Power system operation and maintenance method and system, electronic device, storage medium and product

Abstract

The application discloses a power system operation and maintenance method and system, electronic equipment, storage medium and product, and relates to the technical field of operation and maintenance, and comprises the following steps: in response to an alarm event, analyzing the alarm event to obtain a facility identifier; obtaining a first node corresponding to the facility identifier from a preset knowledge graph; traversing the knowledge graph from the first node as a starting point, determining a second node and a power supply state of the second node according to the electrical on-off state of each node; determining the severity of the second node based on the power supply path where the second node is located; and performing operation and maintenance on the power system based on the second node, the power supply state and the severity. The application has the advantages of improving the operation and maintenance efficiency of the power system and reducing the risk of delayed fault handling caused by human errors.

Description

Technical Field

[0001] This application relates to the field of operation and maintenance technology, and in particular to power system operation and maintenance methods, systems, electronic devices, storage media and products. Background Technology

[0002] In existing power monitoring systems, when power facilities trigger tripping or other alarm events, the determination of the fault's impact range heavily relies on the subjective experience and manual operation of maintenance personnel. These personnel must spend a significant amount of time reviewing complex electrical wiring diagrams and load distribution documents, verifying the interruption of power supply paths downstream of the fault point one by one. This manually-driven analysis method is not only inefficient but also highly susceptible to overlooking critical business loads due to human error when dealing with the multi-layered topology of modern power networks. This leads to delayed fault handling and may potentially trigger cascading business interruptions or equipment damage. Summary of the Invention

[0003] The main purpose of this application is to provide a power system operation and maintenance method, system, electronic equipment, storage medium and product, which aims to improve the efficiency of power system operation and maintenance and reduce the risk of untimely fault handling caused by human error.

[0004] To achieve the above objectives, this application proposes a power system operation and maintenance method, the method comprising: In response to an alarm event, the alarm event is parsed to obtain a facility identifier, wherein the facility identifier is the identifier of the power facility in the power system that triggered the alarm event; The first node corresponding to the facility identifier is obtained from the preset knowledge graph, wherein the knowledge graph includes nodes corresponding to each power facility in the power system and the relationship between each node, and the relationship between each node is mapped according to the power supply rules of the power system; Starting from the first node, the knowledge graph is traversed, and the power supply status of the second node and the second node are determined according to the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node. Based on the power supply path in which the second node is located, the severity of the second node is determined, wherein the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power outage conditions when a fault propagates to the second node; wherein the step of determining the severity of the second node based on the power supply path in which the second node is located includes: determining a probability value based on the power supply path in which the second node is located, wherein the probability value characterizes the likelihood of a fault propagating to the second node; multiplying the load utilization rate of the second node by the service level weight as the loss value of the second node, wherein the load utilization rate characterizes the current load level of the power facility corresponding to the second node, the service level weight characterizes the importance of the services carried by the power facility corresponding to the second node, and the loss value characterizes the degree of business impact caused by power outage of the power facility corresponding to the second node under fault conditions; and multiplying the probability value by the loss value as the severity of the second node; The power system is operated and maintained based on the second node, the power supply status, and the severity.

[0005] In one feasible embodiment, before the step of obtaining the first node corresponding to the facility identifier from a preset knowledge graph, the method further includes: Obtain the load table of the power system, parse the load table to obtain the attribute information of each power facility and the connection relationship between each power facility, wherein the attribute information characterizes the electrical characteristics and connection location of the power facility; Based on the attribute information and the preset node template, nodes corresponding to each of the power facilities are generated. The types of the node template include at least the upstream busbar of the municipal incoming line, municipal incoming line, medium voltage incoming line cabinet, medium voltage feeder cabinet, transformer, low voltage incoming line cabinet, low voltage feeder cabinet, loads at all levels, plug-in interface, machine and machine component. Based on the connection relationship and the preset power supply rules, the relationship between each node is established. The power supply rules include normal power supply link conduction rules, mutual backup rules between dual backup devices, and switching interlocking rules between mains power and emergency power. The relationship between each node includes at least downstream relationship, upstream relationship, inclusion relationship, belonging relationship, mutual backup relationship, and interlocking relationship. Each node and its relationship are verified, and the nodes and relationships that pass the verification are written into the knowledge graph.

[0006] In one feasible embodiment, the step of establishing the relationship between the nodes based on the connection relationship and the preset power supply rules includes: Based on the aforementioned connection relationships, downstream relationships, upstream relationships, inclusion relationships, and membership relationships are established. If the connection relationship indicates that the mutual backup attribute identifiers of the two power facilities are the same, then a mutual backup relationship is established between the nodes corresponding to the two power facilities according to the mutual backup rule; If the connection relationship indicates that the mains power incoming cabinet and the emergency power incoming cabinet exist in pairs, then an interlocking relationship is established between the node corresponding to the mains power incoming cabinet and the node corresponding to the emergency power incoming cabinet according to the switching interlocking rule.

[0007] In one feasible embodiment, the step of traversing the knowledge graph starting from the first node, determining the second node and its power supply status based on the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node, includes: Starting from the first node, perform a topological traversal along the relationships in the knowledge graph; For the current node being traversed, if the electrical on / off state of the downstream node of the current node is on, then the downstream node of the current node is determined as the second node, the power supply state of the downstream node of the current node is marked as off, and the downstream node of the current node is determined as the traversal node and the traversal continues downward. If the electrical connection status of the current node is disconnected or tripped, then query the knowledge graph for backup nodes that have the backup relationship with the current node. If the electrical connection status of the backup node is on, then determine the downstream node of the backup node as the second node, mark the power supply status of the downstream node of the backup node as a transfer power supply status, and continue to traverse downwards after determining the downstream node of the backup node as a traversal node. If the electrical on / off state of the current node is disconnected or tripped, the node type is a medium-voltage incoming cabinet, and the power supply type is mains power, then query the knowledge graph for emergency power supply nodes that have the interlocking relationship with the current node. If the electrical on / off state of the emergency power supply node is on, then determine the downstream node of the emergency power supply node as the second node, mark the power supply state of the downstream node of the emergency power supply node as emergency power supply state, and continue to traverse downwards after determining the downstream node of the emergency power supply node as the traversal node. After the traversal is complete, output the power supply status of each second node.

[0008] In one feasible embodiment, the step of determining the probability value based on the power supply path in which the second node is located includes: If the second node is connected to the first node through a single power supply path, then the product of the conduction probabilities of each node on the power supply path is calculated as the probability value. If the second node is connected to the first node through multiple power supply paths, then calculate a probability value by subtracting the probability that each of the power supply paths is not conductive.

[0009] In one feasible embodiment, the step of performing operation and maintenance on the power system based on the second node, the power supply status, and the severity includes: Obtain fault context data, wherein the fault context data includes at least the electrical on / off status of the first node, the power supply path of the first node, and historical case records similar to the alarm event; The fault context data is converted into a vector representation, and a similarity search is performed in the vector database to obtain the search results. The vector database is obtained by vectorizing expert knowledge documents. The fault context data is input into the rule engine and matched with preset diagnostic rules to calculate the rule matching degree. The diagnostic rules are obtained by transforming the threshold conditions and topological features in the expert knowledge document. The fault context data, the search results, and the rule matching degree are input into a large language model for fusion reasoning to generate a first diagnostic result. The confidence level of the first diagnostic result is calculated to obtain the posterior probability, and the posterior probability and the first diagnostic result are used as the diagnostic result corresponding to the alarm event.

[0010] In one feasible embodiment, the step of calculating the confidence level of the first diagnostic result to obtain the posterior probability includes: The rule matching degree is mapped to a first conditional probability, the similarity of the search results is mapped to a second conditional probability, and the completeness of the fault context data is mapped to a third conditional probability. Under the conditional independence assumption, the logarithms of the first conditional probability, the second conditional probability, and the third conditional probability are taken and weighted and summed to obtain the fused logarithmic probability. The logarithmic probability is then normalized to obtain the posterior probability.

[0011] Furthermore, to achieve the above objectives, this application also proposes a power system operation and maintenance system, which includes: The monitoring layer is used to respond to alarm events, parse the alarm events to obtain facility identifiers, wherein the facility identifiers are the identifiers of the power facilities in the power system that triggered the alarm events; The knowledge layer is used to store a preset knowledge graph and obtain the first node corresponding to the facility identifier from the knowledge graph. The knowledge graph includes nodes corresponding to each power facility in the power system and the relationships between the nodes. The relationships between the nodes are mapped according to the power supply rules of the power system. A twin layer, connected to the monitoring layer, is used to maintain the electrical on / off status of each node; An intelligent layer, connected to the knowledge layer and the twin layer, is used to traverse the knowledge graph starting from the first node, determine the power supply status of the second node based on the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node; and determine the severity of the second node based on the power supply path in which the second node is located, wherein the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power failure conditions when the fault propagates to the second node; wherein the intelligent layer is used to determine a probability value based on the power supply path in which the second node is located, wherein the probability value characterizes the possibility of the fault propagating to the second node; and take the product of the load utilization rate of the second node and the business level weight as the loss value of the second node, wherein the load utilization rate characterizes the current load level of the power facility corresponding to the second node, the business level weight characterizes the importance of the business carried by the power facility corresponding to the second node, and the loss value characterizes the degree of business impact caused by the power interruption of the power facility corresponding to the second node under fault conditions; and take the product of the probability value and the loss value as the severity of the second node; The application layer, connected to the intelligent layer, is used to perform operation and maintenance on the power system based on the second node, the power supply status, and the severity.

[0012] In addition, to achieve the above objectives, this application also proposes an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the power system operation and maintenance method as described above.

[0013] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the power system operation and maintenance method described above.

[0014] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the power system operation and maintenance method described above.

[0015] In this application, in response to an alarm event, the alarm event is parsed to obtain a facility identifier, where the facility identifier is the identifier of the power facility that triggered the alarm event in the power system; from a pre-defined knowledge graph, the first node corresponding to the facility identifier is obtained, where the knowledge graph includes nodes corresponding to each power facility in the power system and the relationships between nodes, the relationships between nodes being mapped according to the power supply rules of the power system; starting from the first node, the knowledge graph is traversed, and the power supply status of the second node and the second node is determined according to the electrical on / off status of each node, where the second node is a node on the fault propagation path of the first node; based on the power supply path where the second node is located, the severity of the second node is determined, where the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power outage conditions when the fault propagates to the second node; based on the second node, the power supply status, and the severity, the power system is operated and maintained.

[0016] This application triggers fault analysis through alarms, combines a pre-set power knowledge graph that integrates power supply rules, traverses the graph starting from the node corresponding to the faulty power facility, determines the nodes and corresponding power supply status on the fault propagation path by combining real-time electrical status, and further assesses the fault severity of each node before carrying out operation and maintenance. It solves the technical problems of low efficiency of existing manual analysis, inability of existing digital solutions to distinguish the functional differences of different power supply paths, and inability to complete accurate fault analysis by combining real-time status. It has the advantages of automatically analyzing the impact range of power faults, improving the operation and maintenance efficiency of power systems, and reducing the risk of untimely fault handling caused by human error. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the power system operation and maintenance method of this application. Figure 2 This is a schematic diagram of the process for determining a serious situation in an embodiment of the power system operation and maintenance method of this application; Figure 3 This is a flowchart illustrating Embodiment 2 of the power system operation and maintenance method of this application. Figure 4This is a schematic diagram of knowledge graph construction provided for an embodiment of the power system operation and maintenance method of this application; Figure 5 A simplified flowchart is provided for one embodiment of the power system operation and maintenance method of this application; Figure 6 This is a flowchart illustrating Embodiment 4 of the power system operation and maintenance method of this application; Figure 7 This is a schematic diagram of the architecture of the power system operation and maintenance system in the embodiments of this application; Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the power system operation and maintenance method in this application embodiment.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions. The following description uses an electronic device as an example to illustrate this embodiment and the subsequent embodiments.

[0024] Based on this, the embodiments of this application provide a power system operation and maintenance method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the power system operation and maintenance method of this application. In this embodiment, the power system operation and maintenance method includes steps S10 to S50: Step S10: In response to an alarm event, the alarm event is parsed to obtain the facility identifier, wherein the facility identifier is the identifier of the power facility that triggered the alarm event in the power system.

[0025] Alarm events are data records generated by a power monitoring system when it detects electrical parameters exceeding preset thresholds or receives activation signals from protective devices. Conditions triggering alarm events include current exceeding rated values, voltage deviating from normal ranges, abnormal power output, frequency fluctuations, switch status changes, and protective relay tripping. The power monitoring system collects electrical quantities through sensors, smart meters, and protective devices deployed in substations and distribution rooms. When the collected values ​​meet the alarm triggering conditions, the system generates an alarm event data packet containing a timestamp, device location, alarm type, and fault source information.

[0026] A power system is an electrical energy production and consumption system composed of power generation, transmission, transformation, distribution, and consumption. Power facilities are the specific devices in a power system that perform the functions of transmitting, transforming, distributing, or using electrical energy, including busbars, incoming line cabinets, feeder cabinets, transformers, load switches, and electrical equipment. A facility identifier is a unique identification code assigned to each power facility, used to locate specific equipment in monitoring systems and knowledge graphs. The encoding format of facility identifiers includes equipment asset numbers, process location codes, or unique serial numbers assigned by the system, without specific restrictions.

[0027] The alarm event data packet is read, and the fault source device identifier field is extracted from the data packet. This field value is then output as the facility identifier. In one feasible implementation, the parsing process extracts the device ID field from the alarm data in JSON or XML format through string matching; in another feasible implementation, the parsing process reads the device encoding field from the binary alarm message using a predefined data structure offset.

[0028] Step S20: Obtain the first node corresponding to the facility identifier from the preset knowledge graph. The knowledge graph includes the nodes corresponding to each power facility in the power system and the relationships between the nodes. The relationships between the nodes are mapped according to the power supply rules of the power system.

