Electric power fire-fighting emergency decision-making method and system based on knowledge graph

Abstract

The invention provides an electric power fire-fighting emergency decision-making method and system based on a knowledge graph, and relates to the technical field of artificial intelligence, and the method comprises the steps: constructing an electric power fire-fighting knowledge graph, integrating transformer station equipment ledger data, real-time monitoring data, external environment data and historical case data, and building an association relationship among multi-source data, thereby obtaining an electric power fire-fighting emergency decision-making result; a structured knowledge base supporting emergency decision making is obtained; performing dynamic risk reasoning based on a structured knowledge base and fire alarm data transmitted in real time, identifying a fire behavior influence range and a key risk point, and obtaining a preliminary differentiated emergency disposal scheme based on a reasoning result; and extracting spatial features of the key risk points and the monitoring positions, dynamically determining and updating boundary parameters of the fire behavior influence area through a spatial feature fitting algorithm, and defining the dynamically updated fire behavior influence area. According to the method, the whole-process intelligence from data integration to instruction execution is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a power fire emergency decision-making method and system based on knowledge graphs. Background Technology

[0002] With the continuous advancement of smart grid construction, the scale of substation equipment is constantly expanding. The combined effects of aging electrical equipment, extreme weather conditions, and complex operation and maintenance environments have led to a surge in fire risks. Power system fires cause economic losses, with substation cable fires and transformer explosions accounting for a particularly high proportion, posing a serious threat to the safe and stable operation of the power grid. Against this backdrop, traditional power fire emergency decision-making models are no longer adequate for the actual needs of complex scenarios, mainly due to three core bottlenecks: First, the emergency response plans have a low degree of structure. Most existing plans are presented in text form, and computers cannot automatically parse the key operational information and related logic. For example, the statement in the relevant procedures that the nearest power supply should be cut off first often leads to problems such as delayed operation judgment and poor step connection in actual execution due to the lack of support from the data related to the equipment.

[0003] Second, there is insufficient multi-source data fusion capability. Equipment ledger data and real-time monitoring data generated during the daily operation of substations, such as temperature, smoke concentration, external meteorological data and historical accident cases, are often scattered and stored in different systems, forming information silos. Some smart fire protection platforms have realized the basic data collection function, but have failed to establish deep semantic relationships such as equipment spatial correlation and the influence of meteorological factors on fire, which makes it difficult for the accuracy of fire early warning to meet actual emergency needs.

[0004] Third, dynamic decision-making and response are lagging behind. Traditional emergency decision-making relies heavily on human experience and judgment. In the face of a dynamic fire spread, it is impossible to update the boundaries of the fire-affected area and potential risk points in real time. From the perspective of actual operation, the time from alarm triggering to the generation of a preliminary response plan in a substation fire is relatively long, which is far from matching the golden response window required for fire emergency response. It is easy to miss the best rescue opportunity due to decision-making delays. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a power fire emergency decision-making method and system based on knowledge graphs, so as to realize the intelligentization of the entire process from data integration to command execution.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a power fire emergency decision-making method based on knowledge graphs, the method comprising: A knowledge graph for power fire protection is constructed by integrating substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data, and establishing the relationships between multiple data sources to obtain a structured knowledge base that supports emergency decision-making. Based on a structured knowledge base and real-time fire alarm data, dynamic risk reasoning is performed to identify the fire's impact range and key risk points. Based on the reasoning results, preliminary differentiated emergency response plans are obtained. Spatial features of key risk points and monitoring locations are extracted, and the boundary parameters of the fire-affected area are dynamically determined and updated through a spatial feature fitting algorithm, thus defining the dynamically updated fire-affected area. Based on the fire-affected area, the fire-affected area is divided into multiple analysis units according to rules. Real-time fire parameters are collected in each analysis unit, and a comprehensive adjustment parameter is calculated based on the parameter change characteristics. Based on the comprehensive adjustment parameter, the preliminary differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. The multi-role operation instruction set is adapted to the functional authority of the executing entity and the terminal perception requirements, and then processed for multi-modal information adaptation and visualization rendering. It is then distributed to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

[0007] Furthermore, a power fire protection knowledge graph is constructed by integrating substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data, and establishing relationships between multiple data sources to obtain a structured knowledge base that supports emergency decision-making, including: Integrate substation equipment ledger data, real-time monitoring data, external environment data, and historical case data to form a multi-source heterogeneous data source; Based on multi-source heterogeneous data sources, entities, attributes, and relationships in the power fire protection field are extracted, and an ontology schema is defined. Based on the ontology model, multi-source data are fused and mapped to construct a power fire protection knowledge graph containing entities, attributes, and relationships. The knowledge graph of power fire protection is stored in a graph database to form a structured knowledge base that supports emergency decision-making.

[0008] Furthermore, based on a structured knowledge base and real-time fire alarm data, dynamic risk reasoning is performed to identify the fire's impact range and key risk points. Based on the reasoning results, preliminary differentiated emergency response plans are derived, including: Receive real-time fire alarm data and associate and map the fire alarm data with entities in a structured knowledge base; Based on the reasoning rules of knowledge graphs, risk propagation reasoning is performed on alarm entities after association mapping, dynamically identifying equipment and facilities directly threatened and potentially affected by fire, and determining the scope of fire impact and key risk points. Based on the identified fire impact range and key risk points, and combined with historical case data and emergency response plan rules, a preliminary differentiated emergency response plan is generated.

[0009] Furthermore, spatial features of key risk points and monitoring locations are extracted, and the boundary parameters of the fire-affected area are dynamically determined and updated using a spatial feature fitting algorithm, thus defining the dynamically updated fire-affected area, including: By querying the structured knowledge base, the spatial coordinates and layout information of the identified key risk points and their surrounding monitoring locations are extracted; Based on spatial coordinates and layout information, a topological network is constructed that reflects the spatial connections and proximity relationships between key risk points, monitoring locations, and surrounding equipment and facilities. The real-time fire parameter data is combined with the topology network and processed by a spatial feature fitting algorithm. Based on the spatial distance between the monitoring point and the key risk point, the real-time fire parameter values ​​and their changing trends, weights are assigned to the monitoring data at different locations. An interpolation calculation method is used to dynamically calculate the fire intensity distribution of the entire area and calculate the boundary parameters of the fire-affected area. Based on the obtained boundary parameters, the geometric range of the fire-affected area is defined and updated in real time. The geometric range is represented in the form of a set of boundary coordinates or a region outline.

[0010] Furthermore, based on the fire-affected area, the fire-affected area is regularly divided into multiple analysis units. Real-time fire parameters are collected within each analysis unit, and a comprehensive adjustment parameter is calculated based on the parameter change characteristics, including: Based on the determined, dynamically updated fire impact area, a gridding method based on geographic coordinates or predefined dimensions is used to regularly divide the area into several independent spatial analysis units. Based on each spatial analysis unit, corresponding monitoring equipment is associated or deployed according to the geographical location, and real-time fire parameters within the coverage area of ​​that unit are collected. The time-series change characteristics of real-time fire parameters collected in each spatial analysis unit are analyzed. The rate of change, extreme values ​​and fluctuation amplitude of each parameter per unit time are calculated by sliding window, and characteristic indicators that characterize the dynamic evolution of the fire situation in each unit are extracted. Based on the characteristic indicators of each unit, weighting factors are assigned to them according to the degree of contribution of the characteristic indicators to the overall fire situation. A multi-indicator weighted fusion algorithm is used to normalize and aggregate the characteristic indicators of all units to obtain a comprehensive adjustment parameter for quantitative assessment and dynamic reflection of the current overall fire situation.

