code, and the system automatically verifies their existence and validity in the current model. As another optional implementation, the source point can be automatically triggered by real-time alarm events pushed by the SCADA system, and the target points are generated in batches according to a preset rule base, such as all trip-related terminals in the same interval, to support concurrent searches of multiple targets.

[0019] In this embodiment, the process of obtaining the search source point and target point based on the initialized search parameters is as follows: a secondary digital model of the substation is constructed using an SDD / SPD file, and the equipment information and interconnection information of the substation in the secondary digital model are obtained; the search source point and target point information are obtained according to the received instructions; and the search hierarchy is determined based on the connection relationship between the search source point and the target point to constrain the search depth.

[0020] In this embodiment, a substation secondary system digital model is obtained by defining the full object model of the substation secondary system using SDD / SPD files in a structured XML format. This differs from traditional CAD drawings (manual drafting) or unstructured text descriptions, offering strong consistency, traceability, and machine readability, providing a structured knowledge base for subsequent automated loop searches. Defining the full object model of the substation secondary system using structured XML format is a common method and will not be elaborated upon here.

[0021] Next, equipment-level attribute information, such as equipment type, manufacturer, model, GUCI, cabinet number, installation location coordinates, terminal block number, and physical number of each terminal, is extracted from the SDD / SPD file of the substation secondary digital model. Simultaneously, connection relationship triplets are parsed, including source device.terminal, connection node, and target device.terminal, and further associated with the cable and cable core levels to reconstruct the complete physical link path, such as: terminal X1:1 of protection device A → connection node CN-203 → cable CBL-456 → cable core CBL-456-A1 → terminal Y2:3 of intelligent terminal B. This includes not only explicit cable connections but also implicit logical connections, the latter being determined through the SPD. <goosecontrol>and <sampledvaluecontrol>The LD / LN / DO paths referenced within the block are mapped.

[0022] When the monitoring system detects that a circuit breaker issues an "SF6 low pressure lockout" alarm, it automatically sets the circuit breaker as the target point and sets all upstream protection devices in its GOOSE subscription relationship as candidate source point sets; or according to the fault recording file in the dispatch instruction, it parses the starting trip signal source (such as "RCS-931AM protection action output") as the source point and the circuit breaker controlled by the action as the target point.

[0023] The search hierarchy is determined by the connection relationship between the source and target points. The source point is used as the first-level baseline node. Based on the parsed connection graph in SDD / SPD, a one-way breadth-first search (BFS-like propagation) is performed, adding one level with each propagation until the target point's node is first covered. This process does not elaborate on the specific path, but only establishes a layer-to-layer reachability map. For example: source point A ∈ → Reachable device set {B1, B2} ∈ → The set of reachable devices {C3, C4, C5} ∈ →… →Target point D5 ∈ Therefore, the minimum theoretical number of levels is determined to be C=5, and C is used as the maximum level threshold for the forward probing phase of the FBS algorithm. Nodes exceeding this level are not included in the search set. This hierarchical relationship is not a fixed topological hierarchy, but a dynamically calculated shortest hop count path depth, which can effectively avoid redundant loop expansion and interference from irrelevant branches.

[0024] In this embodiment, the SDD / SPD file provides a standardized and parsable initial data source, ensuring the integrity and standardization of devices and connection relationships. A multimodal instruction mechanism enables flexible and precise specification of source / target points, adapting to both manual intervention and automatic system triggering scenarios. A hierarchical constraint mechanism based on dynamic calculation of connection relationships transforms the abstract "search depth" into a concrete and quantifiable layer threshold C, giving the FBS algorithm's forward probing process clear boundaries and termination conditions. These three elements together constitute the initialization closed loop of the search task, ensuring both the engineering reliability of the input data and providing the search process with responsiveness and computational controllability to meet actual operational needs.

[0025] The implementation example achieves automated acquisition and semantic parsing of structured data from substation secondary systems, avoiding information distortion and omissions caused by manual data entry; it supports rapid task configuration in complex business scenarios through a command-driven source / target point dynamic binding mechanism; and it significantly compresses the search space of the FBS algorithm's forward probing through connection-relationship-driven hierarchical depth constraints, ensuring the accuracy of loop location while reducing the algorithm's time complexity from... The level (k is the average degree of the network) is reduced to The system employs a multi-level architecture, where C represents the number of levels and M represents the average number of devices per level. This significantly improves the real-time performance of searches and the utilization rate of system resources. Ultimately, it supports a basic capability for secondary full-circuit intelligent search that can be implemented in engineering, deployed in batches, and seamlessly integrated with existing digital substation systems.

[0026] Step S002: Establish a search hierarchy path based on the search source point and target point, and formulate a search strategy based on the search hierarchy path.

[0027] Understandably, the search hierarchy path is not a specific edge sequence in traditional graph theory, but rather an abstract hierarchical structure with the source vertex as the first layer and the target vertex's layer as the final layer. Each layer represents the set of all reachable devices within a specific hop count from the source vertex. The essence of the search hierarchy path is the forward probing phase of the FBS algorithm: starting from the source vertex, based on the connection relationships defined by the connecting nodes, the reachable device set is expanded layer by layer outwards. Each expansion forms a layer set, where the shortest path hop count between devices and the source vertex is the same. When the target vertex first appears in a certain layer set, the probing terminates, and the resulting hierarchical sequence is the search hierarchy path.

[0028] For example, if source point A can reach target point D5 in 2 hops, then the hierarchical path is: , , , ,in The aforementioned search strategy is based on generating a forward search strategy from the hierarchical path. Its core principle is to limit the search scope to connections between adjacent layer sets, prohibiting cross-layer or intra-layer traversal, thereby expanding the search space from... Compress to This significantly reduces computational complexity. As an optional implementation, the layer set expansion depth can be dynamically configured, such as a default of 5 layers and a maximum of 8 layers. When the target point is not found after reaching the preset number of layers, the system returns a "no valid path" message and logs it. As another optional implementation, layer set construction can introduce weighting factors to assign higher priority to high-reliability devices, such as dual-channel fiber optic connections, allowing devices in the same layer to be sorted by weight and prioritizing the detection of high-confidence path branches.

[0029] In a further embodiment, the step of establishing a search hierarchy path based on the search source point and the target point, and formulating a search strategy based on the search hierarchy path, involves: acquiring associated devices that are connected to the search source point and the target point; classifying the associated devices according to the progressive connection relationship from the search source point to the target point; and assigning the classified devices to the layer sets determined according to the hierarchy relationship, thereby obtaining a multi-layered structured search set arranged sequentially according to the progressive connection relationship from the search source point to the target point; obtaining the search hierarchy path based on the arrangement order of the multi-layered structured search set; and formulating a forward search strategy based on the search hierarchy path, starting from the layer set where the search source point is located and sequentially searching towards the layer set where the target point is located.

[0030] In a further embodiment, classifying associated devices according to the progressive connection relationship from the search source point to the target point, and assigning the classified devices to the layer sets determined according to the hierarchical relationship, involves: determining the search layer number of the multi-layer structure search set according to the hierarchical relationship; classifying associated devices according to the connection order from the search source point to the target point, and assigning the classified associated devices to the respective layer sets of the multi-layer structure search set.