[0029] A knowledge graph is a graph data organization structure composed of nodes and the relationships connecting them. In the context of power system operation and maintenance, each node in the knowledge graph corresponds one-to-one with a power facility, and each node stores the type, attributes, and identification information of the corresponding power facility. The relationships in the knowledge graph represent the electrical connections or functional associations between power facilities, including energy transfer direction, backup associations, and power switching constraints.

[0030] Power supply rules are the physical power supply logic of a power system under normal operation and fault conditions. These rules include normal power supply link conduction rules, mutual backup rules between dual backup devices, and switching interlocking rules between mains power and emergency power. Normal power supply link conduction rules characterize the direction and path of electrical energy transfer from upstream equipment to downstream loads. Mutual backup rules characterize the redundancy relationship between two or more power facilities, allowing backup facilities to take over power supply when the primary facility is out of service. Switching interlocking rules characterize the interlocking relationship between the primary power supply and the emergency power supply, automatically activating the emergency power supply or allowing manual activation when the primary power supply is disconnected.

[0031] The first node is the node in the knowledge graph corresponding to the power facility that triggered the alarm event, i.e., the fault source node. The process of obtaining the first node involves querying the knowledge graph to find nodes whose facility identifiers in the node attributes match the parsed facility identifiers. In one feasible implementation, the process uses a graph database query language to perform attribute filtering queries, locating nodes with matching facility identifiers in the node attribute index. In another feasible implementation, the process directly looks up the internal number of the node corresponding to the facility identifier using a hash mapping table, and then uses this internal number to locate the node object in the knowledge graph.

[0032] Step S30: Traverse the knowledge graph starting from the first node, and determine the power supply status of the second node and the second node based on the electrical on / off status of each node. The second node is a node on the fault propagation path of the first node.

[0033] Electrical on / off status refers to the electrical state of the power facilities corresponding to a node, including normal on, off, and tripped states. Normal on state indicates that the equipment is energized and current can flow normally. Off state indicates that the equipment is electrically isolated due to tripping or power loss. Tripped state indicates that the equipment automatically disconnects the circuit due to protection action. Electrical on / off status is collected in real-time by the power monitoring system and stored in the status database.

[0034] The second node is a node affected by the fault source node along the fault propagation path. Power supply status is a classification of the power source for the second node under fault conditions, including power outage status, backup power supply status, and emergency power supply status. Power outage status indicates that the node cannot obtain power through the primary power supply path. Backup power supply status indicates that the node obtains backup power through mutual backup facilities. Emergency power supply status indicates that the node obtains backup power through an emergency power source.

[0035] Starting from the first node, the traversal visits adjacent nodes along the relationships between nodes. The traversal algorithm is not limited to a specific implementation and can be set according to actual needs. In one feasible implementation, a breadth-first search algorithm is used for traversal, with a first-in-first-out queue to manage the nodes to be processed, and visited nodes are marked to avoid duplicate processing. In another feasible implementation, a depth-first search algorithm is used for traversal, with a stack structure to manage the nodes to be processed, and after traversing along a single path to the end, backtracking to the branch point to continue traversing.

[0036] Step S40: Based on the power supply path where the second node is located, determine the severity of the second node, where the severity characterizes the degree of business loss caused by the power facilities corresponding to the second node under power failure conditions when the fault propagates to the second node.

[0037] A power supply path is a route from the first node to the second node, consisting of a series of sequentially connected nodes and relationships. The process of determining the power supply path where the second node is located includes recording the sequence of nodes and relationships traversed during the traversal from the first node to the second node, or obtaining path information by backtracking the connection relationship between the second node and the first node in the knowledge graph.

[0038] Severity is a numerical indicator that quantifies the impact of a fault on a second node. In one feasible implementation, determining severity involves determining a probability value and a loss value, then multiplying the probability value by the loss value to obtain the severity. The probability value characterizes the likelihood of a fault propagating to the second node and is calculated based on the electrical on / off status of each node along the power supply path. The loss value characterizes the degree of business impact caused by the power facilities corresponding to the second node under power outage conditions. In another feasible implementation, the loss value can also be used as the severity. The specific settings can be configured according to actual needs and are not limited here.

[0039] Please refer to Figure 2 Step S40, the step of determining the severity of the second node based on the power supply path in which the second node is located, includes: Step S401: Determine the probability value based on the power supply path where the second node is located, where the probability value represents the possibility of a fault propagating to the second node; The probability value is a numerical parameter that quantifies the likelihood of a fault propagating from the first node to the second node. The probability value ranges from zero to one; the closer the probability value is to one, the higher the certainty of the fault propagating to the second node. The closer the probability value is to zero, the lower the probability of the fault propagating to the second node.

[0040] A power supply path is a pathway formed by sequentially connecting nodes and relationships between a first node and a second node. Power supply paths can be single or multiple. A single power supply path means the second node receives power from the first node through a single pathway; multiple power supply paths mean the second node receives power from the first node through two or more independent pathways, with mutual backup or emergency redundancy between the pathways. For a single power supply path, the probability value is determined based on the conduction status of each node along the path. All nodes on the path must remain conductive for a fault to propagate along the path to the second node. If any node on the path is disconnected or tripped, the propagation probability of that path is zero. For multiple power supply paths, the probability value is determined by combining the conduction status of each independent power supply path. A fault must simultaneously disconnect all power supply paths to affect the second node. If at least one power supply path remains conductive, the probability of the fault propagating to the second node is significantly reduced.

[0041] In one feasible embodiment, step S401, the step of determining the probability value based on the power supply path where the second node is located, includes: Step S4011: If the second node is connected to the first node through a single power supply path, calculate the product of the conduction probabilities of each node on the power supply path as the probability value. The second node is connected to the first node via a single power supply path, meaning there is only one path between the first and second nodes, formed by sequentially connecting nodes and relationships. There are no alternative paths provided by backup nodes or emergency power supply nodes along this path. The second node relies solely on this single path to obtain power. In this single power supply path, the nodes are connected in series in terms of power supply reliability. Power can only be transferred from the first node to the second node when all nodes on the path are conducting. If any node on the path is disconnected or tripped, the path is interrupted, and the second node cannot obtain power through this path.

[0042] The calculation of the product of the conduction probabilities of each node along the power supply path, as the probability value, is based on the reliability principle of series systems. The conduction probability is the probability that a node will maintain electrical continuity under fault conditions. In a single power supply path, a fault propagating from the first node to the second node requires passing through every intermediate node along the path. Only when all nodes on the path remain conductive can the fault propagate completely to the second node. Therefore, the conduction probability of the entire path is equal to the product of the conduction probabilities of each node along the path. If any node on the path is in an open or tripped state, the conduction probability of that node is zero, resulting in a zero propagation probability for the entire path.

[0043] In a specific implementation, the number of power supply paths between the second node and the first node can be determined. If only a single power supply path exists, the sequence of all nodes on that path is extracted. The conduction probability of each node on the path is queried or calculated. The conduction probabilities of all nodes are multiplied together, and the resulting product is used as the probability value of the fault propagating to the second node. This probability value directly reflects the likelihood of the fault propagating to the second node along the only path.

[0044] Step S4012: If the second node is connected to the first node through multiple power supply paths, calculate a probability value by subtracting the probability that each power supply path is not conductive.

[0045] The second node is connected to the first node through multiple power supply paths, meaning that there are two or more independent paths between the first and second nodes. These paths may include a main power supply path and a backup path, or a main power supply path and an emergency power supply path. Each path is electrically independent; the disconnection of one path does not affect the continuity of the others. As long as at least one path remains continuous, the second node can maintain power supply.

[0046] The probability value is calculated by subtracting the probability that all power supply paths are not conductive from the probability of 1, based on the reliability principle of parallel systems. For multiple power supply paths, the probability that a fault propagates to the second node is equal to the complement of the probabilities that all power supply paths are simultaneously not conductive. The probability that each power supply path is not conductive is equal to 1 minus the probability that the path is conductive. The probability that all power supply paths are not conductive is equal to the product of the probabilities that all paths are not conductive. Subtracting this product from 1 yields the probability that at least one power supply path remains conductive. This probability value is the probability that the fault propagates to the second node.

[0047] First, determine the number of power supply paths between the second node and the first node. When multiple power supply paths are determined, calculate the conduction probability of each power supply path. Convert the conduction probability of each path into its non-conductivity probability. Multiply the non-conductivity probabilities of each path together, and then subtract the product from one to obtain the probability value of the fault propagating to the second node. This probability value reflects the possibility that the fault will affect the second node after simultaneously cutting off all redundant paths.

[0048] Step S402: The product of the load utilization rate of the second node and the service level weight is used as the loss value of the second node. The load utilization rate represents the current load level of the power facility corresponding to the second node, the service level weight represents the importance of the service carried by the power facility corresponding to the second node, and the loss value represents the degree of business impact caused by the power outage of the power facility corresponding to the second node under fault conditions. Load utilization rate is the ratio of the real-time operating load of the power facility corresponding to the second node to its rated capacity. The calculation process for load utilization rate can be as follows: obtain the real-time power value of the power facility from the real-time status data, obtain the rated capacity value of the power facility from the attribute information, and divide the real-time power value by the rated capacity value to obtain the load utilization rate. Load utilization rate characterizes the degree of load saturation of the power facility at the current moment; the higher the load utilization rate, the heavier the production task the power facility is undertaking when a fault occurs.

[0049] Business level weights are pre-configured level coefficients based on the criticality of the business carried by the power facility corresponding to the second node. The configuration process for business level weights can be as follows: based on the functional positioning of the power facility in the production process, the capacity loss caused by downtime, and the scope of its impact on upstream and downstream processes, a level identifier and corresponding weight value are assigned to the power facility. A higher business level weight indicates a greater impact of the business carried by the power facility on the overall production system.

[0050] The loss value is a parameter that quantifies the impact on services caused by power outages due to fault conditions on the power facilities corresponding to the second node. The loss value is calculated by multiplying the load utilization rate by the service level weight. The calculation process involves reading the load utilization rate of the second node and the service level weight of the power facilities corresponding to that node, and multiplying the two values ​​to obtain the loss value. The loss value characterizes the overall impact of power loss on services at the second node, assuming the fault has propagated to it. The load utilization rate and service level weight jointly determine the magnitude of the loss value; nodes with high load and high service levels have higher loss values.

[0051] Step S403: The product of the probability value and the loss value is used as the severity of the second node.

[0052] Severity is a quantitative indicator that combines the probability of fault propagation with the degree of business loss. The severity is calculated by multiplying the probability value by the loss value. The product of the probability and loss values ​​represents the comprehensive risk of the fault's impact; the probability value reflects the likelihood of the fault reaching the node, and the loss value reflects the degree of business impact after the node loses power. Multiplying the two considers both the electrical probability of fault propagation and the business consequences of a node losing power. For nodes where the primary power supply path is definitively interrupted and cannot be restored through backup or emergency paths, the probability value approaches one, and the severity is primarily determined by the loss value. For nodes with backup or emergency paths that are operational, the probability value approaches zero, and the severity decreases accordingly.

[0053] This embodiment couples the real-time electrical probability of fault propagation with the dynamic quantification of service loss to form a severity index in the form of expected loss. This index reflects whether a fault is likely to reach a certain node, and more importantly, it reflects the actual service impact caused by the power outage of that node. This allows maintenance personnel to quantify and sort multiple affected nodes based on the severity value, prioritizing the handling of nodes with high severity. This enables the precise allocation of limited maintenance resources to the most critical fault-affected links, avoiding priority mismatch or omission of critical loads due to human experience judgment. This improves the efficiency of power system maintenance and reduces the risk of untimely fault handling caused by human error.

[0054] Step S50: Perform operation and maintenance on the power system based on the second node, power supply status, and severity.

[0055] The operation and maintenance (O&M) of a power system is a process of performing subsequent processing based on the results of fault impact analysis. The O&M process is not limited here; for example, it can output an impact list, which includes a set of second-level nodes, the power supply status of each second-level node, and the severity of each second-level node, used to show O&M personnel the scope of the fault impact and the priority of handling. In one feasible implementation, the O&M process further includes operations such as generating fault diagnosis reports based on the impact list, but the specifics are not limited here.

[0056] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 Before step S20, which involves obtaining the first node corresponding to the facility identifier from a preset knowledge graph, the method further includes: Step S01: Obtain the load table of the power system, parse the load table to obtain the attribute information of each power facility and the connection relationship between each power facility, wherein the attribute information represents the electrical characteristics and connection location of the power facility.

[0057] A load table is a structured data file that records the configuration parameters and connection information of various power facilities in a power system. The load table is organized in tabular form, with each row corresponding to a power facility or a connection record between facilities. The sources of load tables include power system design documents, as-built drawings, equipment ledgers, or configuration export files from monitoring systems.

[0058] Attribute information is a set of parameters representing the electrical characteristics and connection locations of power facilities, obtained from the load table. Electrical characteristics include the equipment's rated capacity, rated voltage, rated current, and load type. Connection locations include the plant area, building, floor, room number, and coordinates in the system diagram. Attribute information is used to distinguish the physical characteristics and installation locations of different nodes in the knowledge graph.

[0059] Connection relationships are the associated data representing the electrical connections or hierarchical affiliation between power facilities, obtained from parsing the load table. Connection relationships include power supply links between upstream and downstream equipment, backup connections between equipment at the same level, and inclusion relationships between equipment and components. The process of parsing the load table can involve: reading the table file, identifying header fields, extracting cell data row by row, and mapping the extracted data to attribute information fields and connection relationship fields. In one feasible implementation, the parsing process is implemented through structured queries or script reading. In another feasible implementation, the parsing process uses a data conversion tool to uniformly convert heterogeneous load tables into a standard intermediate format before extracting attribute information and connection relationships.

[0060] Step S02: Based on attribute information and preset node templates, generate nodes corresponding to each power facility. The types of node templates include at least the upstream busbar of the municipal incoming line, municipal incoming line, medium-voltage incoming line cabinet, medium-voltage feeder cabinet, transformer, low-voltage incoming line cabinet, low-voltage feeder cabinet, loads at all levels, plug-in interfaces, equipment and equipment components.