[0011] Furthermore, based on comprehensive adjustment parameters, the initial differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operational instruction set for different execution roles is generated, including: The comprehensive adjustment parameters are compared with the preset adjustment thresholds of the contingency plan, and the optimization direction and intensity of the preliminary differentiated emergency response plan are determined based on the comparison results. Based on the optimization direction and intensity, the emergency response rules and historical handling cases in the structured knowledge base are invoked to dynamically adjust the initial plan's handling measures, resource allocation plan, and execution priority, thereby obtaining the latest handling strategy adapted to the current fire situation; The latest handling strategy is broken down into specific, actionable tasks, and based on the responsibilities and permissions of the execution roles and the terminal type, operation instruction sets are generated and packaged for different execution roles.

[0012] Furthermore, the multi-role operation instruction set is adapted and visualized based on the functional authority of the executing entity and the terminal's perception requirements. This is then distributed to the corresponding role's terminal devices via a secure communication protocol, driving multi-terminal collaborative execution of emergency response operations, including: Receive operation instruction sets for different execution roles, and parse the instruction content and its corresponding target terminal attributes; Based on the obtained instruction content and target terminal attributes, combined with the functional authority of the executing entity and the terminal's perception requirements, the instruction content is adapted and converted into a multimodal information format; Visualize and render multimodal information to generate an interactive command interface that includes graphics, text, voice, and augmented reality elements; The multimodal instruction information generated by rendering is distributed to the terminal devices of the corresponding user roles through an encrypted secure communication protocol; Based on the instruction information, the system drives each terminal device to receive and display the corresponding instructions, and coordinates to trigger and execute the corresponding emergency response operations.

[0013] Secondly, a knowledge graph-based power fire emergency decision-making system includes: The acquisition module is used to build a power fire protection knowledge graph. By integrating substation equipment ledger data, real-time monitoring data, external environmental data and historical case data, and establishing the relationship between multi-source data, a structured knowledge base that supports emergency decision-making is obtained. The calculation module is used to perform dynamic risk reasoning based on a structured knowledge base and real-time fire alarm data, identify the fire impact range and key risk points, and obtain preliminary differentiated emergency response plans based on the reasoning results; extract the spatial features of key risk points and monitoring locations, and dynamically determine and update the boundary parameters of the fire impact area through a spatial feature fitting algorithm, thereby defining the dynamically updated fire impact area. The adjustment module is used to divide the fire-affected area into multiple analysis units based on the fire-affected area, collect real-time fire parameters in each analysis unit, and calculate a comprehensive adjustment parameter based on the parameter change characteristics. Based on the comprehensive adjustment parameter, the initial differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. The processing module is used to adapt and visualize the multi-role operation instruction set according to the functional authority of the executing subject and the terminal perception requirements, and distribute it to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

[0014] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0015] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0016] The above-described solution of the present invention has at least the following beneficial effects: By constructing a power fire protection knowledge graph to integrate substation equipment ledgers, real-time monitoring, external environment, and historical case data, and establishing multi-source data associations, a structured knowledge base supporting decision-making is formed. Based on knowledge graph reasoning rules and spatial feature fitting algorithms, combined with real-time alarm data, dynamic risk reasoning, fire impact range identification, and real-time updating of boundary parameters are achieved. The fire impact area is divided into grids, and comprehensive adjustment parameters are calculated through multi-index weighted fusion algorithms to dynamically optimize emergency plans. According to the functional authority of the executing entity and the terminal perception requirements, the operation instruction set is adapted to multi-modal information, visualized, and distributed securely via communication. This overcomes the technical problems in traditional power fire emergency decision-making, such as unstructured emergency plans leading to delayed operational judgment and poor step connection; information silos formed by scattered multi-source data resulting in insufficient accuracy of fire early warning; and reliance on human experience leading to delayed dynamic decision-making response and inability to match the golden window of fire emergency response. Thus, the entire process of power fire emergency response, from data integration to instruction execution, is intelligent, improving the response efficiency of emergency decision-making, ensuring the efficient collaborative execution of emergency response operations by multi-role terminal devices, and effectively reducing the impact range of fire accidents. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a knowledge graph-based power fire emergency decision-making method provided by an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of a knowledge graph-based power fire emergency decision-making system provided by an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0020] like Figure 1 As shown, an embodiment of the present invention proposes a power fire emergency decision-making method based on knowledge graphs, the method comprising the following steps: Step 1: Construct a power fire protection knowledge graph by integrating substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data, and establishing the relationship between multiple data sources to obtain a structured knowledge base that supports emergency decision-making. Step 2: Based on the structured knowledge base and real-time fire alarm data, perform dynamic risk reasoning to identify the fire impact range and key risk points. Based on the reasoning results, obtain a preliminary differentiated emergency response plan. Extract the spatial features of key risk points and monitoring locations, and dynamically determine and update the boundary parameters of the fire impact area through a spatial feature fitting algorithm to define the dynamically updated fire impact area. Step 3: Based on the fire-affected area, the fire-affected area is divided into multiple analysis units according to rules. Real-time fire parameters are collected in each analysis unit, and a comprehensive adjustment parameter is calculated based on the parameter change characteristics. Based on the comprehensive adjustment parameter, the preliminary differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. Step 4: The multi-role operation instruction set is adapted to multi-modal information and visualized by the functional authority of the executing entity and the terminal perception requirements. It is then distributed to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

[0021] In this embodiment of the invention, a power fire protection knowledge graph is constructed to integrate substation equipment ledgers, real-time monitoring, external environment, and historical case data, and to establish relationships between multi-source data, forming a structured knowledge base that supports emergency decision-making. Based on this structured knowledge base and real-time fire alarm data, dynamic risk reasoning is performed to identify the fire's impact range and key risk points, generating preliminary differentiated emergency response plans. Simultaneously, spatial features are extracted and the boundary parameters of the fire-affected area are dynamically updated using a spatial feature fitting algorithm. The fire-affected area is regularly divided into analysis units, real-time fire parameters are collected, and comprehensive adjustment parameters are calculated based on parameter change characteristics to optimize the preliminary plan. This technology generates multi-role operation instruction sets; adapts and visualizes these sets according to the functional authority of the executing entity and the terminal's perception requirements, and then distributes them to the corresponding terminals via a secure communication protocol. This overcomes the technical problems of traditional power fire emergency decision-making, such as the fragmentation of multi-source data forming information silos that cannot support accurate decision-making; the difficulty in dynamically identifying the scope of fire impact and risk points; the lack of differentiation and real-time nature of emergency plans; poor instruction adaptability; and low efficiency of multi-terminal collaborative execution. It achieves orderly connection and dynamic optimization of the entire process from data integration to instruction execution in power fire emergency decision-making, improving the adaptability of emergency plans and the collaborative execution of multiple roles.