[0031] It should be noted that the acquisition of associated devices connected to the search source and target points refers to: using SDD or SPD files as the data source, parsing the structured objects defined therein, such as connection nodes, cables, cable cores, and IEDs (Intelligent Electronic Devices), to identify all secondary devices that are directly electrically connected to the search source or target point through physical cables or logical terminals. The search source point is such as the output terminal of a fault protection device, and the target point is such as the input terminal of a circuit breaker trip coil, including but not limited to relay protection devices, measurement and control devices, merging units, intelligent terminals, operating boxes, terminal blocks, and intermediate relays. This process does not rely on manual drawing interpretation but achieves automatic topology traversal through XMLSchema validation and XPath path matching, ensuring the completeness and reproducibility of associated device identification.

[0032] It should be noted that the classification of associated devices based on the progressive connection relationship from the search source point to the target point means: taking the search source point as the first-level set reference, and according to the signal flow and functional hierarchy relationship between the devices, the associated devices are classified into layers according to their logical step positions in the source and target signal transmission chains. For example, output relays, optocoupler modules, and tripping plates directly connected to the source point are classified into the second-level set; re-operating relays and intermediate relays in the control box connected to the output terminals of devices in the second-level set are classified into the third-level set; the next level down, devices directly connected to the tripping coil of the circuit breaker mechanism box are classified into the fourth-level set; and the final target point, such as the layer where the circuit breaker tripping coil is located, is the target layer. This classification does not depend on the physical installation location of the devices, but strictly follows the logical connection path defined in the IEC 61850 standard to ensure that the layering results conform to the functional logic of the secondary circuit.

[0033] The determination of the search layer number of the multi-layer structure search set based on hierarchical relationships refers to the following: Before the FBS algorithm is started, based on the topology complexity of the substation secondary system, equipment connection density, and maintenance response timeliness requirements, the maximum allowable search depth C is preset or dynamically derived. This C is the total number of layers in the multi-layer structure search set. For example, in the secondary circuit scenario of a medium-sized 220kV substation, C=5 is set, indicating that the search can extend to the 5th layer of equipment from the search source point. In a large 500kV smart substation, due to the long protection links and deep logic nesting, it can be dynamically increased to C=7. This number of layers is neither a fixed constant nor an upper limit for traversing the entire network, but rather a flexible boundary parameter with the termination criterion being whether the target point falls into a certain layer of the search set. Its physical implementation is as follows: During the construction of the search hierarchy path, the count is incremented by 1 each time a device expansion in the breadth direction is completed. When the current layer number is equal to C and the target point has not yet been captured, an alarm is triggered and a message "target unreachable" is displayed to avoid invalid calculations.

[0034] The classification of associated devices according to the order of connection from the search source point to the target point refers to: starting from the search source point, based on the cable connection relationships defined in the SDD / SPD file, such as the connection relationships of cable, cable core, connection node, terminal, and device in sequence, identifying the set of devices with a direct electrical path to the source point level by level along the signal flow, and sorting and classifying them according to the shortest path hops between them and the source point. For example, relays and control box terminal devices directly connected to the source point via a single cable core are classified as Level 2 devices; measurement and control devices and merging units connected via an intermediate connection node (such as a terminal block) are classified as Level 3 devices; and intelligent terminals and switch ports reached after one more transfer are classified as Level 4 devices. This classification process does not depend on the device type or functional role, but strictly follows the directed transitivity of the physical connection topology, ensuring that each layer of devices is a necessary node to form a potential path.

[0035] It should be noted that classifying the devices into layers determined by hierarchical relationships means: establishing an independent device set container (Layer Set) for each layer, with each container storing the GUCI, device type, cabinet number, terminal number, and connection relationship metadata of all devices in that layer; the upstream and downstream connection relationships between the layers are recorded through directed edges, forming a directional hierarchical graph structure. Specifically, the first layer set container only contains the search source point itself; the second layer set container contains all devices directly connected to the source point; the third layer set container contains all devices directly connected to the devices in the second layer set but not appearing in the first or second layers, and so on. Inter-layer index relationships are established between the containers through pointer mapping, supporting inter-layer jump queries with O(1) time complexity. This classification process synchronously performs deduplication verification. If a device's GUCI has already appeared in a higher-level container, the classification is skipped to prevent duplicate calculations caused by loops.

[0036] It should be noted that the multi-layered search set obtained according to the progressive connection relationship from the search source point to the target point means: the first layer set (source point), the second layer set, the third layer set, ... up to the layer where the target point is located are arranged in ascending order according to the jump number to form an ordered layer sequence. This sequence implicitly contains the minimum logical path length constraint for the signal to propagate from the source to the target. The devices within each layer are unordered, but there is a strict unidirectional dependency relationship between layers. That is, the device in the i-th layer can only send signals to the device in the i+1-th layer and cannot drive across layers or in reverse. This structure is essentially an abstraction of the original mesh topology by a directed acyclic graph (DAG), stripping away redundant feedback branches and bypass connections, and retaining the core control flow path.

[0037] It should be noted that obtaining the search hierarchy path based on the arrangement order of the multi-layer structure search set refers to converting the layer sequence into a virtual hierarchy coordinate path, the form of which is: (k is the layer number of the target point). This path does not describe a specific device, but defines the hierarchical transition order that the search must follow. This path serves as the algorithm navigation skeleton to guide subsequent search actions to stay on the main logical chain and prevent invalid stays in irrelevant branches.

[0038] It should be noted that the aforementioned forward search strategy, which involves starting from the set of layers containing the search source point and sequentially searching towards the set of layers containing the target point based on the search hierarchy path, means that only searches from the set of layers containing the target point are allowed. Starting from the layer device, search for its location. All downstream connected devices in the layer are now complete. Only after layer search is started Layer-by-layer search ensures that the entire process is deterministic and interruptible, without backtracking, skipping, or concurrently scanning non-adjacent layers.

[0039] Working Process: Automatic identification of associated devices provides the data foundation for hierarchical structure; logical classification of progressive connection relationships ensures that the hierarchical structure conforms to the functional flow of the secondary system; the structured organization of the hierarchical set supports efficient hierarchical indexing; the ordered arrangement of multi-layered structural search sets forms the path navigation skeleton; the search hierarchical path transforms the abstract hierarchy into executable coordinate instructions; the forward search strategy ultimately translates these instructions into deterministic, low-overhead layer-by-layer probing actions. Through the above steps, this application achieves the replacement of blind traversal with directional guidance in the large-scale network connection relationship of substation secondary systems. By decoupling the hierarchy and compressing the search space, the search process focuses on the core logical chain between the source and the target, thereby effectively solving the problems of long search time, high resource consumption, and insufficient fault loop location accuracy of existing DFS / BFS algorithms in scenarios with numerous connection nodes and complex relationships. This achieves the technical effects of improving search efficiency, reducing computational resource consumption, and enhancing loop location accuracy.

[0040] Step S003: Determine the target loop path of the target point in reverse order based on the search hierarchy path.