[0061] A node template is a predefined model that defines the data structure of nodes in a knowledge graph. A node template specifies the type labels, attribute field sets, and data formats that a node should contain. Each node template corresponds to a type of power facility, ensuring that facilities of the same type have the same data structure in the knowledge graph.

[0062] The process of generating nodes can be as follows: For each power facility, select a node template that matches its type, fill the attribute fields of the node template with the parsed attribute information, and generate a node instance with a unique identifier, type label, and attribute values. The unique identifier is generated based on the combination of equipment location, power type, and serial number in the attribute information.

[0063] The node template types cover facility categories in the power system, including at least the upstream busbar of the municipal incoming line, municipal incoming line, medium-voltage incoming line cabinet, medium-voltage feeder cabinet, transformer, low-voltage incoming line cabinet, low-voltage feeder cabinet, loads at all levels, plug-in interfaces, equipment, and equipment components. The upstream busbar template of the municipal incoming line defines the attribute structure of the urban power grid access point. The municipal incoming line template defines the attribute structure of the main power supply access point of the power consumption area. The medium-voltage incoming line cabinet template defines the attribute structure of the incoming switchgear of the medium-voltage distribution system. The medium-voltage feeder cabinet template defines the attribute structure of the outgoing switchgear of the medium-voltage distribution system. The transformer template defines the attribute structure of voltage transformation equipment. The low-voltage incoming line cabinet template defines the attribute structure of the incoming switchgear of the low-voltage distribution system. The low-voltage feeder cabinet template defines the attribute structure of the outgoing switchgear of the low-voltage distribution system. The load templates at all levels define the attribute structure of the power loads at all levels. The plug-in interface template defines the attribute structure of the power distribution interface device. The equipment template defines the attribute structure of the production equipment. The equipment component template defines the attribute structure of the components of the production equipment.

[0064] Step S03: Based on the connection relationship and the preset power supply rules, establish the relationship between each node. The power supply rules include the normal power supply link conduction rules, the mutual backup rules between dual backup devices, and the switching interlocking rules between mains power and emergency power. The relationship between each node includes at least the downstream relationship, upstream relationship, inclusion relationship, belonging relationship, mutual backup relationship, and interlocking relationship.

[0065] Power supply rules are mapping criteria that transform the physical power supply logic of a power system into knowledge graph relationship types. Power supply rules originate from power system design specifications, operating procedures, and power supply architecture documents. Power supply rules define the logic for the transfer, switching, and transmission of electrical energy between devices under normal and fault conditions.

[0066] The process of establishing a relationship can be as follows: identifying the physical connections between nodes based on the connection relationships, determining the relationship type that the physical connection should be mapped to based on the power supply rules, and finally establishing a connection of that relationship type between the nodes. In one feasible implementation, the relationship establishment process is implemented through batch processing by a rule engine, which reads the connection relationships and power supply rules and outputs a relationship establishment instruction. In another feasible implementation, the relationship establishment process is implemented through conditional judgment; for each pair of nodes with a connection relationship, the conditions in the power supply rules are matched sequentially, the relationship type is determined, and then written.

[0067] Normal power supply link conduction rules characterize the direction and path logic of power transmission from upstream equipment to downstream load. Mutual backup rules between dual backup devices characterize the redundancy logic between two or more power facilities, where the backup facility takes over power supply when the primary facility is out of service. Switching interlock rules between mains power and emergency power represent the interlocking switching logic between the main power supply and emergency power supply, where the emergency power supply is activated when the main power supply is disconnected.

[0068] Downstream relationships are established according to normal power supply link conduction rules, representing the flow of electrical energy from upstream nodes to downstream nodes. Upstream relationships are established according to normal power supply link conduction rules, representing the traceability of electrical energy from downstream nodes to upstream nodes, and are the opposite of downstream relationships. Inclusion relationships are established according to equipment composition hierarchy, representing the relationship where a higher-level device includes a lower-level component. Belonging relationships are established according to equipment composition hierarchy, representing the relationship where a lower-level component belongs to a higher-level device, and are the opposite of inclusion relationships. Backup relationships are established according to the backup rules between dual-path backup devices, representing the relationship where two nodes provide backup power to each other. Interlocking relationships are established according to the interlocking rules for switching between mains power and emergency power, representing the interlocking switching relationship between mains power nodes and emergency power nodes.

[0069] Step S04: Verify each node and the relationships between each node, and write the nodes and relationships that pass the verification into the knowledge graph.

[0070] Validation is the process of checking the data integrity and logical consistency of generated nodes and established relationships. The purpose of validation is to identify and exclude data that does not conform to the physical logic of the power system and the specifications of the knowledge graph data. Writing is the process of persistently storing the validated nodes and relationships in the knowledge graph database.

[0071] Validating each node includes checking whether its unique identifier is globally unique, whether its naming conforms to predetermined rules, and whether its attribute fields are complete. Validating each relationship includes, but is not limited to, checking whether the starting and ending nodes of the relationship exist, whether the relationship type matches the node type combination, and whether the relationship forms a loop or isolated nodes that are not mutually redundant.

[0072] In one feasible implementation, the integrity check includes 11 check rules, specifically including: (1) Uniqueness check: the unique identifier of all nodes must be globally unique; (2) Naming consistency check: the naming rules of upstream and downstream nodes should be consistent; (3) Electrical logic consistency check: the power supply type of the primary load must match the upstream transformer; (4) Power supply link integrity check: all load nodes must have a complete power supply link; (5) Key node isolation detection: busbars and transformers must not exist in isolation; (6) Ring power supply detection: power supply loops are not allowed (except for mutual backup relationships); (7) Hierarchical relationship check: the power supply hierarchy must conform to the defined relationship order; (8) Field null value handling check: null value nodes in upstream and downstream relationships should be skipped correctly; (9) Transformer mutual backup chain integrity check: mutual backup transformers must appear in pairs; (10) Load capacity check: the load current must not exceed the capacity of the upstream equipment; (11) Voltage level matching check: the voltage levels of upstream and downstream must match.

[0073] In one feasible embodiment, step S03, the step of establishing the relationship between nodes based on the connection relationship and the preset power supply rules, includes: Step S031: Establish downstream relationships, upstream relationships, inclusion relationships, and membership relationships based on the connection relationships; Connection relationships are association information derived from load tables, representing the physical connections or hierarchical affiliations between power facilities. Connection relationships include power supply direction information between upstream and downstream equipment, and hierarchical affiliation information between the overall equipment and its components.

[0074] The process of establishing downstream relationships can be as follows: identify the power supply side node identifier and the power receiving side node identifier in the connection relationship; locate the node corresponding to the power supply side node identifier and the node corresponding to the power receiving side node identifier in the knowledge graph; and establish a downstream relationship from the power supply side to the power receiving side between the two. The downstream relationship represents the direction of electrical energy transmission from the upstream node to the downstream node.

[0075] The process of establishing an upstream relationship can be as follows: identify the power supply side node identifier and the power receiving side node identifier in the connection relationship; locate the node corresponding to the power supply side node identifier and the node corresponding to the power receiving side node identifier in the knowledge graph; and establish an upstream relationship from the power receiving side to the power supply side between the two. The upstream relationship and the downstream relationship constitute a pair of inverse relationships.

[0076] The process of establishing an inclusion relationship can be as follows: identify the overall device identifier and component identifier in the connection relationship; locate the nodes corresponding to the overall device identifier and the component identifier in the knowledge graph; and establish an inclusion relationship between the two, pointing from the whole to the component. An inclusion relationship represents the hierarchical containment of a higher-level device over a lower-level component.

[0077] The process of establishing a membership relationship can be as follows: identify the overall device identifier and component identifier in the connection relationship; locate the node corresponding to the overall device identifier and the node corresponding to the component identifier in the knowledge graph; and establish a membership relationship between the two, pointing from the component to the whole. Membership and inclusion relationships constitute a pair of inverse relationships.

[0078] Step S032: If the connection relationship indicates that the mutual backup attribute identifiers of the two power facilities are the same, then a mutual backup relationship is established between the nodes corresponding to the two power facilities according to the mutual backup rules. The mutual backup attribute identifier is an attribute parameter obtained from the load table that represents the mutual backup relationship of power facilities. The mutual backup attribute identifier is stored in the attribute information of the power facilities in coded form. Power facilities with the same mutual backup attribute identifier are configured as mutual backups in the physical power supply architecture.

[0079] The process of determining whether the mutual backup attribute identifiers of two power facilities are the same can be as follows: read the mutual backup attribute identifier field from the attribute information of the two power facilities and compare whether the values ​​of the two fields are equal. If the field values ​​are equal, then the two power facilities are determined to meet the attribute matching conditions in the mutual backup rules.

[0080] The process of establishing a backup relationship can be as follows: Locate the nodes corresponding to two power facilities with the same backup attribute identifier in the knowledge graph, and establish a backup relationship between the two nodes. The backup relationship is bidirectional, representing the association between the two nodes as backup power supplies for each other. The backup relationship does not represent the direction of power transmission, but rather represents the redundancy logic that when the power facility corresponding to one node goes out of service, the power facility corresponding to the other node can take over the power supply.

[0081] The role of mutual backup rules is to transform the physical power supply logic of dual backup in a power system into a relation type in a knowledge graph. Through mutual backup relationships, the knowledge graph can express the physical possibility of the power supply path shifting from the primary facility to the backup facility under fault conditions.

[0082] Step S033: If the connection relationship indicates that the mains power incoming cabinet and the emergency power incoming cabinet exist in pairs, then an interlocking relationship is established between the node corresponding to the mains power incoming cabinet and the node corresponding to the emergency power incoming cabinet according to the switching interlocking rules.

[0083] The indication that the mains power incoming cabinet and the emergency power incoming cabinet exist as a pair comes from the power configuration record in the connection relationship or attribute information. When the connection relationship indicates that both the mains power incoming cabinet and the emergency power incoming cabinet are configured in the same power supply circuit or the same power supply level, it is determined that the two exist as a pair.

[0084] The switching interlocking rule is the interlocking logic that characterizes the activation of the emergency power supply when the mains power supply is disconnected. The switching interlocking rule defines the electrical interlocking relationship between the mains power incoming cabinet and the emergency power incoming cabinet. When the node corresponding to the mains power incoming cabinet is in the disconnected state, the node corresponding to the emergency power incoming cabinet can be activated to maintain downstream power supply.

[0085] The process of establishing an interlocking relationship can be as follows: Locate the node corresponding to the mains power incoming cabinet and the node corresponding to the emergency power incoming cabinet in the knowledge graph, and establish an interlocking relationship between the two nodes. The interlocking relationship is bidirectional, representing the interlocking switching constraints between the two nodes. The interlocking relationship does not represent the normal direction of power transmission, but rather the physical constraint logic of power switching under fault conditions.

[0086] The role of interlocking relationships is to transform the electrical interlocking logic between mains power and emergency power in a power system into a relationship type in a knowledge graph. Through interlocking relationships, the knowledge graph can express the physical possibility of switching the power supply path from mains power to emergency power under fault conditions.

[0087] For example, in one feasible implementation, please refer to Figure 4This provides a specific process for constructing a knowledge graph, which is used to automatically generate nodes and relationships in the knowledge graph based on the load table of the power system. The knowledge graph construction is completed offline before the power system operation and maintenance methods are executed. The constructed knowledge graph is used for subsequent fault propagation analysis and may specifically include S100-S700.

[0088] S100, Data Reading. Read the load table of the power system. The format of the load table is not limited here; for example, it can be stored in Excel format. Load table types include, but are not limited to: municipal incoming line information table, process load table, first building load table, second building load table, and third building load table. In the load table, each row of data corresponds to a power facility or a connection relationship between facilities. During the data reading process, the table file can be read row by row, header fields can be identified, and cell data can be extracted row by row.

[0089] S200, Parsing and Standardization. The system performs deduplication on the read data, identifying and removing duplicate rows based on unique identifiers. Field values ​​are standardized; for example, the "Level 1 Total Container Load" node in the first building load table and the "Level 1 Container" node in the second building load table are standardized to a single "Level 1 Load" node.

[0090] After parsing, the attribute information of each power facility and the connection relationships between them are obtained. The attribute information represents the electrical characteristics and connection location of the power facility, including rated capacity, power supply type, load type, and backup transformer serial number; the connection relationships represent the power supply link direction between upstream and downstream equipment, the backup association between equipment at the same level, and the hierarchical affiliation between equipment and its components.

[0091] S300, Node Generation. Based on attribute information and preset node templates, nodes corresponding to each power facility are generated. In this embodiment, the node template defines the type labels, attribute field sets, and data formats that the node should contain. Each node template corresponds to a type of power facility, and there are no specific limitations here. For example, in one feasible embodiment, the types of node templates include, but are not limited to: upstream busbar of municipal incoming line, municipal incoming line, medium-voltage incoming line cabinet, medium-voltage feeder cabinet, transformer, low-voltage incoming line cabinet, low-voltage feeder cabinet, primary load, plug-in interface, secondary load, tertiary load, quaternary load, machine, and machine component.

[0092] Taking a transformer node as an example, the node generation process can be as follows: the node type label for the transformer node is "Transformer"; the unique identifier is generated based on the combination of "location" + "power type" + "number", such as "F1aa001", where F1 represents the location, aa represents the power type, and 001 is the number; the attribute fields can be data such as rated capacity, power type, load type, and backup transformer serial number extracted from the load table. For each power facility, the system selects a node template that matches its type, fills the parsed attribute information into the attribute fields of the node template, and generates a node instance with a unique identifier, type label, and attribute values.

[0093] S400, Relationship Generation. Based on connection relationships and preset power supply rules, relationships between nodes are established. Power supply rules include normal power supply link conduction rules, mutual backup rules between dual backup devices, and switching interlocking rules between mains power and emergency power. Relationships between nodes include six categories: downstream relationships, upstream relationships, inclusion relationships, belonging relationships, mutual backup relationships, and interlocking relationships.