[0022] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Integrate substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data to form a multi-source heterogeneous data source. Specifically, this includes: first, exporting equipment ledger data from the power company's substation equipment management system. This data contains the name, model, installation location, commissioning time, and rated parameter information of all equipment within the substation. Then, using monitoring equipment deployed within the substation, such as temperature sensors, smoke sensors, and flame detectors, real-time monitoring data such as equipment operating temperature, ambient smoke concentration, and the presence of flames are collected. Simultaneously, the system connects to the meteorological data platform of the meteorological department to obtain real-time wind speed, wind direction, precipitation, temperature, and other external environmental data for the substation area. In addition to environmental data, historical case data of past substation fire accidents were collected from the accident archive system of the fire department and the safety management archives of the power company. The data includes information such as the time of the accident, the equipment involved, the cause of the fire, the development process of the fire, the response measures, and the results of the response. Then, the acquired equipment ledger data, real-time monitoring data, external environmental data, and historical case data were initially sorted out. The unstructured historical case text data was converted into structured tabular data, and the real-time monitoring data of different formats were uniformly converted into JSON format. Finally, the data was integrated to form a multi-source heterogeneous data source covering basic equipment information, real-time status information, external environmental information, and historical accident information.

[0023] Step 1.2: Based on multi-source heterogeneous data sources, extract entities, attributes, and relationships in the power and fire protection field, and define the ontology schema. Specifically, this includes: based on the integrated multi-source heterogeneous data sources, using a combination of manual annotation and natural language processing technology, extracting equipment entities (transformers, circuit breakers, and cables) from equipment ledger data; extracting entities for temperature, smoke concentration, and flame monitoring indicators from real-time monitoring data; extracting entities for wind speed and wind direction from external environmental data; and extracting entities for equipment overheating, cable short-circuit fire risks, power outage operations, and emergency response to sprinkler system activation from historical case data. Then, for each extracted entity, determine its corresponding attributes, such as the attributes of the equipment entity. The attributes of the entities include installation location and rated temperature. The attributes of the monitoring indicator entities include monitoring threshold and current value. The attributes of the risk entities include risk level and triggering conditions. The attributes of the handling entities include operation steps and execution time. Then, the relationships between different entities are sorted out. For example, there is an occurrence relationship between the transformer and equipment entities and the equipment overheating risk entity. There is a need to execute the equipment overheating risk entity and the power outage operation handling entity. Finally, based on the extracted entities, attributes and relationships, an ontology modeling tool is used to construct an ontology pattern for the power fire protection field. This pattern clearly defines the definitions of various entities, the specific content of attributes, and the types and constraints of relationships between entities, forming a unified knowledge framework for the power fire protection field.

[0024] Step 1.3: Based on the ontology schema, the multi-source data is fused and mapped to construct a power fire protection knowledge graph containing entities, attributes, and relationships. Specifically, this includes: First, according to the defined ontology schema, mapping processing is performed on each type of data in the multi-source data. This involves mapping each piece of equipment information from the equipment ledger data to the equipment entity in the ontology schema, mapping the equipment model, installation location, and other information to the attributes of the equipment entity; mapping each monitoring indicator value from the real-time monitoring data to the monitoring indicator entity in the ontology schema, mapping the value to the attributes of the monitoring indicator entity; mapping each environmental factor data from the external environment data to the environmental factor entity in the ontology schema, mapping the specific value of the environmental factor to the attributes of the environmental factor entity; mapping each risk event from the historical case data to the risk entity in the ontology schema; and mapping each handling action to the ontology. The model identifies entities for handling issues and then addresses data conflicts during the mapping process. For example, when equipment installation location data from different sources are inconsistent, the equipment location data recorded in the power company's GIS system is used as the standard. When an abnormal fluctuation occurs in the value of a certain indicator in the real-time monitoring data, it is corrected by combining monitoring data from adjacent time periods. Then, based on the entity relationships defined in the ontology model, associations are established between data from different data sources. For example, the real-time temperature monitoring data of a transformer is associated with the transformer equipment entity in the ontology model, the transformer equipment entity is associated with the equipment overheating risk entity, and the equipment overheating risk entity is associated with the corresponding power outage operation handling entity in historical cases. Finally, through the above fusion and mapping process, a power fire protection knowledge graph is constructed, which includes various entities such as equipment, monitoring indicators, environmental factors, risks, and handling, as well as entity attributes and inter-entity relationships.

[0025] Step 1.4 involves storing the power fire protection knowledge graph in a graph database to form a structured knowledge base that supports emergency decision-making. Specifically, this includes: selecting Neo4j graph database as the storage medium based on the data scale and query requirements of the power fire protection knowledge graph. This database can efficiently store and manage complex relationships between entities. Then, following the data storage format of Neo4j graph database, each entity in the power fire protection knowledge graph is converted into a node in the database; each attribute of an entity is converted into a node's attribute key-value pair; and each relationship between entities is converted into an edge between nodes. Simultaneously, corresponding relationship type attributes are set for the edges. Finally, the converted node data and edge data are imported using Neo4j's import tool. The system imports data into the database in batches, creating data indexes during the import process. Indexes are created for commonly used query fields such as equipment name, risk type, and response measures to improve the efficiency of data querying during subsequent emergency decision-making. Then, a data update mechanism for a knowledge graph is established. When new equipment is added to the substation, causing changes in the equipment ledger data, the corresponding equipment nodes and relationships are added to the graph database in a timely manner. When real-time monitoring data is updated in real time, the attribute values ​​of the corresponding monitoring indicator nodes in the graph database are updated synchronously. When new historical fire cases occur, the corresponding risk nodes, response nodes, and related relationships are added to the graph database. Finally, a structured knowledge base that can be updated in real time and supports efficient querying and correlation analysis is formed.

[0026] In this embodiment of the invention, a multi-source heterogeneous data source is constructed by integrating substation equipment ledgers, real-time monitoring, external environment, and historical case data. Based on this data source, entities, attributes, and relationships in the field of power fire protection are extracted and ontology schemas are defined. The multi-source data is then fused and mapped according to the ontology schemas to construct a power fire protection knowledge graph containing entities, attributes, and relationships. This graph is then stored in a graph database to form a structured knowledge base that supports emergency decision-making. This approach effectively overcomes the technical problems in traditional power fire emergency decision-making, where multi-source data is scattered across different systems, lacks unified semantic associations, forms information silos, and computers cannot efficiently analyze data value or form a structured knowledge system to support decision-making. This approach achieves the orderly integration and deep semantic association of multi-source heterogeneous data, constructing a standardized and structured power fire protection knowledge base, and ensuring the efficiency of emergency decision-making from a data foundation level.