[0041] Understandably, the reverse determination of the target loop path refers to the backward tracking phase of the FBS algorithm: starting from the target point, tracing back layer by layer to the source point along the inter-layer connection mapping table recorded during the forward probing process, ultimately piecing together a precise connection chain pointing from the target point back to the source point. Here, the inter-layer connection mapping table is a list of all predecessor devices for each device in the previous layer. This process does not rely on global topology traversal; it only needs to access the cached inter-layer adjacency relationships, resulting in a time complexity of O(H), where H is the total number of layers in the hierarchical path. For example, in the aforementioned hierarchical path, the predecessor device of D5 is... , The precursor is , The predecessor is A, therefore the loop path is The forward expansion is The target loop path specifically refers to the continuous link of terminal-cable core-terminal identified by the forward sequence. Each link corresponds to a real cable connection, possessing physical feasibility and a closed signal logic path, ensuring functional integrity. As an optional implementation, when a device has multiple predecessors, the system can automatically select the optimal path according to preset rules, such as selecting the path with the shortest tripping delay or the path with the fewest fiber optic cross-sections. As another optional implementation, the backtracking process supports interactive intervention, allowing maintenance personnel to manually specify the next-hop predecessor device at any layer set, achieving manually guided path correction.

[0042] Reference Appendix Figure 2 As shown in the embodiment, the step of determining the loop path of the target point in reverse according to the search hierarchy path includes: Step S301: Determine the layer set where the target point is located.

[0043] Understandably, determining the layer set where the target point is located refers to: in the multi-layer structure search set constructed in step S002, locating the specific layer number to which the target point is assigned based on the device affiliation layer information recorded during the forward probing process. This layer number is dynamically generated and persistently stored by the FBS algorithm during the forward probing phase, and its value is equal to the number of layers from the search source point that can reach the target point with the fewest connection hops.

[0044] For example, if the search source point is in the first layer set, the secondary winding terminals of the current transformer directly connected to it constitute the second layer set, the input terminals of the merging unit connected to these winding terminals via cable cores constitute the third layer set, and the target point is the TRIP output terminal of a line protection device, and it first appears in the fifth layer set during forward probing, then the system determines that the target point is located in the fifth layer set. For example, the fifth layer set might include the output relay of a protection device, the trip coil terminal of a circuit breaker, or the source point of a fault alarm signal. This layer set not only includes the target point itself but also all other related devices included in this layer at the same probing depth, thus providing a complete set of candidate nodes for subsequent backtracking.

[0045] Step S302: According to the reverse search strategy of the forward search strategy, backtracking is performed starting from the layer set where the target point is located until the layer set where the search source point is located is located.

[0046] Understandably, the backward search strategy following the forward search strategy, which involves backtracking from the layer set where the target point is located until the layer set where the search source point is located, refers to performing top-down path reconstruction based on the inter-layer connection mapping table established and cached during the forward probing phase. LLCM stores the connection relationship between each layer device and its upstream device in key-value pairs.

[0047] For example: Device GUCI="PROT_L1_TRIP01" in layer 5 → Device GUCI="MU_LINE1_OUT03" in layer 4; "MU_LINE1_OUT03" in layer 4 → "CT_L1_S2_T1" in layer 3; "CT_L1_S2_T1" in layer 3 → "CT_L1_S2" in layer 2; "CT_L1_S2" in layer 2 → "CT_L1" in layer 1. The backtracking process involves looking up tables layer by layer, extracting predecessor nodes, and adding each predecessor node to the backtracking path queue until the layer set containing the search source node is reached. This process does not re-traverse the original topology graph, relying only on the lightweight hierarchical index built during the forward probing phase, significantly reducing computational overhead.

[0048] Step S303: Based on the connection relationship of each associated device in the backtracking path, obtain the connection path from the target point to the search source point, and obtain the loop path.

[0049] Understandably, obtaining the connection path from the target point to the search source point based on the connection relationships of various associated devices in the backtracking path, and thus obtaining the loop path, means: arranging the GUCI sequences of devices obtained sequentially during the backtracking process in reverse chronological order, such as ["PROT_L1_TRIP01", "MU_LINE1_OUT03", "CT_L1_S2_T1", "CT_L1_S2", "CT_L1"], to form a standard signal flow path from the search source point to the target point; then, combining this with the terminal connection descriptions, cable core definitions, and connection node IDs of the corresponding devices in the SDD / SPD file, reconstructing the complete physical connection link, including the cable type, laying path, terminal number, and electrical interface type of each segment. This path is output in structured JSON format, containing four-dimensional attribute fields: device layer, physical layer, electrical layer, and functional layer, for subsequent visualization and risk analysis.

[0050] In this embodiment, hierarchical location provides spatial anchors for backtracking, the reverse search strategy relies on the hierarchical index established by forward probing to achieve efficient path reconstruction, connection relationship parsing ensures both physical and logical accuracy of the loop path, secondary loop topology enables precise definition of the fault impact domain, and tripping strategy formulation transforms the topology analysis results into executable relay protection action schemes. Data flow in each stage is unidirectional and progressive, with closed-loop status feedback, forming a complete technical chain of "location—reconstruction—modeling—decision making".

[0051] The implementation scheme achieves the following through the above steps: In a large-scale mesh topology of a substation secondary system, the target loop path can be reconstructed with high accuracy without global traversal; it overcomes the shortcomings of traditional DFS / BFS algorithms, such as high path reconstruction delay, large resource consumption, and susceptibility to topology noise interference caused by blind search; and through the tight coupling design of secondary loops and tripping strategies, the fault isolation range is compressed from the traditional "interval level" to the "functional link level", significantly improving the selectivity, speed, and reliability of relay protection, effectively solving the core problems of insufficient fault loop location accuracy, lack of ability to quickly generate the entire loop path by combining fault alarm signals, and difficulty in supporting efficient fault handling.

[0052] Step S004: Establish a loop topology diagram based on the loop path, and perform visualization processing on the loop topology diagram to obtain a visualization model.

[0053] Understandably, the loop topology diagram is a dedicated graphical representation generated based on the loop path obtained in step S003. Its nodes are all device terminals and key connection nodes in the path, and the edges are the cable core entities connecting these nodes, forming a logically closed but not necessarily physically looped directed graph. The establishment of the loop topology diagram is accomplished by calling a graphics engine, which includes D3.js or Qt Graphics View. The steps include: first, parsing the geometric installation information of each GUCI corresponding device in the loop path, the geometric installation information comes from the Location element in SDD / SPD, and determining the initial coordinates of the nodes; second, reading the start and end terminal IDs of the cable cores, automatically generating Bézier curve edges, and labeling the cable specifications; and finally, integrating the device icon library to render the node styles. The visualization process involves the fusion of information from three dimensions: physical layer information, such as equipment installation location, cable routing, and cabinet space layout; electrical layer information, such as rated voltage / current, insulation class, and GOOSE / SV communication parameters; and functional layer information, such as protection logic type, signal purpose, and control access level. These three layers are bound to corresponding graphical elements through a unified data model, such as a UML class diagram definition, supporting real-time display of multi-dimensional attributes on mouse hover. The visualization model is an interactive 3D / 2D scene rendered from the above fusion, supporting zooming, panning, device highlighting, path coloring, and topology expansion / collapse.