[0094] Based on the connection relationships, downstream relationships, upstream relationships, inclusion relationships, and belonging relationships are established. Specifically, the power supply side node identifier and the power receiving side node identifier in the connection relationship are identified, and a downstream relationship from the power supply side to the power receiving side and an upstream relationship from the power receiving side to the power supply side are established between them; the overall equipment identifier and the component identifier in the connection relationship are identified, and an inclusion relationship from the whole to the component and a belonging relationship from the component to the whole are established between them.

[0095] If the connection relationship indicates that the mutual backup attribute identifiers of two power facilities are the same, then a mutual backup relationship is established between the nodes corresponding to the two power facilities according to the mutual backup rules. For example, when the "mutual backup transformer serial number" attribute of two transformer nodes is the same, a two-way mutual backup relationship is established between them.

[0096] If the connection relationship indicates that the mains power incoming cabinet and the emergency power incoming cabinet exist in pairs, then an interlocking relationship is established between the node corresponding to the mains power incoming cabinet and the node corresponding to the emergency power incoming cabinet according to the switching interlocking rules. For example, when the nodes of "Mains Power Incoming Cabinet" and "E-Power Incoming Cabinet" in the same row exist in the municipal incoming information table, a two-way interlocking relationship is established.

[0097] S500, Integrity Verification. Multiple verifications are performed on each node and its relationships. Exemplarily, in one feasible implementation, these verifications include, but are not limited to: consistency verification, power supply link integrity verification, critical node isolation detection, ring power supply detection, hierarchical relationship verification, field null value handling verification, transformer mutual backup chain integrity detection, load capacity verification, and voltage level matching verification. Specifically, uniqueness verification requires that the unique identifiers of all nodes be globally unique; naming consistency verification requires that the naming rules for upstream and downstream node numbers be consistent; electrical logic consistency verification requires that the power supply type of the primary load matches the upstream transformer; power supply link integrity verification requires that all load nodes have a complete power supply link; critical node isolation detection requires that buses and transformers must not exist in isolation; ring power supply detection requires that power supply loops are not allowed, except for mutual backup relationships; hierarchical relationship verification requires that the power supply hierarchy conforms to the defined relationship order; field null value handling verification requires that null value nodes in upstream and downstream relationships be correctly skipped; transformer mutual backup chain integrity detection requires that mutual backup transformers appear in pairs; load capacity verification requires that the load current does not exceed the capacity of the upstream equipment; and voltage level matching verification requires that the voltage levels of upstream and downstream devices match.

[0098] S600: If the verification is successful, the verified nodes and relationships will be written into the graph database. The graph database adopts a graph data organization structure, which consists of nodes and relationships connecting nodes. It stores the verified node instances and relationship connections to form a complete knowledge graph.

[0099] If the S700 fails verification, a verification report and failure record are generated based on the failed data. These are then manually reviewed and re-imported, but not written to the knowledge graph. The verification report records the identifier of the failed node or relationship, the reason for the failure, and its location in the load table. After the operations and maintenance personnel correct the load table data based on the verification report, they re-execute the data reading to integrity verification process until all data passes verification and is written to the graph database.

[0100] In one feasible embodiment, step S30, which involves traversing the knowledge graph starting from the first node and determining the power supply status of the second node based on the electrical on / off status of each node, includes: Step S301: Starting from the first node, perform a topological traversal along the relationships in the knowledge graph; Topological traversal is a data processing procedure that starts from a starting node and visits adjacent nodes along the relationships between nodes in a knowledge graph. The purpose of traversal is to visit all nodes on the fault propagation path and obtain the electrical on / off status and power supply status of each node. The traversal starts from the first node, which is the fault source node, and the traversal unfolds layer by layer along the relationships extending outward from the first node.

[0101] The traversal algorithm is not limited to a specific implementation. In one feasible implementation, the traversal uses a breadth-first search algorithm, employing a queue to manage the nodes to be visited and processing each node sequentially according to its level. In another feasible implementation, the traversal uses a depth-first search algorithm, employing a stack structure to manage the nodes to be visited, delving along a single path to the end and then backtracking. In yet another feasible implementation, the traversal uses a recursive algorithm, implementing path deepening and backtracking through a function call stack.

[0102] During the traversal, a record of visited nodes can be maintained to avoid the same node being processed repeatedly. Starting from the current node, the downstream relationships, mutual backup relationships, and locking relationships of the current node in the knowledge graph are queried, and the direction of the next traversal is determined based on the direction of each relationship.

[0103] Step S302: For the current node being traversed, if the electrical on / off state of the downstream node of the current node is on, then the downstream node of the current node is determined as the second node, the power supply state of the downstream node of the current node is marked as off, and the downstream node of the current node is determined as the traversed node and the traversal continues downward. The current node is the node being processed during the traversal. Downstream nodes are nodes located below the current node in the power supply link, connected through downstream relationships. The electrical on / off state is "on," indicating that the power facilities corresponding to the downstream node are energized and current can flow normally.

[0104] Query the operating status field of the power facility corresponding to the downstream node from the status database or real-time status mirror, and read whether the value of the field is in the conducting state. If the reading result is in the conducting state, it is determined that the fault can continue to propagate downstream along the main power supply path. Add the identifier of the downstream node to the second node set, and assign the power supply status field of the downstream node to the power failure state. The power failure state indicates that the node cannot obtain normal power supply through the main power supply path and is within the scope of the fault.

[0105] The downstream node is added to the set of nodes to be processed, allowing the traversal algorithm to continue querying the downstream relationships, mutual backup relationships, and locking relationships of this downstream node in subsequent steps. This process ensures that the layer-by-layer extension of the fault propagation path is completely traced.

[0106] Step S303: If the electrical on / off state of the current node is off or tripped, query the knowledge graph for backup nodes that have a backup relationship with the current node. If the electrical on / off state of the backup node is on, determine the downstream node of the backup node as the second node, mark the power supply state of the downstream node of the backup node as the transfer power supply state, and continue to traverse downwards after determining the downstream node of the backup node as the traversal node.

[0107] Specifically, the electrical on / off status of the current node being either disconnected or tripped indicates that the power facilities corresponding to the current node have been taken out of operation, and the main power supply path is interrupted at this node. The disconnected status indicates that the equipment is in an electrically isolated state due to tripping or power loss, while the tripped status indicates that the equipment has automatically disconnected the circuit due to protection action.

[0108] Retrieve nodes with a backup relationship to the current node from the knowledge graph. Backup relationships are pre-established during the knowledge graph construction phase based on the matching results of backup attribute identifiers. Perform a query using graph query language, starting from the current node and proceeding along the adjacent nodes of the backup relationship type, returning the identifiers and attributes of the backup nodes. Query the operating status field of the power facilities corresponding to the backup nodes from the status database; if the electrical on / off status of the backup node is "on," the backup power supply path is determined to be available, and the fault propagation path is transferred from the current node to the backup node.

[0109] Add the downstream nodes connected to the backup node through downstream relationships to the second node set; assign the power supply status field of the downstream node to the value of "transfer power supply status". "Transfer power supply status" indicates that the node obtains backup power through the backup path, and although the main power supply path is interrupted, service is not interrupted. Add the downstream nodes of the backup node to the node set to be processed, allowing the traversal algorithm to continue tracking the fault propagation downstream of the backup path.

[0110] Step S304: If the electrical on / off state of the current node is off or tripped, the node type is medium voltage incoming cabinet and the power supply type is mains power, then query the knowledge graph for emergency power supply nodes that have a blocking relationship with the current node. If the electrical on / off state of the emergency power supply node is on, then determine the downstream node of the emergency power supply node as the second node, mark the power supply state of the downstream node of the emergency power supply node as emergency power supply state, and continue to traverse downwards after determining the downstream node of the emergency power supply node as the traversal node.

[0111] When the current node meets all the conditions, an emergency power supply path query is triggered. The conditions include: the electrical on / off status is off or tripped, the node type is a medium-voltage incoming cabinet, and the power supply type is mains power. These three conditions together indicate that the mains power supply has been taken out of operation on the high-voltage side, which is a prerequisite for triggering the emergency power supply.

[0112] Retrieve nodes with a locking relationship to the current node from the knowledge graph. Locking relationships are pre-established during the knowledge graph construction phase based on the paired relationship between mains power supply cabinets and emergency power supply cabinets. Perform a query using a graph query language, starting from the current node and proceeding along adjacent nodes of the locking relationship type, returning the identifier and attributes of the emergency power supply node. Query the operating status field of the power facility corresponding to the emergency power supply node from the status database; if the electrical on / off status of the emergency power supply node is "conducting," the emergency power supply path is determined to be available, and the fault propagation path switches from the mains power node to the emergency power supply node.

[0113] Add the downstream nodes connected to the emergency power node through downstream relationships to the second node set; assign the power supply status field of the downstream node to the emergency power supply status. The emergency power supply status indicates that the node obtains backup power through the emergency power supply path. Add the downstream nodes of the emergency power node to the node set to be processed, so that the traversal algorithm can continue to track the fault propagation downstream of the emergency power supply path.

[0114] Step S305: After the traversal is completed, output each second node and its power supply status.

[0115] There are no unprocessed nodes in the set of nodes to be processed, indicating that all paths reachable from the first node have been visited and processed, and a complete set of second nodes and the power supply status flags of each second node are obtained.

[0116] The second set of nodes and the power supply status of each node are organized into structured data. This structured data includes a list of node identifiers, the power supply status field value for each node, and the severity value for each node. Output formats include data files, interface responses, or displayed data. In one feasible implementation, the output process can generate an impact list data object. This impact list contains an array of second nodes, with each array element recording the node identifier, node type, power supply status, and severity, for subsequent fault diagnosis or query processing.

[0117] For example, in one feasible implementation, a topological traversal is performed in the knowledge graph starting from the first node. Based on the electrical on / off status of each node, the second node on the fault propagation path and its power supply status are automatically determined. Please refer to [reference needed]. Figure 5 The specific process can be: S1, Receive alarm events. When the monitoring layer receives an alarm event pushed by the power monitoring master station system, it parses the data packet of the alarm event, extracts the fault source device identifier as the facility identifier, which is the unique identification code of the power facility that triggered the alarm event in the system. Based on the facility identifier, it queries the graph database of the knowledge layer for node objects with matching attributes, locates the first node, and the first node is the fault source node.

[0118] S2, Initialize the traversal queue. Create a traversal queue, add the first node to the queue, and mark it as visited.

[0119] S3, check if the traversal queue is empty. If the traversal queue is empty, it means that all paths reachable from the first node have been visited and processed, and jump to S14; if the traversal queue is not empty, pop a node from the head of the traversal queue as the current node, and perform downstream relationship judgment, mutual backup relationship judgment, and locking relationship judgment on the current node.

[0120] S4, Determine the downstream relationships of the current node. Query all downstream relationships of the current node from the knowledge graph. If a downstream relationship exists, for each downstream node of the current node, obtain the electrical on / off status of that downstream node from the real-time state mirror of the twin layer. If no downstream relationship exists, proceed to S7 to perform the mutual backup relationship determination.

[0121] S5, Determine the status of the downstream node of the current node. If the electrical connection status of the downstream node is on, the system determines the downstream node as the second node and proceeds to S6. If the electrical connection status of the downstream node is off or tripped, the downstream path is interrupted, the power system operation and maintenance system will no longer traverse along the downstream path, and will jump to S7 to determine the mutual backup relationship.

[0122] S6, Mark the node affected by the power outage. The power system operation and maintenance system identifies this downstream node as the second node, marks its power supply status as out of power, adds it to the traversal queue and marks it as visited, so that the power system operation and maintenance system can continue to query the downstream relationships, backup relationships and interlocking relationships of this downstream node in subsequent loops, and then returns to S3.

[0123] S7, query the electrical on / off status of the current node. If the current node's electrical on / off status is off or tripped, query the knowledge layer for backup nodes with a mutual backup relationship to the current node, obtain the electrical on / off status of the backup nodes, and then execute S8. If the current node's electrical on / off status is on, return to S3.

[0124] S8, Cross-connection path conduction judgment. The power system operation and maintenance system judges whether the electrical on / off state of the cross-connection node is on. If the electrical on / off state of the cross-connection node is on, then execute S9; if the current node does not have a cross-connection relationship, or the electrical on / off state of the cross-connection node is off or tripped, then jump to S10 to perform the interlocking relationship judgment.

[0125] S9, Mark the load to be transferred. Determine that the backup power supply path is available, identify the downstream node of the backup node as the second node, mark the power supply status of the downstream node of the backup node as the transfer power supply status, add the downstream node of the backup node to the traversal queue and mark it as visited, so that the power system operation and maintenance system can continue to track the fault propagation situation downstream of the backup path, and then return to S3.

[0126] S10, Interlocking Condition Judgment. The power system operation and maintenance system determines whether the current node simultaneously meets the following three conditions: electrical on / off state is off or tripped, node type is medium-voltage incoming cabinet, and power supply type is mains power. If all three conditions are met, S11 is executed; if the current node does not meet the above three conditions, the power system operation and maintenance system directly returns to S3.

[0127] S11, Query emergency power supply nodes. The power system operation and maintenance system queries the knowledge layer for emergency power supply nodes that have an interlocking relationship with the current node, obtains the electrical on / off status of the emergency power supply node, and then executes S12.

[0128] S12, Emergency Path On / Off Judgment. The power system operation and maintenance system determines whether the electrical on / off status of the emergency power supply node is on. If the electrical on / off status of the emergency power supply node is on, then execute S13; if the electrical on / off status of the emergency power supply node is off or tripped, then the power system operation and maintenance system directly returns to S3.

[0129] S13, Mark the emergency power supply load. The power system operation and maintenance system determines that the emergency power supply path is available, identifies the downstream node of the emergency power supply node as the second node, marks the power supply status of the downstream node of the emergency power supply node as emergency power supply status, adds the downstream node of the emergency power supply node to the traversal queue and marks it as visited, so that the power system operation and maintenance system can continue to track the fault propagation situation downstream of the emergency power supply path, and then returns to S3.