[0027] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Receive real-time fire alarm data and map it to entities in a structured knowledge base. Specifically, this involves receiving real-time fire alarm data from the fire alarm system deployed within the substation. The data includes the monitoring device number that triggered the alarm, the coordinates of the monitoring location, the specific parameter values ​​that triggered the alarm (e.g., temperature, smoke concentration), and the alarm occurrence time. Then, the power fire protection knowledge graph in the structured knowledge base is retrieved, and entity types related to the alarm data are selected from the knowledge graph, including equipment entities such as transformers and circuit breakers, and monitoring indicator entities such as temperature. For example, for smoke concentration and location entities such as a certain area of ​​a substation, the monitoring device number in the alarm data is matched with the device entity in the knowledge graph, the monitoring location coordinates are associated with the location entity in the knowledge graph, and the alarm parameter value is mapped with the attribute corresponding to the monitoring index entity in the knowledge graph. For example, the 85℃ alarm data returned by the temperature sensor numbered T3 is associated and mapped with the equipment entity of transformer No. 3, the location entity of area A where transformer No. 3 is located, the temperature monitoring index entity and its rated threshold of 65℃ in the knowledge graph, so as to achieve accurate docking between real-time alarm data and structured knowledge.

[0028] Step 2.2: Based on the reasoning rules of the knowledge graph, risk propagation reasoning is performed on the alarm entities after association mapping to dynamically identify equipment and facilities directly threatened or potentially affected by the fire, and to determine the scope of the fire's impact and key risk points. Specifically, this includes: using pre-set reasoning rules based on the power fire protection knowledge graph in the structured knowledge base. These rules are formulated based on spatial relationships between equipment entities, such as adjacent equipment; power supply relationships, such as upstream and downstream power supply equipment; and risk transmission relationships, such as equipment overheating causing adjacent equipment failures. For example, when the temperature monitoring index attribute value of a certain equipment entity exceeds the rated threshold, the equipment entity is in an overheated state, and the fire will spread to adjacent equipment entities with spatial relationships within 10 minutes. Risk propagation is then performed on the alarm entities after association mapping, such as the overheating alarm entity of transformer No. 3. The reasoning process involves real-time adjustments to the reasoning logic based on the latest monitoring data, such as whether the temperature continues to rise or whether the smoke concentration is spreading. It uses the relationship chains between entities in the knowledge graph to gradually deduce the impact. For example, the overheating of transformer No. 3 leads to the deduction that it may affect the adjacent circuit breaker No. 2, cable tray equipment and facilities in area A. This dynamically identifies equipment and facilities directly threatened by the fire, such as transformer No. 3 itself and equipment within 10 meters of it, as well as potentially affected equipment and facilities, such as the main transformer No. 1 connected via power supply lines. Simultaneously, based on the importance of the equipment and facilities in the substation's power supply system, such as whether they are core power supply equipment or whether they affect the stability of the regional power grid, the boundaries of the fire's impact range are determined, such as an area with a radius of 20 meters centered on transformer No. 3. Key risk points are identified, such as the cable tray in area A near the main power supply line.

[0029] Step 2.3: Based on the identified fire impact range and key risk points, and combining historical case data with emergency response plan rules, generate a preliminary differentiated emergency response plan. This includes: retrieving historical case data from the structured knowledge base that matches the currently identified fire impact range (e.g., a 20-meter radius area) and key risk points (e.g., cable trays in area A). Screening these cases for scenarios similar to the current fire, such as past cases of similar transformers overheating and causing small-scale fires at a substation, where the key risk point is also cable trays. Extracting effective response measures from these historical cases, such as cutting off the power supply to the corresponding transformer, activating the automatic sprinkler system in the vicinity, and assigning personnel to monitor the cable tray temperature on-site. Simultaneously, retrieving standardized emergency response plan rules stored in the knowledge base. These rules include response procedures for different fire scenarios, such as the power-off sequence after equipment overheating alarms, etc. Prioritizing the handling of risk points, such as prioritizing the handling of core power supply equipment, and adjusting handling requirements under different external environments, such as strengthening fire spread monitoring in windy weather, integrates handling measures from historical cases with emergency plan rules. Combining the specific characteristics of the current fire situation, such as the changing trends of alarm parameters, whether the temperature is continuously rising, and the actual state of key risk points, such as whether cable trays have experienced localized high temperatures, a preliminary differentiated emergency response plan is generated. For example, in the current situation of overheating of transformer No. 3 and light wind in area A, the plan clearly states that the upstream power supply switch of transformer No. 3 should be cut off first rather than directly cutting off the transformer power supply to avoid voltage fluctuations. At the same time, windproof baffles should be deployed on the basis of activating the sprinkler system in area A. For the non-critical risk point of circuit breaker No. 2, personnel should be arranged to monitor its status every 5 minutes to ensure that the plan is adapted to the current specific fire situation rather than adopting a uniform handling procedure.

[0030] In this embodiment of the invention, the technology of associating real-time fire alarm data with entities in a structured knowledge base, the risk propagation reasoning method based on knowledge graph reasoning rules, and the mechanism of generating disposal plans by combining historical case data with emergency plan rules overcome the technical problems of dynamic response lag, untimely identification of fire impact range and key risk points, and lack of targeted emergency plans caused by reliance on human experience in traditional emergency decision-making. Thus, it realizes the dynamic identification of equipment threatened by fire and potential risk points, and quickly generates preliminary differentiated emergency disposal plans adapted to the real-time fire situation, providing timely and effective decision support for the golden response window of fire emergency response.

[0031] In a preferred embodiment of the present invention, step 2 above may include: Step 2.4 involves querying a structured knowledge base to extract the spatial coordinates and layout information of the identified key risk points and their surrounding monitoring locations. Specifically, this includes retrieving the power fire protection knowledge graph through the structured knowledge base's query interface, inputting the identified key risk point identifiers such as cable trays in area A and the high-temperature point of transformer No. 3, and extracting the spatial coordinate data corresponding to these key risk points from the knowledge graph, including latitude and longitude coordinates and relative coordinates within the substation. Simultaneously, information on all monitoring locations within a 50-meter radius of the key risk points is extracted, covering the sensor numbers corresponding to the monitoring locations, such as temperature sensor T5 and smoke sensor Y3, the installation spatial coordinates of each sensor, the relative positional relationship between the sensor and the key risk point (e.g., located 3 meters east of the key risk point), and the layout information of surrounding equipment and facilities, such as whether there are cable trenches near the sensors, whether the equipment spacing is less than 2 meters, and whether there are firebreaks. The extracted spatial coordinates are then categorized and organized according to the key risk points, forming a spatial information set centered on each key risk point.

[0032] Step 2.5: Based on spatial coordinates and layout information, construct a topology network reflecting the spatial connections and proximity relationships between key risk points, monitoring locations, and surrounding equipment and facilities. Specifically, this includes: using network modeling tools to construct a topology network based on the extracted spatial coordinates and layout information; designating key risk points and monitoring locations as core nodes and monitoring nodes, respectively; and designating surrounding equipment and facilities, such as adjacent circuit breakers (No. 2) and cable trenches, as associated nodes. Calculate the straight-line distance and actual connectivity between nodes based on spatial coordinates. For example, if the straight-line distance between the cable tray in area A (key risk point) and monitoring node T5 is 3 meters and there are no physical obstacles, then a direct proximity connection edge is established between the two nodes. If transformer No. 3 (key risk point) and associated node No. 2 (circuit breaker) are connected via a power supply line with a line length of 8 meters, then a composite connection edge combining power supply association and spatial proximity is established. Simultaneously, label the attribute information of each connection edge in the topology network, including node distance, connectivity type (e.g., spatial connectivity, functional connectivity), and the presence of fire protection facilities. Finally, a topology network is formed that clearly reflects the tightness of spatial connections and proximity relationships between key risk points, monitoring locations, and surrounding equipment and facilities.