[0054] Reference Appendix Figure 3 As shown in the embodiment, the steps of establishing a loop topology map based on the loop path and visualizing the loop topology map to obtain a visualization model include: Step S400: Analyze the physical layer information, electrical layer information and functional layer information of all devices in the loop path, and construct a multi-dimensional digital twin model by fusing the physical layer information, electrical layer information and functional layer information for visualizing the loop topology.

[0055] The physical layer information, electrical layer information, and functional layer information are obtained from the parsing of the SDD / SPD file. The physical layer information includes the spatial installation location of the equipment within the substation, mechanical connection relationships, and environmental constraint parameters. The electrical layer information includes the equipment's rated electrical parameters, signal type, transmission characteristics, and fault electrical response characteristics. The functional layer information includes the equipment's logical role in the secondary circuit, control logic dependencies, signal flow semantics, and functional safety level.

[0056] The construction of a multi-dimensional digital twin model by fusing physical layer information, electrical layer information, and functional layer information includes: A unified spatiotemporal semantic framework is constructed, using the device's GUCI as the sole anchor point to map three types of heterogeneous information to the same coordinate system. The physical layer adopts a dual reference of WGS-84 geographic coordinate system + local millimeter-level 3D mesh coordinate system; the electrical layer is bound to the IEC61850 Logical Device (LD) namespace; and the functional layer is attached to the behavioral sequences in the IEC 62559 use case model. The three are associated across layers through the `hasPhysicalProperty`, `hasElectricalProperty`, and `hasFunctionalRole` object attributes in the device ontology. As an optional embodiment, when a device lacks information from a certain layer, the system adopts a fuzzy inference mechanism: confidence intervals are generated based on the statistical mean of similar devices to ensure model integrity.

[0057] The multi-dimensional digital twin model uses a graph database (Neo4j or TigerGraph) as its underlying storage. Nodes represent devices or connection points, and edges represent connection relationships (including cables, optical fibers, and wireless links). Each edge carries three attribute labels: physical layer (length, cross-sectional area), electrical layer (impedance, attenuation), and functional layer (signal type, priority). The model supports LOD (Level of Detail) hierarchical rendering: the global view displays the cabinet-level topology, and when drilling down to the terminal level, it automatically loads physical wiring photos and electrical parameter heatmaps.

[0058] In a further embodiment, the creation of a shadow mode isolated from the running system is: Step S401: Simultaneously create a shadow mode isolated from the running system to ensure the safety of all planning and simulation operations.

[0059] The Shadow Mode refers to a virtual simulation environment that is logically completely isolated from the actual secondary operation system of the substation, highly faithful in data, and capable of real-time mapping in timing. It is not limited to a static graphical interface display, but rather a multi-dimensional digital twin that carries the static attributes, connection relationships, dynamic behaviors, and evolutionary trends of the equipment. This mode does not participate in the execution of real control commands and does not send trip or lockout signals to primary equipment, ensuring that all analysis, simulation, and operation are completed within safety boundaries. This meets the mandatory requirements of GB / T 36275—2018 Technical Specification for Safety Protection of Intelligent Substations regarding "logical isolation between the production control area and the management information area" and "physical / logical isolation between the simulation test environment and the real-time monitoring system."

[0060] In a further embodiment, creating a shadow mode isolated from the running system means: Step S410: Construct a clone model containing information data of all devices in the loop based on the loop topology diagram.

[0061] The loop topology graph is a directed connected graph derived from the target loop path. Its nodes are secondary devices uniquely identified by GUCI, such as protection devices, smart terminals, merging units, optical cable splice boxes, terminal blocks, etc., and the edges are the connection nodes, cable cores and cable three-level connection relationships defined in the SDD / SPD file. The clone model refers to a structured digital copy built in memory or a lightweight database based on this topology graph, which includes, but is not limited to, the following dimensions of information: Basic metadata of the equipment: GUCI, equipment type, name of the associated IED, cabinet number, installation location coordinates, terminal number; Connection semantic information: Each edge carries the connection direction, cable type, number of cores, insulation class, and transmission medium type; Functional configuration snapshot: IED internal logical node instantiation parameters, GOOSE / SV subscription / publishing list, CRC checksum, configuration version number.

[0062] The cloned model is serialized and stored in JSON-LD format and semantically annotated using a Schema.org-compatible ontology model, supporting subsequent semantic alignment with external knowledge graphs, including a fault case library and an equipment defect knowledge library.

[0063] Step S411: Obtain the dynamic status data of all devices in the loop topology diagram. The dynamic status data includes real-time health index, operating parameters, and alarm status.

[0064] In this example, dynamic status data is collected in real time through the substation control layer communication network or process layer network, with a collection period of ≤500ms, timestamp accuracy at the millisecond level, and validity verified by data quality marking. The real-time health index is a dimensionless normalized value, ranging from 0.0 to 1.0, calculated by the device's embedded diagnostic module or the site-wide intelligent analysis unit. Its calculation model includes: Hardware layer: weighted fusion based on temperature sensor readings, power supply ripple coefficient, Flash erase / write cycles, and FPGA configuration verification failure frequency; Communication layer: dynamic correction based on GOOSE message packet loss rate, SV sample value dispersion, and heartbeat timeout count; Functional layer: based on protection action logic criterion compliance, input jitter frequency, and output command response delay deviation. This HI value supports threshold-based alarm grading (HI < 0.3 for severe anomalies, 0.3 ≤ HI < 0.6 for alert status, and HI ≥ 0.6 for normal status), and can be replaced with a predictive health score (Predictive HI) trained based on an LSTM neural network. Inputting a historical 72-hour HI sequence and ambient temperature and humidity data, it outputs the HI decay trend for the next 24 hours. Operating parameters include, but are not limited to: device operating temperature (°C), DC power supply voltage (Vdc), CPU utilization (%), memory remaining rate (%), measured GOOSE release interval (ms), SV sampling rate deviation (ppm), and fiber optic link optical power (dBm). The alarm status adopts the AlarmStatus enumeration type defined by the IEC 61850 standard, covering three levels: Level 1 (general alarm), Level 2 (critical alarm), and Level 3 (urgent alarm). Each alarm is accompanied by a timestamp, acknowledgment status, suppression flag, and associated logical node path.

[0065] Step S412: Integrate the dynamic state data into the clone model to obtain the shadow pattern.

[0066] In this embodiment, the fusion refers to establishing a two-way binding relationship between dynamic state data and device entities in the clone model, including: Attribute mapping layer: Pre-sets dynamic attribute slots for each device node in the clone model, such as "healthIndex", "operatingTemp", and "alarmList", and matches them with the device identifier field in the real-time data stream using GUCI as the primary key; Timing synchronization layer: The PTP protocol is used to align the cloned model system clock with the substation clock source to ensure that HI update events, alarm trigger events and model state changes are traceable under a unified timeline; Data fusion strategy: A weighted average method is used to generate a fused HI for multi-source HI values ​​of the same device; a sliding window filter is used to suppress noise for high-frequency operating parameters; and a "set upon first trigger, reset after manual confirmation" mechanism is adopted for alarm status to avoid misjudgment caused by instantaneous interference. The shadow mode formed after fusion not only has static topology integrity but also dynamic behavioral realism. For example, when the HI value of a certain merging unit drops sharply from 0.72 to 0.41, the corresponding node in the shadow mode automatically changes color (from green to orange) and highlights "Sampling delay exceeds standard (current 28ms > threshold 15ms)" and the associated GOOSE subscription IED list in its attribute panel; when a smart terminal triggers a Level 3 alarm "Output relay sticking", the device icon in the shadow mode flashes red and automatically highlights all the circuit breaker smart terminal nodes connected to its outgoing circuits, forming a visualized path for risk transmission. As an optional implementation, this fusion process supports hot-swappable data source access, allowing the dynamic loading of new sensor data streams or third-party diagnostic interfaces without stopping the operation of the shadow mode, improving system scalability.