[0130] S14, Output Impact List. After the traversal queue is empty and the main loop ends, the power system operation and maintenance system outputs all the second nodes identified during the traversal process and the power supply status corresponding to each second node. The output format includes a structured data object, where each second node records the node identifier, node type, power supply status, and severity. This structured data object is used for subsequent operation and maintenance steps.

[0131] During the traversal process described above, maintaining a record of visited nodes prevents the same node from being processed repeatedly. In a specific implementation, the traversal algorithm is not limited; a breadth-first search algorithm can be used, employing a first-in-first-out queue to manage the nodes to be processed, and processing each node sequentially according to its level. In another feasible implementation, a depth-first search algorithm can also be used, employing a stack structure to manage the nodes to be processed, traversing a single path to the end and then backtracking to the branch point to continue traversing. The specific implementation can be configured according to actual needs.

[0132] Based on the first and / or second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to the first and / or second embodiments described above can be referred to the above description and will not be repeated hereafter. Based on this, step S401, the step of determining the probability value based on the power supply path where the second node is located, includes: Step S4011: If the second node is connected to the first node through a single power supply path, calculate the product of the conduction probabilities of each node on the power supply path as the probability value. The second node is connected to the first node via a single power supply path, meaning there is only one path between the first and second nodes, formed by sequentially connecting nodes and relationships. There are no alternative paths provided by backup nodes or emergency power supply nodes along this path. The second node relies solely on this single path to obtain power. In this single power supply path, the nodes are connected in series in terms of power supply reliability. Power can only be transferred from the first node to the second node when all nodes on the path are conducting. If any node on the path is disconnected or tripped, the path is interrupted, and the second node cannot obtain power through this path.

[0133] The calculation of the product of the conduction probabilities of each node along the power supply path, as the probability value, is based on the reliability principle of series systems. The conduction probability is the probability that a node will maintain electrical continuity under fault conditions. In a single power supply path, a fault propagating from the first node to the second node requires passing through every intermediate node along the path. Only when all nodes on the path remain conductive can the fault propagate completely to the second node. Therefore, the conduction probability of the entire path is equal to the product of the conduction probabilities of each node along the path. If any node on the path is in an open or tripped state, the conduction probability of that node is zero, resulting in a zero propagation probability for the entire path.

[0134] In a specific implementation, the number of power supply paths between the second node and the first node can be determined. If only a single power supply path exists, the sequence of all nodes on that path is extracted. The conduction probability of each node on the path is queried or calculated. The conduction probabilities of all nodes are multiplied together, and the resulting product is used as the probability value of the fault propagating to the second node. This probability value directly reflects the likelihood of the fault propagating to the second node along the only path.

[0135] Step S4012: If the second node is connected to the first node through multiple power supply paths, calculate a probability value by subtracting the probability that each power supply path is not conductive.

[0136] The second node is connected to the first node through multiple power supply paths, meaning that there are two or more independent paths between the first and second nodes. These paths may include a main power supply path and a backup path, or a main power supply path and an emergency power supply path. Each path is electrically independent; the disconnection of one path does not affect the continuity of the others. As long as at least one path remains continuous, the second node can maintain power supply.

[0137] The probability value is calculated by subtracting the probability that all power supply paths are not conductive from the probability of 1, based on the reliability principle of parallel systems. For multiple power supply paths, the probability that a fault propagates to the second node is equal to the complement of the probabilities that all power supply paths are simultaneously not conductive. The probability that each power supply path is not conductive is equal to 1 minus the probability that the path is conductive. The probability that all power supply paths are not conductive is equal to the product of the probabilities that all paths are not conductive. Subtracting this product from 1 yields the probability that at least one power supply path remains conductive. This probability value is the probability that the fault propagates to the second node.

[0138] First, determine the number of power supply paths between the second node and the first node. When multiple power supply paths are determined, calculate the conduction probability of each power supply path. Convert the conduction probability of each path into its non-conductivity probability. Multiply the non-conductivity probabilities of each path together, and then subtract the product from one to obtain the probability value of the fault propagating to the second node. This probability value reflects the possibility that the fault will affect the second node after simultaneously cutting off all redundant paths.

[0139] Step S402: The product of the load utilization rate of the second node and the service level weight is used as the loss value of the second node. The load utilization rate represents the current load level of the power facility corresponding to the second node, the service level weight represents the importance of the service carried by the power facility corresponding to the second node, and the loss value represents the degree of business impact caused by the power outage of the power facility corresponding to the second node under fault conditions. Load utilization rate is the ratio of the real-time operating load of the power facility corresponding to the second node to its rated capacity. The calculation process for load utilization rate can be as follows: obtain the real-time power value of the power facility from the real-time status data, obtain the rated capacity value of the power facility from the attribute information, and divide the real-time power value by the rated capacity value to obtain the load utilization rate. Load utilization rate characterizes the degree of load saturation of the power facility at the current moment; the higher the load utilization rate, the heavier the production task the power facility is undertaking when a fault occurs.

[0140] Business level weights are pre-configured level coefficients based on the criticality of the business carried by the power facility corresponding to the second node. The configuration process for business level weights can be as follows: based on the functional positioning of the power facility in the production process, the capacity loss caused by downtime, and the scope of its impact on upstream and downstream processes, a level identifier and corresponding weight value are assigned to the power facility. A higher business level weight indicates a greater impact of the business carried by the power facility on the overall production system.

[0141] The loss value is a parameter that quantifies the impact on services caused by power outages due to fault conditions on the power facilities corresponding to the second node. The loss value is calculated by multiplying the load utilization rate by the service level weight. The calculation process involves reading the load utilization rate of the second node and the service level weight of the power facilities corresponding to that node, and multiplying the two values ​​to obtain the loss value. The loss value characterizes the overall impact of power loss on services at the second node, assuming the fault has propagated to it. The load utilization rate and service level weight jointly determine the magnitude of the loss value; nodes with high load and high service levels have higher loss values.

[0142] Step S403: The product of the probability value and the loss value is used as the severity of the second node.

[0143] Severity is a quantitative indicator that combines the probability of fault propagation with the degree of business loss. The severity is calculated by multiplying the probability value by the loss value. The product of the probability and loss values ​​represents the comprehensive risk of the fault's impact; the probability value reflects the likelihood of the fault reaching the node, and the loss value reflects the degree of business impact after the node loses power. Multiplying the two considers both the electrical probability of fault propagation and the business consequences of a node losing power. For nodes where the primary power supply path is definitively interrupted and cannot be restored through backup or emergency paths, the probability value approaches one, and the severity is primarily determined by the loss value. For nodes with backup or emergency paths that are operational, the probability value approaches zero, and the severity decreases accordingly.

[0144] Furthermore, in one feasible implementation, severity is used to sort and classify the second nodes. When outputting the impact list, each second node is sorted according to its severity value, so that maintenance personnel can identify the nodes that need to be dealt with first under fault conditions. Nodes with high severity correspond to power facilities with high load, high business level and high probability of fault propagation, and have a higher handling priority in fault response.

[0145] For example, in one feasible implementation, the process of determining the severity of the second node based on the power supply path in which the second node is located can be S4001-S4003: S4001, determine the probability value based on the power supply path of the second node. The probability value represents the likelihood of a fault propagating to the second node. When determining the probability value, first determine the number of power supply paths between the second node and the first node.

[0146] Specifically, if the second node is connected to the first node via a single power supply path, it means that there is only one path between the first and second nodes, formed by sequentially connecting nodes and relationships. There are no alternative paths provided by backup nodes or emergency power supply nodes on this path. According to the reliability principle of a series system, a fault propagating from the first node to the second node requires passing through every intermediate node on the path. Only when all nodes on the path remain conductive can the fault propagate completely to the second node. Therefore, the conduction probability of the entire path is equal to the product of the conduction probabilities of each node on the path. If any node on the path is disconnected or tripped, the conduction probability of that node is zero, resulting in a zero propagation probability for the entire path. For example, if the second node is connected to the first node via a single power supply path, the probability value can be expressed as:

[0147] Among them, P k This represents the conduction probability of a single power supply path k connecting the second node and the first node, i.e., the probability that a fault propagates to the second node along a single path. This represents the probability that node e on power supply path k remains conductive under fault conditions; path k This represents the power supply path k, which is the path formed by connecting nodes and relationships sequentially from the first node to the second node. This represents the probability that all nodes on the power supply link k remain conductive under fault conditions.

[0148] If the second node is connected to the first node through multiple power supply paths, it indicates that there are two or more independent paths between the first and second nodes. According to the reliability principle of parallel systems, the probability of a fault propagating to the second node is equal to the complement of the probabilities that all power supply paths are simultaneously non-conductive. The non-conductive probability of each power supply path is equal to 1 minus the conduction probability of that path. The probability that all power supply paths are non-conductive is equal to the product of the non-conductive probabilities of each path. Subtracting this product from 1 yields the probability that at least one power supply path remains conductive; this probability value is the probability of a fault propagating to the second node. For example, if the second node is connected to the first node through multiple power supply paths, the probability value can be expressed as:

[0149] Where i is the second node in the current severity calculation; Let F represent the probability of a fault propagating to the second node i under multiple power supply paths, i.e., the probability that at least one path is active; F is the fault event; path k Indicates the power supply path k; This represents the product operator over all power supply paths; This represents the probability that all nodes on the k-th power supply path remain conductive; This represents the probability that all paths are not workable; This represents the probability that node e on the power supply link remains conductive under fault conditions.

[0150] S4002 uses the product of the load utilization rate of the second node and the business level weight as the loss value of the second node.

[0151] The loss value represents the degree of business impact caused by power outages due to fault conditions in the power facilities corresponding to the second node. The load utilization rate represents the current load level of the power facilities corresponding to the second node. The business level weight represents the importance of the business carried by the power facilities corresponding to the second node.

[0152] Load utilization is the ratio of the real-time operating load of the power facility corresponding to the second node to its rated capacity. Service level weight is a pre-configured level coefficient based on the criticality of the services carried by the power facility corresponding to the second node. The loss value is calculated by multiplying the load utilization rate by the service level weight; this product represents the overall impact of power outage at the second node on the services, assuming the fault has propagated to that node.

[0153] For example, the loss function can be expressed as:

[0154] Among them, L i This represents the loss value at node i. The load utilization rate of node i, i.e., the real-time load rate, is the ratio of the real-time operating load of the power facility corresponding to that node to its rated capacity. This indicates the load importance of node i, i.e., the service level weight, which represents the importance of the services carried by the power facilities corresponding to that node.

[0155] S4003 uses the product of the probability value and the loss value as the severity of the second node.

[0156] Severity is a quantitative indicator that combines the probability of fault propagation with the degree of business loss. This product considers both the electrical probability of the fault reaching the second node and the degree of business impact after the node loses power. For example, the severity calculation model can be expressed as: (1) Define the severity of node i as its expected loss under fault conditions:

[0157] Where i is the second node currently calculating severity; R(i) is the severity of node i and its expected loss under failure conditions; E[·] represents the conditional expectation operator, and Loss i This represents the degree of service loss caused by power outage at node i under fault conditions, i.e., the loss value L. i .

[0158] (2) In the event of a fault The severity is then expanded to be the product of the failure propagation probability (i.e., the probability value) and the loss value:

[0159] Where i is the node index, which is the second node in the current severity calculation; R i Indicates the severity of node i. The probability value of the fault propagating to node i represents the likelihood that the fault will be transmitted from the first node to node i along the power supply path. This represents the loss value when node i is affected; F represents the failure event.

[0160] (3) Combining the above steps, the severity calculation model can be represented as follows:

[0161] Where R(i) represents the severity of node i; P(F→i) represents the probability of the fault propagating to node i; U i C represents the load utilization rate of node i; i This represents the business level weight of node i.

[0162] Based on the first, second, and / or third embodiments of this application, the content that is the same as or similar to that in Embodiment 1 described above in the fourth embodiment of this application can be referred to the above description and will not be repeated hereafter. Based on this, please refer to... Figure 6 Step S50, based on the second node, power supply status, and severity, involves the following steps for operating and maintaining the power system: Step S501: Obtain fault context data, wherein the fault context data includes at least the electrical on / off status of the first node, the power supply path of the first node, and historical case records similar to the alarm event; Fault context data is a multi-source heterogeneous data collection used for fault diagnosis reasoning. It is aggregated from knowledge graphs, real-time status databases, and historical databases.

[0163] The electrical on / off status of the first node refers to the on / off state of the power facility corresponding to the fault source node at the time of the alarm event. This status is read from the real-time status database or real-time status mirror and includes normal on-state, off-state, and tripped states. The electrical on / off status is used to determine whether the fault source facility has been taken out of operation and whether the fault has caused a physical power outage.

[0164] The power supply path of the first node is a pathway formed by sequentially connecting nodes and relationships between the power source node and the first node. This power supply path is obtained from the knowledge graph through graph queries, including the sequence of nodes and relationships along the path. The power supply path is used to show the location of the fault source node in the power grid topology, as well as the upstream and downstream range that the fault may affect.

[0165] Historical case records similar to alarm events are past fault records retrieved from historical or vector databases that are similar to the current alarm event in terms of event type, device type, or fault phenomenon. Historical case records include the historical alarm time, historical root cause, historical handling measures, and historical recovery results. Historical case records are used to provide a reference for current fault diagnosis.

[0166] Step S502: Convert the fault context data into a vector representation, perform a similarity search in the vector database, and obtain the search results. The vector database is obtained by vectorizing expert knowledge documents. In one feasible implementation, the vector database is a power fault knowledge index pre-built based on expert knowledge documents after vectorization. The construction of the vector database precedes the fault diagnosis phase to ensure that the knowledge base has completed data preparation before fault diagnosis is triggered.