[0033] Step 2.6: The real-time fire parameter data acquired from monitoring is combined with the topology network and processed using a spatial feature fitting algorithm. Based on the spatial distance between monitoring points and key risk points, real-time fire parameter values ​​and their changing trends, weights are assigned to monitoring data at different locations. An interpolation method is then used to dynamically calculate the fire intensity distribution across the entire area and determine the boundary parameters of the fire-affected area. Specifically, this includes acquiring fire parameter data from the substation's real-time monitoring system for key risk points and surrounding monitoring locations, including real-time temperature values, smoke concentration values, temperature change rate increase every 5 minutes, and smoke diffusion speed at each monitoring point. These real-time fire parameter data are then associated and bound to the topology network constructed in Step 2.5. Corresponding real-time parameter attributes are added to each monitoring node in the topology network. Subsequently, the spatial feature fitting algorithm is activated to process the associated data. The algorithm first assigns basic weights based on the spatial distance between monitoring points and key risk points. Monitoring points closer to critical risk points have higher weights. For example, a point within 1 meter has a weight of 0.9, and a point 3-5 meters away has a weight of 0.6. The weights are then adjusted based on the magnitude and trend of real-time fire parameters. For instance, a monitoring point with a real-time temperature exceeding 90°C and rising by 10°C every 5 minutes has its weight increased by 0.2, while a monitoring point with a real-time smoke concentration below the alarm threshold and with stable changes has its weight decreased by 0.1. After weight allocation, a spatial interpolation method is used to supplement the fire parameter data of areas without monitoring equipment in the topology network based on the weighted fire parameters of each monitoring point. For example, the temperature distribution between two adjacent monitoring points can be estimated by using their weighted temperature values, thereby dynamically generating a heat map of the fire intensity distribution in the entire area. The boundary parameters of the fire-affected area are determined based on the fire intensity distribution. For example, the outermost coordinates of the area covered by the temperature and smoke concentration values ​​corresponding to a moderate fire intensity level are used as the boundary parameters.

[0034] Step 2.7: Based on the obtained boundary parameters, define and update the geometric range of the fire-affected area in real time. The geometric range is represented by a set of boundary coordinates or a regional outline. Specifically, this includes: converting the calculated boundary parameters into specific spatial coordinate data; if the boundary parameters correspond to a region with a temperature exceeding 85℃, then selecting the spatial coordinates of all monitoring points with a temperature reaching 85℃ and the interpolated virtual points with a temperature reaching 85℃, and connecting these coordinates sequentially in a clockwise or counterclockwise order to form the geometric range of the fire-affected area represented by a set of boundary coordinates or a planar regional outline. Simultaneously, a data update trigger mechanism is set. The system receives new real-time monitoring data every 30 seconds, re-executes the calculation process in step 2.6 to update the boundary parameters, and adjusts the geometric range of the fire-affected area based on the updated boundary parameters. For example, when new monitoring data shows that the temperature in the eastern area rises to 85℃, the corresponding coordinate point of the area is added to the boundary coordinate set to expand the eastern boundary of the geometric range. When the temperature in the western area drops below 85℃, the corresponding coordinate point of the area is deleted from the boundary coordinate set to shrink the western boundary of the geometric range. This ensures that the geometric range of the fire-affected area can be updated in real time to keep up with the changes in the fire situation, and intuitively reflects the changes in the spatial range after the fire spreads or is controlled.

[0035] In this embodiment of the invention, by employing techniques such as extracting spatial coordinates and layout information of key risk points and monitoring locations from a structured knowledge base, constructing a topological network reflecting spatial connectivity and proximity relationships, and combining real-time fire parameters with spatial feature fitting algorithms to allocate weights and using interpolation calculations, the technical problems of traditional emergency decision-making, such as the inability to accurately define the boundaries of fire-affected areas in real time, the difficulty in dynamically reflecting the fire spread trend, and the unknown fire status in unmonitored areas, are overcome. This enables dynamic estimation of the fire intensity distribution across the entire area, accurately calculates the boundary parameters of the fire-affected area, and updates its geometric range in real time, providing a visualized and dynamic spatial distribution basis for emergency response.

[0036] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Based on the determined, dynamically updated fire impact area, a gridding method based on geographic coordinates or predefined dimensions is used to divide the area into several independent spatial analysis units. Specifically, this includes: obtaining the set of boundary coordinates of the dynamically updated fire impact area; if geographic coordinates are used as the basis for gridding, the fire impact area is divided into several rectangular spatial analysis units at intervals of 0.0001 degrees of longitude and 0.0001 degrees of latitude, using the predefined coordinate system inside the substation as the reference. The geographic coordinate range of each unit is clear and non-overlapping; if predefined dimensions are used as the basis for division, the unit size is determined according to the equipment layout density of the substation. For example, the unit size is set to 5 meters × 5 meters in densely populated areas and 10 meters × 10 meters in sparsely populated areas. The fire impact area is evenly divided into multiple independent square spatial analysis units using coordinate calculation tools. During the division process, it is ensured that all units completely cover the fire impact area, and there are no gaps or overlaps between adjacent units. Each unit is assigned a unique identifier number.

[0037] Step 3.2: Based on each divided spatial analysis unit, associate or deploy corresponding monitoring equipment according to geographical location, and collect real-time fire parameters within the coverage area of ​​the unit. Specifically, this includes: for each divided spatial analysis unit, query the existing monitoring equipment deployment information of the substation recorded in the structured knowledge base according to the unique identifier number and the corresponding geographical coordinate range, and filter out the monitoring equipment whose coordinates fall within the unit range, such as temperature sensors, smoke concentration sensors, flame detectors, etc., and establish the association between the unit and the existing monitoring equipment; if there is no existing monitoring equipment in a spatial analysis unit, deploy temporary monitoring equipment, such as portable temperature detectors and mobile smoke sensors, according to the location characteristics of the unit, such as whether it is close to critical equipment or whether it is a fire spread channel. The temporary equipment is connected to the real-time monitoring system through a wireless communication module; then set the acquisition frequency of each monitoring device, such as once every 10 seconds for densely equipped units and once every 30 seconds for sparsely equipped units, and collect real-time fire parameters within the coverage area of ​​the unit, including ambient temperature, smoke concentration, oxygen content, and whether flames are detected.