[0067] The implementation example achieves the following: A high-fidelity clone model is constructed using a loop topology diagram as its framework, ensuring the accuracy of the spatial structure and connection semantics of the shadow mode; by collecting and fusing three types of dynamic data—HI, operating parameters, and alarm status—at millisecond levels, the clone model is endowed with real-time perception and state evolution capabilities; further, through attribute mapping, time-series synchronization, and multi-source fusion strategies, the shadow mode becomes a digital sandbox that combines static topology consistency, dynamic behavioral realism, and risk transmission traceability. This solution addresses core issues such as the lack of ability to quickly generate full-loop paths and support three-dimensional visualization by combining defect alarm signals, as well as the low efficiency and poor accuracy of physical loop model visualization and editing. It enables maintenance personnel to conduct tripping strategy simulation, fault injection deduction, and contingency plan verification in a zero-risk environment, significantly improving the timeliness and reliability of secondary system defect handling, while providing a secure, reliable, and dynamic data foundation for the aforementioned visualization model.

[0068] Step S402: Use color to distinguish the roles and running status of the loop topology diagram in shadow mode, and then use dynamic flow simulation to simulate the operation process.

[0069] The role coding adopts the color semantic specification recommended by the International Electrotechnical Commission (IEC): newly connected equipment (construction equipment that has not passed acceptance) is highlighted with RAL 5012 (cobalt blue); normally operating equipment (passing all joint commissioning tests and HI≥0.8) is presented with RAL 6017 (grass green); equipment to be removed (offline but with physical connections not disconnected) is marked with a dashed box with RAL 3000 (tomato red); as an optional embodiment, the color scheme can be switched for different business scenarios: in maintenance mode, a color-blind friendly mode (blue / yellow / gray combination) is enabled, and in emergency response mode, key tripping paths are highlighted, such as with flashing gold edges to enhance visual focus.

[0070] In this embodiment, the operating status coding adopts a dynamic gradient strategy: the device HI value of 0.9–1.0 is displayed as pure green, 0.7–0.89 as yellow-green gradient, 0.5–0.69 as orange-yellow gradient, and below 0.5, a red pulse animation is superimposed; the connection cable status is colored according to the GOOSE / SV packet loss rate: ≤0.1% is a dark green solid line, 0.1%–1% is a light green dashed line, and >1% is a red wavy line.

[0071] In this embodiment, the dynamic flow simulation uses the IEC 61850-8-1 MMS protocol as the message carrier, encapsulates virtual signals, such as "circuit breaker trip command", into a `Control` service request, and forwards it hop-by-hop in shadow mode according to the GOOSE publish / subscribe mechanism; the delay of each hop is calculated by superimposing three levels: physical layer cable length × propagation speed (speed of light 0.67c) + electrical layer equipment processing delay + functional layer logic judgment time; the simulation process is visualized in the form of particle flow: blue particles represent trip commands, which flow at a constant speed along the topology path and split at branch points according to logical conditions.

[0072] In this embodiment, the physical layer information described above constitutes the spatial skeleton of the digital twin model, providing wiring constraints and heat dissipation boundaries for electrical layer parameters; electrical layer information defines the physical feasibility of signal transmission and supports the timing correctness verification of functional layer logic; functional layer information endows physical connections with business semantics and drives the conditional branching decisions of dynamic flow simulation; the shadow mode, as the operating carrier integrating the three, not only ensures the consistent mapping of physical, electrical, and functional multi-dimensional information, but also eliminates the risk of disturbance to the real system by simulation operations through a complete isolation mechanism; while color coding and dynamic flow simulation together constitute the human-computer interaction interface, transforming abstract multi-dimensional data into intuitive scenarios that are perceptible, operable, and deducible for operation and maintenance personnel.

[0073] The implementation example achieves a multi-dimensional digital twin model by integrating physical layer spatial layout, electrical layer transmission characteristics, and functional layer logical semantics. This solves the problems of low efficiency and poor accuracy in visualizing and editing physical loop models, enabling loop topology diagrams to not only display connection relationships but also reflect the actual operating conditions and functional intentions of equipment. By creating a shadow mode completely isolated from the running system, the lack of ability to quickly generate full loop paths and support three-dimensional visualization based on defect alarm signals is addressed, ensuring that all planning, simulation, and risk projection operations are completed within a secure sandbox. By employing dual-dimensional color coding for roles and states and dynamic flow simulation based on the IEC 61850 protocol, the problem of insufficient support for efficient defect handling is solved. This allows maintenance personnel to intuitively identify risk nodes, preview handling effects, and generate structured decision recommendations, significantly improving the timeliness and reliability of secondary system defect handling.

[0074] Reference Appendix Figure 5 As shown, the second embodiment further includes step S005, based on the above steps S001-S004: analyzing the loop topology diagram and equipment data to predict potential risks.

[0075] Understandably, the device data refers to real-time and historical operating data that are strongly correlated with each node in the loop topology diagram, including: real-time monitoring data, such as terminal voltage, current sampling values, GOOSE message packet loss rate, and SV synchronization deviation; device health index, such as HI, which is calculated by weighted fusion of multi-source sensor data such as temperature, vibration, and insulation resistance; and historical fault data, such as data stored in a relational database, with fields including fault type, occurrence time, scope of impact, root cause, and repair measures. The analysis process is divided into two levels: The first level is static topology risk identification, which involves traversing all nodes and edges of the loop topology graph and comparing the device HI threshold. For example, HI < 0.6 is marked as "sub-healthy", cable aging years (> 15 years) are marked as "high risk", and terminal contact resistance exceeding the limit (> 50mΩ) is marked as "loose". The second level is dynamic correlation risk inference, which involves semantically matching the identified risk points, such as "B2 device HI = 0.42", with the historical failure case library using TF-IDF + cosine similarity algorithm to retrieve historical cases with similarity > 0.7, such as "in 2023, a station's B2 device tripped and failed to operate due to HI continuously being lower than 0.5". Then, combined with the known evolution path of this case, such as "failure to operate occurred within 72 hours after HI dropped from 0.5 to 0.4", the development trend and failure probability of the current risk point in the next 24 / 48 / 72 hours are predicted, and a structured early warning report is generated. As one optional implementation, risk analysis can be integrated with the real-time simulation engine of a digital twin platform to inject virtual disturbances in shadow mode, such as simulating a momentary open circuit in a cable core; observe the location and time of signal propagation interruption, and quantitatively assess the risk impact boundary. As another optional implementation, the early warning report supports multimodal push and automatically links spare parts inventory with maintenance work order templates to form closed-loop handling recommendations.