[0167] The process of building a vector database can involve receiving expert knowledge documents uploaded by users. The file formats of these expert knowledge documents include, but are not limited to, PDF, Word, Excel, TXT, and Markdown. Document types include standard operating procedures, operating instructions, individual courses, job training materials, and historical accident records. In a specific implementation, the above-mentioned expert knowledge documents uploaded by users can be received through a management interface. Based on the file format, the corresponding parser can be called to extract the plain text content. For example, for PDF format, a PDF parser is used to extract the text layer content; for Word format, a document parser is used to extract paragraph text; for Excel format, a table parser is used to extract cell text; and for TXT and Markdown formats, the text content is read directly.

[0168] The extracted plain text content is input into a large language model, which is then guided by prompt word templates to extract structured knowledge items. The prompt word templates define the output format requirements, including but not limited to fields such as knowledge type, fault scenario description, symptom list, root cause list, diagnostic criteria, treatment actions, and related equipment types. The large language model identifies and extracts structured knowledge items from the document text based on the prompt word templates, outputting knowledge item data containing fault scenarios, symptoms, root causes, diagnostic criteria, and treatment actions.

[0169] The extracted knowledge entries are converted into fixed-dimensional vector representations using an embedding model. For example, this can be achieved by concatenating the scene description, symptoms, and root cause fields from the knowledge entries into a complete text, which is then converted into a fixed-dimensional vector representation using the embedding model. The embedding model maps natural language text to a high-dimensional vector space, ensuring that semantically similar texts have a small distance in the vector space. The generated vectors, along with the corresponding knowledge entry metadata, are stored in a vector database to create a vector index. The metadata includes the knowledge type, source document name, related device type, and extraction timestamp, which are used for result filtering and source tracing during subsequent retrieval.

[0170] In the online diagnostic phase, the fault context data is converted into a vector representation. The text content from the fault context data is concatenated into the retrieval text, and the same embedding model as in the offline phase is used to convert the retrieval text into a vector. A similarity search is then performed in the vector database. Similarity search is achieved by calculating the similarity between the retrieval vector and the vectors of each knowledge entry stored in the vector database, returning the top few knowledge entries with the highest similarity as the search results. The search results include the text content and similarity score of the expert knowledge entries.

[0171] Step S503: Input the fault context data into the rule engine, match it with the preset diagnostic rules, and calculate the rule matching degree. The diagnostic rules are obtained by transforming the threshold conditions and topological features in the expert knowledge document. Diagnostic rules are pre-defined judgment rules that can be executed by the rule engine, derived from threshold conditions and topological features in expert knowledge documents. The construction process for diagnostic rules can be as follows: During the offline knowledge construction phase, threshold conditions and topological features are extracted from the diagnostic criteria fields of the extracted knowledge entries. These threshold conditions and topological features are then converted into a rule format supported by the rule engine and stored in a relational knowledge base. Threshold conditions include upper current limit, lower voltage limit, power threshold, and duration. Topological features include node type combinations, relational path patterns, and hierarchical structure.

[0172] Rule matching is the process of comparing real-time parameters and topology information in fault context data with the conditions of diagnostic rules. The rule engine reads each preset diagnostic rule, comparing the real-time status values ​​in the fault context data with the threshold conditions in the rules, and comparing the power supply paths in the fault context data with the topology features in the rules. When the fault context data meets all or some of the conditions in the diagnostic rules, the rule engine calculates the matching degree of that rule. The matching degree is defined as the ratio of the number of conditions met by the fault context data to the total number of conditions in the rule. The rule with the highest matching degree is selected as the rule matching degree output from all diagnostic rules.

[0173] Step S504: Input the fault context data, retrieval results and rule matching degree into the large language model for fusion reasoning to generate the first diagnostic result; Fusion reasoning is a data processing procedure that integrates multi-source, heterogeneous diagnostic evidence into a unified diagnostic conclusion. It involves concatenating fault context data, vector database retrieval results, and rule matching scores from the rule engine into a structured input text. This structured input text includes real-time data segments, retrieval knowledge segments, and rule matching segments.

[0174] The system inputs the concatenated structured text into a large language model. The large language model performs semantic understanding and logical reasoning on the input text, and, by integrating real-time operating status, historical similar cases, and rule matching results, generates a preliminary judgment on the root cause of the fault. The first diagnostic result is text data containing a description of the root cause of the fault, evidence citations, and handling suggestions. The first diagnostic result is generated by the large language model based on the fusion of multi-source information in the input text, and does not rely on a single source of evidence, thus improving the reliability of fault diagnosis.

[0175] Step S505: Calculate the confidence level of the first diagnostic result to obtain the posterior probability, and use the posterior probability and the first diagnostic result as the diagnostic result corresponding to the alarm event.

[0176] It should be noted that confidence level represents the degree of credibility of the first diagnostic result, while posterior probability is the quantitative expression of this degree of credibility in probabilistic form, with a value range of zero to one. The closer the posterior probability is to one, the higher the credibility of the first diagnostic result; the closer the posterior probability is to zero, the lower the credibility of the first diagnostic result.

[0177] Calculating the confidence level to obtain the posterior probability refers to the complete calculation process of fusing multi-source evidence through a Bayesian probability model. The specific calculation process is not limited here and can be set according to actual needs. For example, the scores of each evidence source can be mapped to conditional probabilities, and then the logarithms can be taken under the conditional independence assumption and weighted summed to obtain the fused logarithmic probability, which can then be normalized to obtain the posterior probability. Alternatively, the scores of each evidence source can be directly weighted and averaged or weighted multiplied, and the result can be mapped to a probability value in the interval of zero to one as the posterior probability.

[0178] The posterior probability and the initial diagnostic result are combined to form the diagnostic result corresponding to the alarm event. The diagnostic result includes the root cause of the fault, the posterior probability, the chain of evidence, and the handling recommendations. The diagnostic result is output in structured data format for operation and maintenance personnel to view and use as a reference for decision-making.

[0179] In one feasible embodiment, step S505, the step of calculating the confidence level of the first diagnostic result to obtain the posterior probability, includes: Step S5051: Map the rule matching degree to the first conditional probability, map the similarity of the search results to the second conditional probability, and map the completeness of the fault context data to the third conditional probability. The first conditional probability is the probability of evidence appearing based on the rule matching degree, assuming the root cause of the fault is true. The second conditional probability is the probability of evidence appearing based on the similarity of the search results, assuming the root cause of the fault is true. The third conditional probability is the probability of evidence appearing based on the completeness of the fault context data, assuming the root cause of the fault is true. These three conditional probabilities correspond to three different sources of diagnostic evidence, used to unify heterogeneous engineering scores within a probabilistic framework.

[0180] Mapping rule matching degree to first conditional probability is the process of converting the matching degree value output by the rule engine into a probability value. The mapping process can be as follows: read the rule matching degree value and directly use that value as the first conditional probability value. The rule matching degree is the ratio of the number of matching conditions output by the rule engine to the total number of rule conditions, and its value ranges from zero to one. The mapped first conditional probability is in the range of zero to one.

[0181] Mapping the similarity of search results to a second conditional probability is the process of converting the similarity score returned by the vector database search into a probability value. The mapping process can be as follows: read the similarity score from the search results and directly use that score as the value of the second conditional probability. Search similarity represents the semantic closeness between the fault context data and the expert knowledge entry, and its value ranges from zero to one. The mapped second conditional probability is also within the zero-to-one range.

[0182] Mapping the completeness of fault context data to a third conditional probability is the process of converting an engineering metric reflecting the degree of data loss into a probability value. The mapping process can be as follows: statistically analyze the fill status of each field in the fault context data, calculate the ratio of the number of filled fields to the total number of fields, and use this ratio as the value of the third conditional probability. The higher the data completeness, the closer the third conditional probability is to one; the more data is missing, the closer the third conditional probability is to zero.

[0183] Step S5052: Under the conditional independence assumption, take the logarithm of the first conditional probability, the second conditional probability, and the third conditional probability, and then sum them up by weight to obtain the fused logarithmic probability. Normalize the logarithmic probability to obtain the posterior probability.

[0184] The conditional independence assumption is a simplifying assumption in Bayesian probability models. It assumes that, given the known root cause of the failure, the three sources of evidence corresponding to the first, second, and third conditional probabilities are independent of each other. That is, the probability of one piece of evidence appearing is not affected by the presence or absence of other evidence. This assumption allows the joint conditional probability to be decomposed into the product of the independent conditional probabilities, thus simplifying the fusion calculation of multi-source heterogeneous evidence into a combination of the conditional probabilities of each piece of evidence.

[0185] Under the conditional independence assumption, the logarithms of the first, second, and third conditional probabilities are taken and then summed with weights to obtain the fused logarithmic probability. This logarithmic probability value represents the strength of the combined evidence support for each candidate root cause.

[0186] The fused logarithmic probabilities are normalized to obtain the posterior probabilities. The normalization process can be as follows: Perform an exponential operation on the logarithmic probability values ​​of all candidate root causes, converting the logarithmic probabilities back to their original probabilities in the probability domain. Divide the original probability value of each candidate root cause by the sum of the original probability values ​​of all candidate root causes, ensuring that each probability value falls within the interval of zero to one and the sum of the probability values ​​of all candidate root causes is one. The normalized probability value is the posterior probability, representing the probability that each fault root cause is valid after fusing multi-source evidence.

[0187] For example, in one feasible implementation, the process of calculating the confidence level of the first diagnostic result to obtain the posterior probability may include S1000-S3000: S1000, posterior probability modeling, can specifically include S1001-S1004: S1001, Problem Modeling, that is, calculating the probability that the root cause H of the fault is true given the evidence E in the first diagnostic result; the specific modeling can be: P(HlE), where P(·) represents the probability function; H represents the candidate root cause (that is, the root cause of the fault determined in the first diagnostic result); E represents the evidence set (that is, the rule matching degree, the similarity of the search results, and the completeness of the fault context data in the first diagnostic result); P(HlE) is the probability that the root cause H of the fault is true given the observation of evidence E.

[0188] S1002, using Bayes' theorem, transforms the problem of finding the probability P(H|E) of the root cause under known evidence into finding the probability P(E|H) of the evidence appearing when the root cause is assumed to be true. The specific manifestations are as follows:

[0189] Wherein, P(HlE) is the probability that the candidate root cause H is valid under the condition of observing multi-source evidence E (rule matching degree, retrieval similarity, data completeness); P(ElH) is the probability of observing the current evidence set E under the assumption that the root cause H is valid, which in this embodiment is obtained by the product of the first conditional probability, the second conditional probability, and the third conditional probability. P(H) is the initial probability of the candidate root cause H occurring before any evidence is available, which can be specifically set by historical statistical frequency or expert experience. P(E) is the total probability of the evidence set E occurring under all possible conditions, serving as a normalization constant.

[0190] S1003, multi-source evidence modeling, that is, decomposing the evidence in the first diagnostic result into rule matching degree, similarity of search results, and completeness of fault context data. It is assumed that these three pieces of evidence do not affect each other under the condition that the root cause of the fault H is known. Thus, the probability P(ElH) of the root cause of the fault is decomposed into the product of three conditional probabilities, avoiding computational complexity caused by the coupling of multi-source evidence. Where E={E rule E retrieval E data}, where E is the total set of evidence, E rule E represents the rule matching degree. retrieval E represents the similarity of search results. data This indicates the completeness of the fault context data.

[0191] In engineering, if we assume conditional independence and that the root cause of the failure, H, is true, then the probability that the root cause of the failure, H, is true is: P(ElH)∝P(E rule lH)·P(E retrieval lH)·P(Edata lH)·P(H) Where P(ElH) is the joint probability of the total evidence set E occurring under the assumption that the root cause of the failure H is true; P( lH) represents the probability of evidence appearing corresponding to the rule matching degree under the assumption that H is true, i.e., the first conditional probability; P(E) retrieval lH) represents the probability of evidence appearing corresponding to the similarity of search results, i.e., the second conditional probability, given that hypothesis H is true; P(E) data lH) represents the probability of evidence appearing corresponding to the completeness of the fault context data under the assumption that H is true, which is also the third conditional probability; P(H) is the prior probability, which is the initial probability of the root cause H of the fault occurring before any evidence is available.

[0192] S1004 combines the Bayesian formula from S1002 with the assumptions from step S1003 to obtain a directly computable mathematical model. This model shows that, given multiple sources of evidence E, the probability that a certain root cause H is true is proportional to the prior probability P(H) of that root cause multiplied by the conditional probability of each of the three independent pieces of evidence supporting that root cause, that is: P(ElH)∝P(E rule lH)·P(E retrieval lH)·P(E data lH)·P(H) S2000 maps rule matching degree to a first conditional probability, similarity of search results to a second conditional probability, and completeness of fault context data to a third conditional probability. Specifically,

[0193]

[0194]

[0195] in, This is the first conditional probability; For rule matching degree; This is the second conditional probability; Similarity of semantic search results; This is the third conditional probability; The completeness of the fault context data.

[0196] In this embodiment, the first conditional probability is the probability of evidence appearing corresponding to the rule matching degree under the assumption that the root cause of the fault is true. Mapping the rule matching degree to the first conditional probability is the process of converting the rule matching degree value output by the rule engine into a probability value. The rule matching degree is the ratio of the number of matching conditions output by the rule engine to the total number of rule conditions, and its value ranges from zero to one. The mapped first conditional probability is in the range of zero to one.

[0197] The second conditional probability is the probability of evidence appearing based on the similarity of the search results, assuming the root cause of the fault is true. Mapping the similarity of search results to the second conditional probability is the process of converting the similarity score returned by the vector database search into a probability value. The similarity of the search results represents the semantic closeness between the fault context data and the expert knowledge entries, and its value ranges from zero to one. The mapped second conditional probability is in the interval between zero and one.