[0038] Step 3.3 involves performing time-series variation characteristic analysis on the real-time fire parameters collected in each spatial analysis unit. This involves calculating the rate of change, extreme values, and fluctuation amplitude of each parameter per unit time using a sliding window, and extracting characteristic indicators representing the dynamic evolution of the fire situation in each unit. Specifically, this includes: retrieving the time-series data of the real-time fire parameters bound to each spatial analysis unit from the data acquisition database; using a sliding window with a time length of 5 minutes to extract the time-series data, with a window sliding step size of 1 minute; and calculating the rate of change of each parameter per unit time for the parameter data within each sliding window, such as the temperature change rate being the average temperature within the current window. The difference between the average temperature in the current window and the average temperature in the previous window is divided by 1 minute. The smoke concentration change rate is the difference between the maximum smoke concentration in the current window and the maximum smoke concentration in the previous window, divided by 1 minute. At the same time, the extreme values ​​of each parameter in the window are calculated, such as the maximum temperature, the minimum temperature, the maximum smoke concentration, and the fluctuation range of the parameters, i.e., the difference between the maximum and minimum values. Based on the calculation results, characteristic indicators that represent the dynamic evolution of the fire situation in each unit are extracted, such as whether the temperature change rate exceeds 5°C per minute, whether the maximum smoke concentration reaches the alarm threshold, and whether the flame detection duration exceeds 1 minute. Each characteristic indicator corresponds to a clear judgment standard.

[0039] Step 3.4: Based on the characteristic indicators of each unit, and considering their contribution to the overall fire situation, weighting factors are assigned. A multi-indicator weighted fusion algorithm is used to normalize and aggregate the characteristic indicators of all units, resulting in a comprehensive adjustment parameter for quantitative assessment and dynamic reflection of the current overall fire situation. Specifically, this includes: analyzing the contribution of the characteristic indicators of each spatial analysis unit to the overall fire situation; if the unit contains core equipment such as transformers and main power supply lines, their temperature change rate and flame detection status have a greater impact on the overall fire situation, and these indicators are assigned a weighting factor of 0.8; if the unit is a general passage area, its smoke concentration, oxygen... Characteristic indicators such as gas content have a low contribution level and are assigned a weighting factor of 0.5. For characteristic indicators of all units, a linear normalization method is used to transform the indicator values ​​to the range of 0 to 1. For example, the temperature change rate is mapped from -2℃ to 10℃ per minute to 0 to 1. The higher the change rate, the larger the normalized value. Then, a multi-indicator weighted fusion algorithm is used to multiply the normalized value of each characteristic indicator in each unit by the corresponding weighting factor to obtain the weighted index value of that unit. Then, the weighted index values ​​of all units are summed to obtain a comprehensive adjustment parameter with a value range of 0 to 100. The higher the parameter value, the more severe the current overall fire situation.

[0040] In this embodiment of the invention, the technical means of dynamically updating the geographical coordinates or predefined size of the fire-affected area to divide the spatial analysis unit, associating or deploying monitoring equipment for each unit to collect real-time fire parameters, analyzing the time-series change characteristics of parameters through a sliding window to extract situation indicators, allocating weighting factors based on the contribution of indicators, and using a multi-indicator weighted fusion algorithm to calculate comprehensive adjustment parameters overcome the technical problems in traditional emergency decision-making, such as the rough assessment of the fire-affected area, the lack of precise analysis of the dynamic evolution characteristics of fire parameters, and the lack of data support for decision-making due to the lack of quantified overall fire situation parameters. This enables refined monitoring and situation decomposition of the fire-affected area, extracts the dynamic evolution law of fire in each local area, and obtains quantifiable and real-time comprehensive adjustment parameters that reflect the overall fire situation, thereby improving the dynamic adaptability of power fire emergency decision-making.

[0041] In a preferred embodiment of the present invention, step 3 above may include: Step 3.5 compares the comprehensive adjustment parameters with the preset adjustment thresholds of the emergency plan. Based on the comparison results, it determines the optimization direction and intensity of the preliminary differentiated emergency response plan. Specifically, this includes retrieving the preset adjustment threshold system from the structured knowledge base. The threshold system is constructed based on the comprehensive adjustment parameters corresponding to different fire situations in historical fire cases and the risk level classification standards specified in the emergency plan rules. The thresholds are divided into three intervals: low-risk interval 0-30, medium-risk interval 31-70, and high-risk interval 71-100. Each interval corresponds to a different optimization direction and optimization intensity standard. For example, the low-risk interval corresponds to the optimization direction of fine-tuning existing response measures, and the optimization intensity... The optimization direction is set at 10%-20%, corresponding to supplementary response measures and adjustments to resource allocation in the medium-risk range, with an optimization intensity of 30%-50%. The optimization direction is set at 60%-100%, corresponding to strengthening core response measures and prioritizing the allocation of key resources in the high-risk range. Then, the calculated current comprehensive adjustment parameter is obtained. For example, if the parameter value is 65, it is compared with the preset threshold range to determine that the parameter is in the medium-risk range. Then, according to the standard corresponding to the range, the optimization direction is determined to be supplementary cooling measures and an increase in the allocation ratio of fire extinguishing resources, with an optimization intensity of 40%. That is, 40% of cooling-related measures need to be added to the initial plan, and the allocation of fire extinguishing resources needs to be increased by 40%.

[0042] Step 3.6: Based on the optimization direction and intensity, invoke the emergency response rules and historical handling cases in the structured knowledge base to dynamically adjust the initial plan's handling measures, resource allocation plan, and execution priority, obtaining the latest handling strategy adapted to the current fire situation. Specifically, this includes: supplementing cooling measures and increasing the proportion and optimization intensity of fire extinguishing resource allocation by 40% based on the determined optimization direction; invoking the emergency response rules and historical handling cases in the structured knowledge base through the query interface; filtering out content related to equipment cooling and fire extinguishing resource allocation in the rules, such as deploying mobile cooling fans when the fire in the transformer area is at medium risk; prioritizing the allocation of dry powder fire extinguishers and fire hoses within a 30-meter range; and simultaneously matching the comprehensive adjustment parameters in historical cases. For cases with fires in the 60-70 range and where the affected area includes transformers, effective cooling measures were extracted, such as checking the cooling fan operation status every 5 minutes and resource allocation plans, such as prioritizing the use of the No. 2 fire cabinet in the substation. Based on these rules and case information, the preliminary differentiated emergency response plan was dynamically adjusted. In terms of response measures, three mobile cooling fans were deployed to the vicinity of the target transformer, and the cooling equipment was inspected every 5 minutes. In terms of resource allocation plans, the originally planned two dry powder fire extinguishers were increased to three, and one additional fire hose was added. In terms of execution priority, the priority of the cooling fan deployment task was raised from the original level three to level two, and it was executed in parallel with the task of cutting off the power supply to the target equipment. Finally, a new response strategy adapted to the current medium-risk fire situation was formed.