[0076] In a further embodiment, the step of analyzing the loop topology diagram and device data to predict potential risks is as follows: Step S500: Obtain real-time monitoring data of all devices and combine it with the visualization model to evaluate and display the device status, thereby predicting the fault risk points of the devices in advance.

[0077] As can be understood, the acquisition of real-time monitoring data for all devices as described in this embodiment refers to the simultaneous collection of multi-source heterogeneous real-time data covering the physical, electrical, and functional layers from the substation secondary system online monitoring platform, IED devices, SCADA system, protection information substation, and edge computing nodes. This data includes, but is not limited to: current RMS value and harmonic components, voltage amplitude and phase deviation, terminal temperature, insulation resistance change rate, GOOSE / SV message packet loss rate and delay jitter, circuit breaker opening and closing coil current waveforms, operating mechanism energy storage pressure, optical port received optical power, and equipment health index, among other dynamic operating parameters. This real-time monitoring data is continuously updated with millisecond to second-level time granularity and accessed via the IEC 61850-8-1 MMS protocol or MQTT protocol to ensure data timeliness and integrity. This forms the basis for providing high-fidelity, low-latency state awareness for risk identification.

[0078] The integration of the visualization model for evaluating and displaying device status refers to dynamically binding the aforementioned real-time monitoring data to the corresponding device entities and connection nodes in a multi-dimensional digital twin model through spatial mapping relationships. For example, the HI value of a protection device is mapped to the surface color temperature of its three-dimensional model, such as 0.0–0.3 for red warning, 0.3–0.7 for yellow alert, and 0.7–1.0 for green normal. A current over-limit alarm signal triggers a pulse-like flashing animation on the corresponding terminal of the device, and an SV link delay exceeding tolerance event is superimposed with a red dashed arrow on the fiber optic connection line and labeled "Delay > 2ms". This embodiment achieves semantic alignment between status information and spatial topology, enabling maintenance personnel to intuitively locate the location and impact range of anomalies.

[0079] The aforementioned early prediction of equipment failure risk points refers to the joint analysis of multi-dimensional equipment status data to identify early abnormal signs that have not yet triggered hard alarms but show a deterioration trend. Examples include: a current transformer's secondary winding terminal temperature continuously rising by 12°C within 4 hours, accompanied by a sudden increase in high-frequency microvolt-level discharge signals; or a smart terminal's input board having three consecutive GOOSE message processing delay standard deviations exceeding a set threshold of 200%. This identification process employs a hybrid reasoning mechanism that integrates a rule engine and a lightweight machine learning model. The rule engine handles explicit threshold-type anomalies, such as temperature > 70°C, while the machine learning model is responsible for uncovering multi-parameter coupled deterioration patterns (such as the co-evolution path of temperature + vibration + partial discharge energy). This implementation can shift the risk identification point to the latent period, avoiding a passive response of "alarm equals failure."

[0080] Step S501: Match the fault risk points with historical fault data in the database to obtain fault cases.

[0081] Understandably, the embodiment performs multimodal similarity matching by calling a structured historical fault knowledge base: first, a coarse screening is performed based on equipment type, voltage level, wiring method, and defect phenomenon keywords (such as "smoke," "failure to operate," and "false signal"); then, cosine similarity calculation is performed using the BERT vector representation of the fault report text and the multidimensional state feature vector of the current risk point to select the Top-K (K=3–5) most similar historical cases; the historical fault knowledge base adopts a hybrid architecture of relational database and graph database, the former stores structured fields, including time, location, equipment ID, cause classification, and handling measures; the latter constructs a "equipment-defect-cause-effect" four-tuple causal graph, supporting deep association retrieval. The embodiment provides a verifiable empirical reference system for the current risk.

[0082] Step S502: The fault risk points are compared and analyzed with the fault cases, and the potential risk points of the system are predicted based on the comparison results.

[0083] In a further embodiment, the fault risk point is compared and analyzed with the fault case. Based on the comparison results, the potential risk points of the system are predicted by comparing the current fault risk point with the matched fault case item by item, analyzing their commonalities and differences, combining the known evolution path of the fault case, predicting the development trend and possible consequences of the current fault risk point, and automatically generating a structured report.

[0084] The step of comparing the current fault risk point with the matched fault cases item by item refers to: using the device-level granularity as a benchmark, aligning the dimensions involved in the fault risk point, such as physical entities, abnormal characterization parameters, operating environment conditions, and triggering timing characteristics, with similar fault samples already archived in the historical fault case library at the field level. This comparison process is implemented using a weighted similarity algorithm, where the weight for device type matching is 0.3, the weight for key parameter deviation ≤ ±15% is 0.25, the weight for environmental condition overlap is 0.2, and the weight for timing feature matching is 0.25; a case is considered a valid match when the overall similarity is ≥ 0.75.

[0085] The analysis of commonalities and differences refers to the following: after initial matching, the system automatically extracts and highlights common characteristics of both devices in terms of equipment topology, functional roles, abnormal signal combinations, and electrical parameter degradation modes. Simultaneously, it identifies differences, such as: the current risk point occurs in equipment put into operation within the last 6 months (within the warranty period), while historical cases mostly occur in equipment that has been in operation for more than 8 years; or the current temperature rise rate is 1.8℃ / min, significantly higher than the historical average of 0.9℃ / min; or the current cabinet lacks an online partial discharge monitoring module, while the cabinets in historical cases have complete status awareness capabilities. These difference analysis results are used to correct the initial assumptions for evolution path prediction. For example, if the current equipment is under warranty but the temperature rise rate doubles, it suggests a possible batch manufacturing defect or abnormal installation process, rather than simple aging.

[0086] The known evolution path combined with fault cases refers to: calling a pre-set fault evolution knowledge graph, which constructs multi-branch directed paths starting from the fault phenomenon and ending with the final functional failure. For example, an increase in terminal contact resistance as the starting node can evolve into two paths: Path 1 (68%) → "local overheating" → "insulation layer carbonization" → "phase-to-phase short circuit"; Path 2 (32%) → "oxide film thickening" → "signal transmission interruption" → "protection failure". Each edge is marked with an average evolution time, such as "local overheating → insulation layer carbonization" taking an average of 3.7 days; acceleration conditions, such as when the cabinet temperature is >55℃, shorten this stage to 8.2 hours; blocking measures, such as adding heat sinks, can extend this stage to 12.5 days. The system retrieves the subsequent node sequence of the corresponding branch based on the current evolution stage of the risk point (determined by real-time parameter fitting), and superimposes the current difference point analysis results to reweight the path. In an optional implementation, the evolution path can be modeled using an LSTM neural network, with the input being 72 hours of continuous multi-source monitoring time-series data (temperature, vibration, partial discharge pulse count, GOOSE link quality), and the output being the probability of occurrence of each failure mode in the next 24–72 hours.