[0198] The third conditional probability is the probability of evidence appearing corresponding to the degree of completeness of the fault context data, assuming the root cause of the fault is valid. Mapping the completeness of fault context data to the third conditional probability is the process of converting an engineering indicator reflecting the degree of data loss into a probability value. The mapping process can be as follows: statistically analyze the fill status of each field in the fault context data, calculate the ratio of the number of filled fields to the total number of fields, and use this ratio as the value of the third conditional probability. The higher the data completeness, the closer the third conditional probability is to one; the more data is missing, the closer the third conditional probability is to zero.

[0199] S3000, under the assumption of conditional independence, involves taking the logarithms of the first, second, and third conditional probabilities, then summing them with weights to obtain a fused logarithmic probability. This logarithmic probability is then normalized to obtain the posterior probability. In other words, the three independent conditional probabilities are fused into a single score representing the strength of comprehensive evidence supporting the root cause of the failure, and then converted into a normalized probability for easier comparison and decision-making. Specifically, this can include S3001-S3002: S3001, the first conditional probability, the second conditional probability, and the third conditional probability are expressed logarithmically, and then weighted and summed to obtain the fused logarithmic probability. The specific formula can be:

[0200] in, The weight parameter represents the data, which can be optimized using historical data; i represents the i-th piece of evidence E; S i That is, the score of the i-th piece of evidence E, which includes in this embodiment. Rule matching degree Similarity of semantic search results The completeness of the fault context data; const represents a constant; Score(H) represents the log probability after fusion; P(H|E) represents the probability that the root cause H of the fault is true.

[0201] S3002, normalize the logarithmic probability to obtain the posterior probability:

[0202] Where Confidence(H) is the posterior probability of the root cause H in the first diagnostic result; e is the natural constant, the base of the exponential function, used to convert the fusion score from the logarithmic domain back to the probability domain; Score(H) is the calculated fusion logarithmic probability value of the root cause H; j is the index of the root cause; in this embodiment, all root causes to be compared can be traversed; H j The j-th root cause of the failure is the j-th among all failure cause hypotheses participating in the normalization comparison; e Score(Hj) The index score of the j-th root cause of failure represents the unnormalized weight of the candidate root cause in the probability domain. j e Score(Hj) This represents the summation of the exponential scores over all candidate root causes.

[0203] In one feasible embodiment, a natural language query function is provided based on the constructed knowledge graph. The specific natural language query process may be as follows: Obtain the query request. Obtain the query request input by the user, and determine the query level based on the complexity and structure of the query request. The query level can be set according to actual needs. For example, in one feasible implementation, the query level includes, but is not limited to: fixed template query, parameterized template query, scenario command query, and natural language query.

[0204] Generate query statements. If the query level is a fixed template query, a preset fixed query template is invoked, and the pre-compiled graph database query statement is executed directly. If the query level is a parameterized template query, a preset parameterized query template is invoked, the parameter values ​​input by the user are filled into the parameterized query template, and the graph database query is executed. If the query level is a scenario instruction query, a preset scenario knowledge base is invoked, and the business scenario is identified through keyword matching or semantic similarity calculation, the business scenario is mapped to the corresponding template query, and executed. If the query level is a natural language query, a large language model is invoked to perform intent recognition and entity extraction on the query request, and a graph database query statement is generated.

[0205] Security constraint verification. After generating the graph database query statement, security constraint verification is performed on it. The content of the security constraint verification is not limited here and can be set according to actual needs. For example, in one feasible implementation, security constraint verification may include, but is not limited to: entity whitelist verification, relation type verification, verification of whether attribute names belong to the standard attribute set of each node type, sensitive operation interception, and query depth verification. Specifically, entity whitelist verification checks whether the node types involved in the graph database query statement belong to the node type whitelist. If the graph database query statement contains entities not on the whitelist, execution is rejected and an error message is returned. Relation type verification checks whether the relation types involved in the graph database query statement belong to the relation type whitelist. If the graph database query statement contains types not on the whitelist, execution is rejected and an error message is returned. Attribute name verification checks whether the attribute name belongs to the standard attribute set of each node type. If the attribute name does not belong to the standard attribute set of each node type, execution is rejected and an error message is returned. Sensitive operation interception means that if the graph database query statement contains keywords related to sensitive operations, such as keywords related to data modification, the query is intercepted and an unsupported message is returned. Query depth validation refers to parsing the path query parameters contained in the graph database query statement, extracting the maximum number of hops, and if the maximum number of hops exceeds a preset threshold, the execution is rejected and a warning message is returned, or the depth parameter in the query statement is automatically truncated before execution.

[0206] Execute the query. After the query statement passes security checks, execute the graph database query to obtain the query results.

[0207] Query result security control. The system counts the number of nodes and paths returned by the query. If the number of nodes or paths exceeds a preset limit, the returned results are truncated, retaining only those within the limit and displaying a truncation flag to inform the user that the results have been truncated. If the limit is not exceeded, the complete results are retained. The system aggregates the above data to generate response data containing structured query results, highlighted paths, and explanatory information, which is then displayed on the user interface.

[0208] Data access constraints. During the query process, all data acquisition from the power monitoring master station system is read-only, and write-back operations are prohibited. The access frequency to the same data point is limited, and duplicate calls are reduced through a local caching mechanism. When the interface response of the power monitoring master station system times out or the number of consecutive failures exceeds a preset threshold, subsequent requests are stopped and cached data is returned. The output query results are all advisory and do not contain any control instructions. Structured audit logs are generated for key operations such as query requests, diagnostic task triggering and completion, and system configuration changes.

[0209] This application also provides a power system operation and maintenance system; please refer to... Figure 7 The aforementioned power system operation and maintenance system includes: The monitoring layer is used to respond to alarm events and parse the alarm events to obtain facility identifiers, where the facility identifier is the identifier of the power facility in the power system that triggered the alarm event; The knowledge layer is used to store the preset knowledge graph and obtain the first node corresponding to the facility identifier from the knowledge graph. The knowledge graph includes the nodes corresponding to each power facility in the power system and the relationships between the nodes. The relationships between the nodes are mapped according to the power supply rules of the power system. The twin layer, connected to the monitoring layer, is used to maintain the electrical on / off status of each node; The intelligent layer, connected to the knowledge layer and the twin layer, is used to traverse the knowledge graph starting from the first node, determine the power supply status of the second node and the second node based on the electrical on / off status of each node, where the second node is a node on the fault propagation path of the first node; and determine the severity of the second node based on the power supply path in which the second node is located, where the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power failure conditions when the fault propagates to the second node; wherein, the intelligent layer is used to determine a probability value based on the power supply path in which the second node is located, the probability value characterizing the likelihood of the fault propagating to the second node; and use the product of the load utilization rate of the second node and the business level weight as the loss value of the second node, where the load utilization rate characterizes the current load level of the power facility corresponding to the second node, the business level weight characterizes the importance of the business carried by the power facility corresponding to the second node, and the loss value characterizes the degree of business impact caused by the power interruption of the power facility corresponding to the second node under fault conditions; and use the product of the probability value and the loss value as the severity of the second node; The application layer, connected to the intelligence layer, is used to perform operation and maintenance on the power system based on the second node, power supply status, and severity.

[0210] The power system operation and maintenance system provided in this application can implement the power system operation and maintenance method in the above embodiments, improve the efficiency of power system operation and maintenance, and reduce the risk of untimely fault handling caused by human error. Compared with the prior art, the beneficial effects of the power system operation and maintenance system provided in this application are the same as those of the power system operation and maintenance method provided in the above embodiments, and other technical features of the power system operation and maintenance system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0211] Specifically, the power system operation and maintenance system provided in this embodiment adopts a five-layer decoupled architecture consisting of a monitoring layer, a twin layer, a knowledge layer, an intelligence layer, and an application layer. Data is transmitted unidirectionally or bidirectionally between the layers through predefined data interfaces. The monitoring layer connects upwards to the twin layer, while the intelligence layer connects downwards to the knowledge layer and the twin layer, and upwards to the application layer, forming a complete data link from data acquisition to operation and maintenance decision-making.

[0212] The monitoring layer is deployed in the monitoring and management center and includes a power monitoring data adapter and frequency control and fuse protection modules. The power monitoring data adapter accesses the power monitoring master station system's interface via read-only requests, periodically retrieving power monitoring data, such as real-time telemetry, telesignal data, alarm events, and event sequence records. When the power monitoring master station system detects that electrical parameters exceed preset thresholds or receives a trip signal from a protection device based on the power monitoring data, it generates an alarm event and pushes it to the monitoring layer. Upon receiving the alarm event, the monitoring layer parses the fault source device identifier field in the data packet to extract the facility identifier, which is a unique identification code for the power facility that triggered the alarm event within the power system operation and maintenance system. The monitoring layer then passes the parsed facility identifier, along with the alarm type and timestamp information from the alarm event, down to the twin layer. In this embodiment, the frequency control and circuit breaker protection module uniformly constrains the access frequency, retry count, and degradation strategy of the power monitoring data adapter. When the interface response of the power monitoring master station system times out or the number of consecutive failures exceeds a preset threshold, the circuit breaker is triggered and cached data is returned to ensure that the power monitoring master station system is read-only, not controlled, and does not exceed its authority.

[0213] The twin layer connects to the monitoring layer and includes a state caching service and a real-time state mirror. The state caching service receives raw telemetry and teleindication data from the monitoring layer, standardizes the data format, and writes it to the real-time state mirror. The real-time state mirror maintains the dynamic attributes of each power facility in key-value pairs, where the electrical on / off status includes normal conduction, disconnection, and tripping states. This electrical on / off status is periodically collected and updated by the monitoring layer. When the intelligent layer initiates a status query request, the twin layer returns the current electrical on / off status of each node through the status query interface, providing real-time data support for fault propagation path analysis.

[0214] The knowledge layer comprises a graph database and a vector database. The graph database stores a pre-defined knowledge graph, where nodes correspond one-to-one with power facilities in the power system. Each node stores a facility identifier, node type, and electrical attributes. Relationships between nodes in the knowledge graph are mapped according to the power system's power supply rules, including at least downstream and upstream relationships representing the direction of power transmission, mutual backup relationships representing redundant logic, interlocking relationships representing power switching constraints, and inclusion and membership relationships representing the composition of equipment hierarchies. The vector database stores vector representations of expert knowledge such as standard operating procedures, operating instructions, single-point courses, job training materials, and historical accident records, used for similar knowledge retrieval. The real-time states of the knowledge layer and the twin layer are strictly decoupled to ensure that the topology does not experience data inconsistencies due to real-time state fluctuations. The knowledge layer receives facility identifiers from the intelligent layer, performs attribute matching queries in the graph database, compares the facility identifiers with those stored in the node attributes, locates the first node, and returns the node identifier, type label, and adjacency relationships of this first node to the intelligent layer.

[0215] The intelligent layer establishes data connections with the knowledge layer and twin layer, and includes a unified application service orchestration and permission gateway, an impact scope analysis engine, a multi-source heterogeneous knowledge fusion diagnostic engine, a redundancy analysis agent, a dispatch evaluation agent, and a natural language query engine. The unified application service orchestration and permission gateway is responsible for receiving application layer requests, performing permission verification, orchestrating graph queries and status queries, and aggregating the results. The impact scope analysis engine is responsible for traversing the knowledge graph starting from the first node, determining the second node and its power supply status based on the electrical on / off status of each node, and calculating the severity based on the power supply path of the second node. The multi-source heterogeneous knowledge fusion diagnostic engine further includes a knowledge base construction module, a vector retrieval module, a rule engine module, a large language model inference module, and a confidence calculation module, responsible for fusing vector retrieval, rule matching, and large language model inference to generate diagnostic results with confidence scores. The redundancy analysis agent identifies single-point failure risks based on the knowledge graph. The dispatch evaluation agent assesses the remaining capacity of candidate access points when a new device is added. The natural language query engine converts natural language requests into controlled graph database query statements.

[0216] The application layer connects with the intelligence layer and includes a knowledge graph view, intelligent query, alarm center, knowledge base management, and knowledge graph management. The application layer receives the set of second-nodes, the power supply status of each second-node, and its severity from the intelligence layer. It organizes this data into a structured impact list and displays the fault propagation path diagram, the list of affected nodes, and the severity ranking results through the front-end interface. Based on the information displayed by the application layer, maintenance personnel perform subsequent maintenance operations on the power system, including but not limited to fault isolation confirmation, backup power supply switching verification, load transfer scheduling, and on-site handling dispatch. The knowledge graph view provides a visual representation of the power system topology, supporting layer control, layout switching, and device detail viewing. Intelligent query supports hierarchical natural language queries, including fixed template queries, parameterized queries, scenario command queries, and natural language queries. The alarm center displays a list of alarm events, providing impact surface analysis and fault propagation path visualization. The knowledge base manages user-uploaded expert knowledge documents, mapping configurations, and rule entries. The knowledge graph management provides functions for building, importing, validating, and versioning the knowledge graph. Each module in the application layer is only responsible for data display and user interaction, and does not perform any reasoning or control logic.

[0217] In this embodiment, the power system operation and maintenance system strictly adheres to the call constraints of the power monitoring system, while following the principle of unidirectional dependency data access between layers. All calls from the monitoring layer to the power monitoring master station system are read-only operations, with no write-back operations prohibited at the code level. Access is limited to the data service interfaces provided by the power monitoring system, including real-time data query interfaces, event query interfaces, and configuration query interfaces. The access frequency to the same data point is limited to a preset threshold, and a local caching mechanism reduces duplicate calls. When the interface response of the power monitoring master station system times out or the number of consecutive failures exceeds a preset number, the circuit breaker opens, stopping subsequent requests and returning cached data.

[0218] Application layer data is obtained through interfaces provided by the intelligence layer. The intelligence layer outputs all diagnostic conclusions, standard operating procedure suggestions, dispatch plans, and other advisory content. The power system operation and maintenance system generates structured audit logs for key operations such as user login / logout, query requests, diagnostic task triggering and completion, and system configuration changes.