[0043] Step 3.7 decomposes the latest response strategy into specific, actionable tasks, and generates and encapsulates operation instruction sets for different roles based on their responsibilities, permissions, and terminal types. Specifically, this includes breaking down the latest response strategy into specific, actionable tasks according to the operational flow. For example, cutting off the upstream power supply to the target transformer is broken down into logging into the substation power monitoring system, locating the upstream switch corresponding to the target transformer, confirming the switch's current status, performing the tripping operation, and recording the operation time. Deploying cooling fans is broken down into receiving three mobile cooling fans, transporting them to a designated location near the target transformer, connecting the power supply, starting the fans, and checking the fan's airflow status. Then, the roles in the substation fire emergency response are identified, including the substation duty officer, on-site firefighters, and fire control room operator, clarifying the responsibilities and permissions of each role, such as the duty officer... Staff are responsible for operating power equipment, firefighters are responsible for on-site equipment deployment and fire monitoring, and control room operators are responsible for system status monitoring. The instruction format is adapted according to the terminal type of each role: the power monitoring terminal used by the duty officer is adapted to text and image instructions, including system login path, switch location screenshots, and text descriptions of operation steps; the handheld terminal used by the firefighter is adapted to concise text and voice instructions, including voice prompts for equipment pickup location, transportation route, and key operation points; the monitoring terminal used by the control room operator is adapted to data monitoring instructions, including the equipment number requiring key monitoring and parameter threshold ranges. The decomposed tasks are categorized by role and packaged into corresponding instruction sets. Each instruction set includes a task identifier, execution time limit, operation standards, and anomaly feedback path, ensuring that each role can obtain operation instructions that conform to their responsibilities and terminal usage habits.

[0044] In this embodiment of the invention, the following technical means are used: comparing comprehensive adjustment parameters with preset contingency plan adjustment thresholds to determine the optimization direction and intensity; dynamically adjusting the disposal measures, resource allocation plans, and execution priorities of the preliminary differentiated emergency response plan by calling emergency plan rules and historical disposal cases in the structured knowledge base; and decomposing the latest disposal strategy into operable task items and generating encapsulated operation instruction sets for different roles based on the responsibilities and permissions of the execution roles and terminal types. Therefore, this overcomes the technical problems in traditional emergency decision-making, such as the lack of quantitative basis for adjusting the disposal plan, which leads to strong blindness; static and fixed plans that cannot adapt to dynamic fire situation changes; and strategies that are not split according to execution roles, which leads to chaotic operations and low execution efficiency. Thus, it achieves precise targeted optimization of the preliminary disposal plan, obtains the latest disposal strategy that is highly adapted to the current fire situation, and enables operation instructions to accurately match the needs of different execution roles, ensuring the efficient and orderly progress of emergency operations.

[0045] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Receive operation instruction sets for different execution roles, and parse the instruction content and its corresponding target terminal attributes. Specifically, this includes: receiving operation instruction sets generated through emergency decision-making instructions for substation operators, on-site firefighters, and fire control room operators; starting the instruction parsing engine to decompose each type of instruction set; extracting core task items from the instruction content, such as cutting off the upstream power supply switch of transformer No. 2, deploying 3 mobile cooling fans in area A, execution time limits such as completing power cut-off within 3 minutes, completing fan deployment within 10 minutes, and operation standards such as confirming the equipment number and status are correct before cutting off the switch, and placing the fans within 5 meters of the transformer. Within a 1-meter radius, in case of operational failure, a fault description is immediately submitted to the control room via the terminal at ventilation points and abnormal feedback paths. Simultaneously, the target terminal attributes corresponding to each type of instruction set are analyzed, including the terminal device type (e.g., power monitoring terminal used by duty officers, handheld smart terminal used by firefighters, large-screen monitoring terminal used in the control room), the communication protocols supported by the terminal (e.g., Modbus protocol for power monitoring terminals, 4G or 5G protocol for handheld terminals), the terminal display parameters (e.g., 720P resolution for handheld terminals, 4K resolution for large-screen terminals), and the associated functional permissions of the terminal (e.g., duty officer terminals have power equipment operation permissions, firefighter terminals have task feedback permissions).

[0046] Step 4.2: Based on the obtained instruction content and target terminal attributes, and considering the functional authority of the executing entity and the terminal's perception requirements, the instruction content is adapted and converted into a multimodal information format. Specifically, this includes: adapting and converting the instruction content and target terminal attributes based on the functional authority of each executing entity and the terminal's perception requirements. For the substation operator's power monitoring terminal, whose function is to operate power equipment and has large-screen interaction and system access capabilities, the instruction to cut off the upstream power supply switch of transformer No. 2 is converted into a graphic and text modal information format containing the power monitoring system login address, a screenshot of the location of the upstream switch of transformer No. 2 in the system interface, and a textual description of the tripping operation steps. For the handheld smart terminal of the on-site firefighter, whose function is to perform on-site equipment deployment and fire monitoring... Given the small size of the terminal screen and the presence of noise interference in the on-site environment, the instructions to deploy three mobile cooling fans in Area A were converted into text prompts including the fan pickup location (e.g., the fire supply warehouse on the west side of the substation), a route map from the warehouse to Area A, instructions to check the fan power supply upon pickup, and voice messages confirming the airflow direction towards the equipment after deployment. The voice prompts were delivered in a clear female voice at a pace of 110 words per minute. For the large-screen monitoring terminal for the fire control room operator, since its function is global monitoring and collaborative scheduling, and the terminal supports multi-window parallel display, the instructions for tracking the task progress of all roles and monitoring equipment operating parameters were converted into data visualization modal information including real-time task completion rate charts, key parameter data curves such as transformer temperature, and abnormal situation pop-up prompts.

[0047] Step 4.3 involves visualizing and rendering the multimodal information to generate an interactive command interface that includes graphics, text, voice, and augmented reality elements. Specifically, this includes: starting the multimodal information rendering engine to process the converted multimodal information; for graphic and textual modal information, using a 3D layout diagram of the substation equipment as the background, highlighting the position of the upstream switch of transformer No. 2 to be operated in red, and marking the operation path from the monitoring system homepage to the switch control interface in green dashed lines; using Microsoft YaHei font for the text, setting the font size on the power monitoring terminal to 14 for large-screen display, and setting the font size on the handheld terminal to 12 to ensure screen stability. Clear display; for voice modal information, professional voice synthesis technology is used for recording, with a 0.6-second pause at key operation prompts such as confirming that the fan power is normal, and the voice tone is strengthened at task time limit reminders such as 5 minutes left to complete deployment, to ensure that on-site firefighters can accurately capture key information; for augmented reality modal information, after the terminal camera collects on-site environmental images, it automatically matches transformers and cooling fans in the substation equipment model library to generate semi-transparent operation guidance signs, such as overlaying arrow icons at the fan deployment location and overlaying dynamic prompts for clicking to open the switch operation button; finally, an interactive command interface is generated.

[0048] Step 4.4 involves distributing the rendered multimodal command information to the terminal devices of the corresponding user roles via an encrypted secure communication protocol. This includes: establishing a dedicated communication link between the emergency decision-making system and each terminal device; encrypting the rendered multimodal command information using SSL or TLS security protocols; generating a unique session key during encryption and storing it in the secure storage system of the system and terminals; and pre-storing a digital certificate issued by the substation safety certification center for authentication before communication. The command distribution scheduling is then initiated, allocating transmission priorities based on the real-time network status of each terminal device. Handheld terminals used by on-site firefighters are prioritized for rapid command reception to maximize emergency response time. The highest priority is given to the power monitoring terminal of the substation operator, the medium priority is given to the large screen monitoring terminal of the fire control room, and the low priority is given to the large screen monitoring terminal of the fire control room. When distributing, an authentication request is first sent to each terminal device. After the terminal returns the digital certificate, the system verifies the validity of the certificate and the consistency with the bound role. After the verification is successful, the multimodal instruction information is encapsulated into data frames according to the communication protocol of the corresponding terminal, such as 4G protocol data frames of handheld terminals and Modbus protocol data frames of power monitoring terminals. During the transmission process, the data transmission progress is monitored in real time. If a network interruption occurs, the system automatically triggers the breakpoint resume mechanism. After the network is restored, the transmission continues from the interruption point to ensure that the instruction information is delivered to the terminal device of the corresponding role completely and securely.