[0087] The prediction of the development trend and possible consequences of the current fault risk point refers to generating a three-dimensional prediction output based on the corrected evolution path: a time dimension, used to provide a critical failure window with high confidence (≥90%), such as "the tripping circuit is expected to be completely open within 48–72 hours"; a spatial dimension, used to locate the high-risk propagation range, such as "it will cause the differential protection of the A set of #2 main transformer to be deactivated, and then expand to the point that the three-sided switches of #2 main transformer cannot be remotely tripped"; and a functional dimension, used to assess the system-level impact, such as "after the main transformer differential protection is lost, the backup protection action time is extended by 0.6 s, which may lead to the fault clearing time exceeding the limit and endangering the thermal stability of the transformer windings."

[0088] In this embodiment, the analysis of commonalities and differences, combined with the known evolution path of the failure cases, to predict the development trend and possible consequences of the current failure risk point specifically involves: First, constructing feature vectors, transforming the state features of the current failure risk point into vectors. ,in This includes factors such as temperature, voltage, load rate, insulation resistance, and commissioning duration. Simultaneously, the initial state features of the matched historical fault cases are characterized into a vector. .

[0089] Next, multidimensional comparison analysis is performed to calculate commonalities and differences, thereby determining the similarity between the two vectors. The calculations include commonality analysis and difference analysis. The commonality analysis is: identification and The characteristics of a high degree of consistency are considered. For example, if both are "the same type of relay", "the ambient temperature is 40°C", and "the circuit logic is a tripping circuit", then the fault mechanism is determined to be highly consistent. The system will assign a high reference weight to this historical case.

[0090] Difference point analysis is: identifying and Deviation characteristics. For example, the current load rate (95%) at the risk point is significantly higher than the load rate of historical cases (70%); or the current insulation resistance is low.

[0091] Furthermore, a weighted prediction model based on the evolution path is used for prediction. For example, assuming the known evolution path of historical failure cases is... That is, after time Subsequently, the equipment evolved from its initial state to a malfunction. For example: overheating -> insulation softening -> short circuit.

[0092] Here, a prediction function is constructed to correct the current timeline. :

[0093] in, For the first The degree of difference of each feature This represents the weight of the influence of this feature on the failure evolution rate.

[0094] The prediction process: If the difference point deteriorates rapidly, such as higher load or higher temperature, the system judges that the current fault evolution speed will be faster than historical cases, and the time window for predicting the fault occurrence will be brought forward. For example, in historical cases, the short circuit occurred after 2 hours, while the current prediction is that the short circuit may occur after 1 hour. If the difference point deteriorates slowly, such as lower ambient temperature or better ventilation, the time window for predicting the fault occurrence will be delayed.

[0095] Consequence prediction process: If the commonalities show that the loop topology and core fault mechanism are completely consistent, the system will directly use the final consequences of historical cases as the current predicted consequences, such as "may lead to protection malfunction" or "may lead to DC system grounding". If the differences involve different critical protection equipment, the system will adjust the consequence level according to the equipment characteristics. For example, if a historical case caused equipment burnout, the consequence may be downgraded to "equipment shutdown" due to the addition of new protection.

[0096] The automatic generation of a structured report refers to the following: The report uses a standardized template defined by XML Schema and includes: a risk summary, including risk ID, equipment GUCI, current health index HI value, and risk level (red / orange / yellow / blue four levels); comparison and tracing, including a table showing the commonalities and differences between the current risk point and matching cases, with confidence levels marked; evolutionary projection, with a Gantt chart showing the time nodes and probability distribution of each evolution stage; impact analysis, including a topology highlight map showing the range of affected equipment, with an IEC 61850 logical link diagram; and handling recommendations, including executable operation instructions generated in priority order, such as: immediate execution items, such as "disconnect the connection of terminal blocks X1:13-X1:14 behind the protection panel of #2 main transformer A"; items to be executed within 2 hours, such as "arrange for infrared thermometry to retest the terminal temperature"; and planned rectification items, such as "replace this batch of terminal blocks, model: WUK-2.5 / 12-ST-5.08".

[0097] In this example, based on fault risk point identification, the system overcomes the limitations of traditional static threshold alarms by constructing a closed-loop analysis chain of comparison, analysis, deduction, and decision-making. By employing a multi-dimensional, item-by-item comparison mechanism, it overcomes the false alarm and missed alarm problems caused by single-parameter threshold methods. By integrating historical case evolution path knowledge, it avoids trend distortion caused by relying solely on real-time data extrapolation. By using differentiated analysis to drive path reweighting, it solves the deficiency of general evolution models in adaptability under specific operating conditions. And by automatically generating structured reports, it eliminates bottlenecks such as low efficiency, inconsistent expression, and poor operability of manual judgment and handling suggestions. Ultimately, the system leaps from detecting anomalies to predicting failures, significantly improving the foresight, accuracy, and executability of risk prevention and control in substation secondary systems, achieving the goal of strengthening secondary equipment status assessment and risk early warning.

[0098] This embodiment of the substation secondary full-circuit search method based on the FBS algorithm uses the FBS algorithm as its core engine. It constructs a hierarchical search space through forward probing, compressing the entire network traversal into inter-layer relationship retrieval, thus solving the problems of low search efficiency and high resource consumption. Through backward tracking, it accurately locates terminal-level circuit paths, avoiding path ambiguity and redundancy inherent in DFS / BFS, and solving the problem of inaccurate positioning. By fusing physical, electrical, and functional information into a three-dimensional model and using shadow mode safety simulation, it transforms the abstract topology into an interactive and verifiable visual model, solving the problems of low efficiency and poor accuracy in physical circuit model visualization. Through deep coupling analysis of loop topology maps and multi-source device data, it upgrades static path finding to dynamic risk deduction, solving the problem of lacking the ability to quickly generate full-circuit paths and support three-dimensional visualization by combining defect alarms. Ultimately, while ensuring the completeness and accuracy of search results, the time taken to search the secondary full circuit of a typical substation was reduced from tens of seconds in DFS / BFS to hundreds of milliseconds in FBS, and the number of searches was reduced by more than an order of magnitude. At the same time, it supports a risk warning lead time of up to 72 hours, which significantly improves the intelligence, refinement and initiative of secondary system operation and maintenance.

[0099] Based on the same inventive concept, the embodiments also disclose an early warning device for substation secondary full-circuit search based on the FBS algorithm, see attached figure. Figure 6 and 7 As shown, it includes the following modules: Acquisition Module 100: Acquires the search source point and target point based on the initialized search parameters. The search parameters include information on all equipment in the substation and their interconnection relationships. Strategy formulation module 200: Establishes a search hierarchy path based on the search source point and target point, and formulates a search strategy based on the search hierarchy path; Loop path determination module 300: Determines the target loop path of the target point in reverse order based on the search hierarchy path; Visualization Module 400: Builds a loop topology map based on the loop path and performs visualization processing on the loop topology map to obtain a visualization model; Risk prediction module 500: Analyzes loop topology diagrams and equipment data to predict potential risks.