[0219] In the offline knowledge construction phase, the power system operation and maintenance system receives expert knowledge documents uploaded by users through the management interface. Supported file formats include, but are not limited to, PDF, Word, Excel, TXT, and Markdown. The power system operation and maintenance system calls the corresponding parser to extract plain text content based on the file format. The extracted text is input into a large language model, which is guided by prompts to extract structured knowledge items. These knowledge items include fault scenarios, symptoms, root causes, diagnostic criteria, and handling actions. The power system operation and maintenance system concatenates the scenario description, symptom, and root cause fields from the extracted knowledge items into text, calls the embedding model to convert it into a fixed-dimensional vector representation, and stores it in a vector database. The power system operation and maintenance system converts the threshold conditions and topological features from the extracted diagnostic criteria into executable diagnostic rules that can be executed by the rule engine and stores them in a relational knowledge base.

[0220] During the online diagnostic phase, when the monitoring layer receives a new alarm event, the power system operation and maintenance system automatically triggers the diagnostic process. The intelligent layer obtains the real-time status of the fault source node and its upstream and downstream nodes from the twin layer, retrieves the power supply topology path of the fault source node from the knowledge layer, and searches the vector database for the most recent historical records similar to the current event type. The intelligent layer calls the impact range analysis engine to perform topology traversal along the downstream relationships, backup relationships, and interlocking relationships in the knowledge graph, starting from the first node corresponding to the fault source node. During the traversal, the impact range analysis engine queries the twin layer for the electrical on / off status of the currently traversed nodes. If the downstream node of the current node is in a conducting state, then the downstream node is identified as the second node, and its power supply status is marked as power failure. If the current node is in a disconnected or tripped state, then the knowledge graph is queried for backup nodes with a mutual backup relationship with this node. If the backup node is conducting, the downstream node of the backup node is identified as the second node and marked as a transfer power supply state. If the current node is a medium-voltage incoming cabinet of mains power type and is in a disconnected or tripped state, then the emergency power supply node with a lockout relationship is queried. If the emergency power supply node is conducting, its downstream node is identified as the second node and marked as an emergency power supply state. After traversal, the impact range analysis engine summarizes all second nodes and their corresponding power supply statuses, and calculates the severity based on the power supply path in which each second node is located. For a single power supply path, the product of the conduction probabilities of each node on the path is calculated as the fault propagation probability value; for multiple power supply paths, the probability value is calculated by subtracting the probability that all paths are not conducting. Simultaneously, the impact scope analysis engine obtains the load utilization rate and business level weight of the second node, multiplies them by the loss value, where the load utilization rate is read from the real-time state mirror of the twin layer, and the business level weight is pre-set in the node attributes of the knowledge layer. The impact scope analysis engine multiplies the probability value by the loss value to obtain the severity of the second node. This severity quantifies the risk of business loss caused when the fault propagates to this node.

[0221] While the impact scope analysis engine performs the aforementioned traversal and severity calculation, the intelligent layer invokes the multi-source heterogeneous knowledge fusion diagnostic engine to perform root cause diagnosis. The diagnostic engine converts the aggregated fault context text into a vector representation, performs a similarity search in the vector database, returns the most relevant expert knowledge entries, and records the highest similarity score. The diagnostic engine inputs the context data into the rule engine, matches it against predefined diagnostic rules, and calculates the rule matching degree. The diagnostic engine concatenates the context text, search results, and rule matching results into structured input text, inputs it into a large language model for fusion inference, and generates a first diagnostic result containing a fault root cause description, evidence citations, and treatment recommendations. The confidence calculation module uses a Bayesian probability model to fuse multi-source diagnostic evidence, mapping the rule matching degree, search similarity, and data completeness to conditional probabilities. Under the assumption of conditional independence, the logarithms are taken, weighted, and summed, then normalized to obtain the posterior probability. The posterior probability and the first diagnostic result are combined to form the diagnostic result corresponding to the alarm event. The intelligent layer outputs the above impact scope analysis results and diagnostic results to the application layer through the unified application service orchestration and permission gateway. The application layer then displays the fault propagation path, the list of affected loads, the root cause diagnosis conclusions, and the handling suggestions.

[0222] Through the static composition of the above five-layer architecture, the data flow of inter-layer data interaction, offline knowledge construction and online diagnosis, as well as the engineering implementation of system call constraints, the dynamic processes of alarm response, fault propagation traversal, severity calculation and operation and maintenance decision-making are supported by the system architecture, ensuring that static topology, real-time status and intelligent reasoning are decoupled at the physical deployment and logical interaction levels.

[0223] This application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the power system operation and maintenance method in Embodiment 1 above.

[0224] The following is for reference. Figure 8 The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of this application. The electronic devices in these embodiments may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0225] like Figure 8 As shown, the electronic device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the electronic device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a display, speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the electronic device to communicate wirelessly or wiredly with other devices to exchange data. Although the diagrams show electronic devices with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented alternatively.

[0226] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0227] The electronic device provided in this application, employing the power system operation and maintenance method described in the above embodiments, can improve the efficiency of power system operation and maintenance and reduce the risk of untimely fault handling caused by human error. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the power system operation and maintenance method provided in the above embodiments, and other technical features of the electronic device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0228] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0229] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0230] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the power system operation and maintenance method in the above embodiments.

[0231] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0232] The aforementioned computer-readable storage medium may be included in an electronic device or may exist independently without being assembled into an electronic device.

[0233] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by an electronic device, cause the electronic device to implement the power system operation and maintenance methods of the various embodiments described above.

[0234] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0235] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0236] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0237] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., computer programs) for executing the above-described power system operation and maintenance method. This improves power system operation and maintenance efficiency and reduces the risk of untimely fault handling caused by human error. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the power system operation and maintenance method provided in the above embodiments, and will not be repeated here.

[0238] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the power system operation and maintenance method described above.

[0239] The computer program product provided in this application can improve the efficiency of power system operation and maintenance and reduce the risk of untimely fault handling caused by human error. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the power system operation and maintenance method provided in the above embodiments, and will not be repeated here.

[0240] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A power system operation and maintenance method, characterized in that, The power system operation and maintenance methods include: In response to an alarm event, the alarm event is parsed to obtain a facility identifier, wherein the facility identifier is the identifier of the power facility in the power system that triggered the alarm event; The first node corresponding to the facility identifier is obtained from the preset knowledge graph, wherein the knowledge graph includes nodes corresponding to each power facility in the power system and the relationship between each node, and the relationship between each node is mapped according to the power supply rules of the power system; Starting from the first node, the knowledge graph is traversed, and the power supply status of the second node and the second node are determined according to the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node. Based on the power supply path in which the second node is located, the severity of the second node is determined, wherein the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power outage conditions when a fault propagates to the second node; wherein the step of determining the severity of the second node based on the power supply path in which the second node is located includes: determining a probability value based on the power supply path in which the second node is located, the probability value characterizing the likelihood of a fault propagating to the second node; multiplying the load utilization rate of the second node by the service level weight as the loss value of the second node, the load utilization rate characterizing the current load level of the power facility corresponding to the second node, the service level weight characterizing the importance of the service carried by the power facility corresponding to the second node, and the loss value characterizing the degree of business impact caused by power outage of the power facility corresponding to the second node under fault conditions; and multiplying the probability value by the loss value as the severity of the second node; The power system is operated and maintained based on the second node, the power supply status, and the severity. Prior to the step of obtaining the first node corresponding to the facility identifier from the preset knowledge graph, the method further includes: Obtain the load table of the power system, parse the load table to obtain the attribute information of each power facility and the connection relationship between each power facility, wherein the attribute information characterizes the electrical characteristics and connection location of the power facility; Based on the attribute information and the preset node template, nodes corresponding to each of the power facilities are generated. The types of the node template include at least the upstream busbar of the municipal incoming line, municipal incoming line, medium voltage incoming line cabinet, medium voltage feeder cabinet, transformer, low voltage incoming line cabinet, low voltage feeder cabinet, loads at all levels, plug-in interface, machine and machine component. Based on the connection relationship and the preset power supply rules, the relationship between each node is established. The power supply rules include normal power supply link conduction rules, mutual backup rules between dual backup devices, and switching interlocking rules between mains power and emergency power. The relationship between each node includes at least downstream relationship, upstream relationship, inclusion relationship, belonging relationship, mutual backup relationship, and interlocking relationship. Each node and its relationship are verified, and the nodes and relationships that pass the verification are written into the knowledge graph. The step of traversing the knowledge graph starting from the first node and determining the power supply status of the second node based on the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node, includes: Starting from the first node, perform a topological traversal along the relationships in the knowledge graph; For the current node being traversed, if the electrical on / off state of the downstream node of the current node is on, then the downstream node of the current node is determined as the second node, the power supply state of the downstream node of the current node is marked as off, and the downstream node of the current node is determined as the traversal node and the traversal continues downward. If the electrical connection status of the current node is disconnected or tripped, then query the knowledge graph for backup nodes that have the backup relationship with the current node. If the electrical connection status of the backup node is on, then determine the downstream node of the backup node as the second node, mark the power supply status of the downstream node of the backup node as a transfer power supply status, and continue to traverse downwards after determining the downstream node of the backup node as a traversal node. If the electrical on / off state of the current node is disconnected or tripped, the node type is a medium-voltage incoming cabinet, and the power supply type is mains power, then query the knowledge graph for emergency power supply nodes that have the interlocking relationship with the current node. If the electrical on / off state of the emergency power supply node is on, then determine the downstream node of the emergency power supply node as the second node, mark the power supply state of the downstream node of the emergency power supply node as emergency power supply state, and continue to traverse downwards after determining the downstream node of the emergency power supply node as the traversal node. After the traversal is complete, output the power supply status of each second node.

2. The power system operation and maintenance method as described in claim 1, characterized in that, The step of establishing the relationship between the nodes based on the connection relationship and the preset power supply rules includes: Based on the aforementioned connection relationships, downstream relationships, upstream relationships, inclusion relationships, and membership relationships are established. If the connection relationship indicates that the mutual backup attribute identifiers of the two power facilities are the same, then a mutual backup relationship is established between the nodes corresponding to the two power facilities according to the mutual backup rule; If the connection relationship indicates that the mains power incoming cabinet and the emergency power incoming cabinet exist in pairs, then an interlocking relationship is established between the node corresponding to the mains power incoming cabinet and the node corresponding to the emergency power incoming cabinet according to the switching interlocking rule.

3. The power system operation and maintenance method as described in claim 1, characterized in that, The step of determining the probability value based on the power supply path where the second node is located includes: If the second node is connected to the first node through a single power supply path, then the product of the conduction probabilities of each node on the power supply path is calculated as the probability value. If the second node is connected to the first node through multiple power supply paths, then calculate a probability value by subtracting the probability that each of the power supply paths is not conductive.

4. The power system operation and maintenance method as described in any one of claims 1 to 3, characterized in that, The steps for operating and maintaining the power system based on the second node, the power supply status, and the severity include: Obtain fault context data, wherein the fault context data includes at least the electrical on / off status of the first node, the power supply path of the first node, and historical case records similar to the alarm event; The fault context data is converted into a vector representation, and a similarity search is performed in the vector database to obtain the search results. The vector database is obtained by vectorizing expert knowledge documents. The fault context data is input into the rule engine and matched with preset diagnostic rules to calculate the rule matching degree. The diagnostic rules are obtained by transforming the threshold conditions and topological features in the expert knowledge document. The fault context data, the search results, and the rule matching degree are input into a large language model for fusion reasoning to generate a first diagnostic result. The confidence level of the first diagnostic result is calculated to obtain the posterior probability, and the posterior probability and the first diagnostic result are used as the diagnostic result corresponding to the alarm event.

5. The power system operation and maintenance method as described in claim 4, characterized in that, The step of calculating the confidence level of the first diagnostic result to obtain the posterior probability includes: The rule matching degree is mapped to a first conditional probability, the similarity of the search results is mapped to a second conditional probability, and the completeness of the fault context data is mapped to a third conditional probability. Under the conditional independence assumption, the logarithms of the first conditional probability, the second conditional probability, and the third conditional probability are taken and weighted and summed to obtain the fused logarithmic probability. The logarithmic probability is then normalized to obtain the posterior probability.

6. A power system operation and maintenance system, characterized in that, The power system operation and maintenance system is used to implement the power system operation and maintenance method as described in claim 1, and the power system operation and maintenance system includes: The monitoring layer is used to respond to alarm events, parse the alarm events to obtain facility identifiers, wherein the facility identifiers are the identifiers of the power facilities in the power system that triggered the alarm events; The knowledge layer is used to store a preset knowledge graph and obtain the first node corresponding to the facility identifier from the knowledge graph. The knowledge graph includes nodes corresponding to each power facility in the power system and the relationships between the nodes. The relationships between the nodes are mapped according to the power supply rules of the power system. A twin layer, connected to the monitoring layer, is used to maintain the electrical on / off status of each node; An intelligent layer, connected to the knowledge layer and the twin layer, is used to traverse the knowledge graph starting from the first node, determine the power supply status of the second node based on the electrical on / off status of each node, wherein the second node is a node on the fault propagation path of the first node; and determine the severity of the second node based on the power supply path in which the second node is located, wherein the severity characterizes the degree of business loss caused by the power facility corresponding to the second node under power failure conditions when the fault propagates to the second node; wherein the intelligent layer is used to determine a probability value based on the power supply path in which the second node is located, wherein the probability value characterizes the possibility of the fault propagating to the second node; and take the product of the load utilization rate of the second node and the business level weight as the loss value of the second node, wherein the load utilization rate characterizes the current load level of the power facility corresponding to the second node, the business level weight characterizes the importance of the business carried by the power facility corresponding to the second node, and the loss value characterizes the degree of business impact caused by the power interruption of the power facility corresponding to the second node under fault conditions; and take the product of the probability value and the loss value as the severity of the second node; The application layer, connected to the intelligent layer, is used to perform operation and maintenance on the power system based on the second node, the power supply status, and the severity.

7. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the power system operation and maintenance method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the power system operation and maintenance method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the power system operation and maintenance method as described in any one of claims 1 to 5.

Images