[0100] In this embodiment, loosely coupled communication is achieved through a standardized data interface, sharing a unified device-connection relationship graph database and a real-time status cache. The device and connection relationship data output by the acquisition module 100 serves as the global basic data source. The strategy formulation module 200 and the loop path determination module 300 share the same hierarchical index structure. The visualization module 400 and the risk prediction module 500 share loop topology graph instances and device status snapshots. Data flow between modules follows an event-driven mechanism: after the acquisition module 100 completes parameter loading, it automatically triggers the strategy formulation module to initiate forward probing; after the strategy formulation module 200 completes hierarchical path construction, it publishes a "path ready" event, driving the loop path determination module 300 to perform backward tracing; after the loop path determination module 300 outputs the target loop path, it notifies the visualization module to generate a topology graph; after the visualization module 400 completes the digital twin model construction and enters the shadow mode ready state, it pushes the initial topology snapshot to the risk prediction module 500, initiating the risk assessment process.

[0101] In the implementation process, upon receiving input from maintenance personnel, such as the source point of terminal K1 of the secondary winding of the A-phase current transformer on the high-voltage side of the #1 main transformer, and the target point of the trip output terminal of the protection device of the 220kV bus tie 212 circuit breaker, the acquisition module quickly parses the SDD file to obtain the equipment and connection relationships of the entire station; the strategy formulation module completes the hierarchical division within 3 layers and generates a forward search strategy; the loop path determination module traces back upward along the recorded inter-layer connection relationships, and outputs a complete trip loop path containing 17 equipment nodes and 16 cable segments within 500ms; visual The module renders a 3D loop topology map in real time based on the path, overlays the physical wiring path and GOOSE signal flow arrows, and simulates the trip command propagation process in shadow mode. The risk prediction module simultaneously retrieves real-time temperature, partial discharge, and communication quality data from 17 devices, identifies three terminal block nodes with a slow upward trend in contact resistance, matches them with the "terminal oxidation leading to failure to operate" fault mode in the historical case library, predicts a 68% probability of failure to operate within 72 hours, and automatically generates a structured early warning report containing handling suggestions such as cleaning and tightening + infrared retesting. The implementation adopts a modular layered design and FBS algorithm-driven approach, avoiding the exponential search expansion of traditional BFS / DFS in a network of thousands of nodes, and solving the core problems of low search efficiency, inaccurate path positioning, and lack of risk extrapolation. This achieves the technical effects of fast response, accurate path reconstruction, early risk warning, and strong handling basis for substation secondary full-circuit search.

[0102] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.< / sampledvaluecontrol> < / goosecontrol>

Claims

1. A search method for the secondary full circuit of a substation based on the FBS algorithm, characterized in that, Includes the following steps: The search source point and target point are obtained based on the initialized search parameters, which include information on all equipment in the substation and their interconnection relationships. Establish a search hierarchy path based on the search source point and target point, and formulate a search strategy based on the search hierarchy path; Determine the target loop path of the target point based on the reverse path of the search hierarchy path; A loop topology diagram is established based on the loop path, and the loop topology diagram is visualized to obtain a visualization model.

2. The search method according to claim 1, characterized in that, The process of obtaining the search source point and target point based on the initialized search parameters involves: constructing a substation secondary digital model using SDD / SPD files, and obtaining the substation's equipment information and interconnection information from the substation secondary digital model. Obtain search source and target point information according to the received instructions; The hierarchical relationship of the search is determined based on the connection between the search source point and the target point, which is used to constrain the search depth.

3. The search method according to claim 1, characterized in that, The process of establishing a search hierarchy path based on the search source point and target point, and formulating a search strategy based on the search hierarchy path, is as follows: Obtain the associated devices that are connected to the search source point and the target point, classify the associated devices according to the progressive connection relationship from the search source point to the target point, and assign the classified devices to the layer sets determined according to the hierarchical relationship to obtain a multi-layer structured search set arranged in sequence according to the progressive connection relationship from the search source point to the target point; The search hierarchy path is obtained based on the arrangement order of the multi-layered search set; Based on the search hierarchy path, a forward search strategy is formulated, starting from the set of layers where the search source point is located and sequentially searching towards the set of layers where the target point is located.

4. The search method according to claim 3, characterized in that, The step of classifying associated devices according to the progressive connection relationship from the search source point to the target point, and then assigning the classified devices to the hierarchical sets determined according to the hierarchical relationship, is as follows: Determine the number of search levels for the multi-level structured search set based on the hierarchical relationship; The associated devices are classified according to the order in which they are connected from the search source point to the target point, and the classified associated devices are respectively assigned to the various layers of the multi-layer structure search set.

5. The search method according to claim 1, characterized in that, The method for determining the target loop path of the target point based on the reverse path of the search hierarchy path is: Determine the layer set where the target point is located; According to the reverse search strategy of the forward search strategy, backtracking starts from the layer set where the target point is located and continues until the layer set where the search source point is located. Based on the connection relationships of each associated device in the backtracking path, the connection path from the target point to the search source point is obtained, thus obtaining the loop path.

6. The search method according to claim 1, characterized in that, The process of establishing a loop topology map based on the loop path and then visualizing the loop topology map yields the following visualization model: Based on the loop path, the physical layer information, electrical layer information, and functional layer information of all devices in the path are analyzed, and a multi-dimensional digital twin model is constructed by fusing the physical layer information, electrical layer information, and functional layer information for visualizing the loop topology. At the same time, a shadow mode is created, isolated from the running system, to ensure the security of all planning and simulation operations; The loop topology diagram is color-coded to distinguish the roles and running states of the shadow mode loop topology diagram, and then the operation process is simulated using dynamic flow.

7. The search method according to claim 6, characterized in that, The creation of a shadow mode isolated from the running system is as follows: Based on the loop topology diagram, construct a clone model containing information data of all devices in the loop; Next, obtain the dynamic status data of all devices in the loop topology diagram. The dynamic status data includes real-time health index, operating parameters, and alarm status. The dynamic state data is fused into the clone model to obtain the shadow pattern.

8. The search method according to claim 1, characterized in that, The search method further includes: analyzing the loop topology diagram and device data to predict potential risks; Among them, the analysis based on the loop topology diagram and equipment data to predict potential risks is as follows: Acquire real-time monitoring data from all devices and combine it with the visualization model to evaluate and display the device status, thereby predicting potential device failure risks in advance. The fault risk points are matched with historical fault data in the database to obtain fault cases; The fault risk points are compared and analyzed with the fault cases, and the potential risk points of the system are predicted based on the comparison results.

9. The search method according to claim 8, characterized in that, The fault risk points are compared and analyzed with the fault cases. Based on the comparison results, the potential risk points of the system are predicted as follows: the current fault risk points and the matched fault cases are compared item by item, their commonalities and differences are analyzed, and combined with the known evolution path of the fault cases, the development trend and possible consequences of the current fault risk points are predicted, and a structured report is automatically generated.

10. An early warning device for substation secondary full-circuit search based on FBS algorithm, characterized in that, Includes the following modules: Acquisition module: Acquires the search source point and target point based on the initialized search parameters, which include information on all equipment in the substation and their interconnection relationships; Strategy formulation module: Establishes a search hierarchy path based on the search source point and target point, and formulates a search strategy based on the search hierarchy path; Loop path determination module: Determines the target loop path of the target point by reverse-engineering the search hierarchy path; Visualization module: Establishes a loop topology map based on the loop path, and performs visualization processing on the loop topology map to obtain a visualization model.

Images