[0049] Step 4.5: Based on the instruction information, drive each terminal device to receive and display the corresponding instructions, and collaboratively trigger and execute the corresponding emergency response operations. Specifically, this includes: after receiving the multimodal instruction information, each terminal device automatically starts the instruction display program; the power monitoring terminal of the substation operator pops up a graphic instruction window on the operation interface, accompanied by a slight prompt sound for 3 seconds, and the window is displayed at the top to ensure that the operator can view it first; the handheld terminal of the on-site firefighter automatically lights up the screen, displays text instructions on the main interface and plays voice prompts, and stops after repeating the voice twice to avoid interfering with on-site operations; the large screen monitoring terminal in the fire control room loads a data visualization instruction interface in the left third area, displaying real-time... Refresh the task progress and equipment parameters of each role; after the role views the instructions, start the emergency response operation. After the operation is completed, click the operation completion confirmation button on the terminal interface. The system will synchronize the operation results to all relevant terminals. For example, after the duty officer completes the power cut-off, the control room screen and the firefighter's handheld terminal will both display a prompt that the power supply to transformer No. 2 has been cut off. If an abnormality occurs during the execution process, such as the cooling fan failing to start, the executing role submits a fault description through the abnormal situation feedback button. After receiving the report, the system will immediately call the fault solution in the structured knowledge base, such as checking whether the fan power interface is loose. If it is loose, replug it and start the fan. The system will also push the solution to the corresponding terminal to guide the executing role in troubleshooting.

[0050] In this embodiment of the invention, the technical means of receiving and parsing operation instruction sets for different execution roles and corresponding target terminal attributes, adapting and converting instructions into multimodal information forms by combining the functional authority of the execution subject and the terminal's perception requirements, visually rendering the multimodal information to generate an interactive instruction interface, distributing instruction information through an encrypted secure communication protocol, and driving various terminal devices to collaboratively receive and display instructions and execute emergency operations are adopted. Therefore, it overcomes the technical problems of traditional emergency decision-making, such as the single instruction form that cannot adapt to the needs of different terminals and roles, the lack of security guarantee for instruction transmission that is prone to information leakage or tampering, and the poor coordination between multiple roles and terminals that leads to delays in operation judgment and chaotic step connection. Thus, it achieves accurate adaptation of operation instructions to different execution roles and terminal devices, ensures the security and integrity of instruction transmission, promotes efficient multi-terminal collaborative execution of emergency response operations, shortens the time from instruction generation to execution, and ensures that key operations are completed within the golden response window for fire emergency response.

[0051] like Figure 2 As shown, embodiments of the present invention also provide a knowledge graph-based power fire emergency decision-making system, comprising: The acquisition module is used to build a power fire protection knowledge graph. By integrating substation equipment ledger data, real-time monitoring data, external environmental data and historical case data, and establishing the relationship between multi-source data, a structured knowledge base that supports emergency decision-making is obtained. The calculation module is used to perform dynamic risk reasoning based on a structured knowledge base and real-time fire alarm data, identify the fire impact range and key risk points, and obtain preliminary differentiated emergency response plans based on the reasoning results; extract the spatial features of key risk points and monitoring locations, and dynamically determine and update the boundary parameters of the fire impact area through a spatial feature fitting algorithm, thereby defining the dynamically updated fire impact area. The adjustment module is used to divide the fire-affected area into multiple analysis units based on the fire-affected area, collect real-time fire parameters in each analysis unit, and calculate a comprehensive adjustment parameter based on the parameter change characteristics. Based on the comprehensive adjustment parameter, the initial differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. The processing module is used to adapt and visualize the multi-role operation instruction set according to the functional authority of the executing subject and the terminal perception requirements, and distribute it to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

[0052] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A power fire emergency decision-making method based on knowledge graphs, characterized in that, The method includes: A knowledge graph for power fire protection is constructed by integrating substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data, and establishing the relationships between multiple data sources to obtain a structured knowledge base that supports emergency decision-making. Based on a structured knowledge base and real-time fire alarm data, dynamic risk reasoning is performed to identify the fire's impact range and key risk points. Based on the reasoning results, preliminary differentiated emergency response plans are obtained. Spatial features of key risk points and monitoring locations are extracted, and the boundary parameters of the fire-affected area are dynamically determined and updated through a spatial feature fitting algorithm, thus defining the dynamically updated fire-affected area. Based on the fire-affected area, the fire-affected area is divided into multiple analysis units according to rules. Real-time fire parameters are collected in each analysis unit, and a comprehensive adjustment parameter is calculated based on the parameter change characteristics. Based on the comprehensive adjustment parameter, the preliminary differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. The multi-role operation instruction set is adapted to the functional authority of the executing entity and the terminal perception requirements, and then processed for multi-modal information adaptation and visualization rendering. It is then distributed to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

2. The power fire emergency decision-making method based on knowledge graphs according to claim 1, characterized in that, A power fire protection knowledge graph is constructed by integrating substation equipment ledger data, real-time monitoring data, external environmental data, and historical case data, and establishing relationships between multiple data sources to obtain a structured knowledge base that supports emergency decision-making, including: Integrate substation equipment ledger data, real-time monitoring data, external environment data, and historical case data to form a multi-source heterogeneous data source; Based on multi-source heterogeneous data sources, entities, attributes, and relationships in the power fire protection field are extracted, and an ontology schema is defined. Based on the ontology model, multi-source data are fused and mapped to construct a power fire protection knowledge graph containing entities, attributes, and relationships. The knowledge graph of power fire protection is stored in a graph database to form a structured knowledge base that supports emergency decision-making.

3. The power fire emergency decision-making method based on knowledge graphs according to claim 2, characterized in that, Based on a structured knowledge base and real-time fire alarm data, dynamic risk reasoning is performed to identify the fire's impact range and key risk points. Based on the reasoning results, preliminary differentiated emergency response plans are derived, including: Receive real-time fire alarm data and associate and map the fire alarm data with entities in a structured knowledge base; Based on the reasoning rules of knowledge graphs, risk propagation reasoning is performed on alarm entities after association mapping, dynamically identifying equipment and facilities directly threatened and potentially affected by fire, and determining the scope of fire impact and key risk points. Based on the identified fire impact range and key risk points, and combined with historical case data and emergency response plan rules, a preliminary differentiated emergency response plan is generated.

4. The power fire emergency decision-making method based on knowledge graphs according to claim 3, characterized in that, Spatial features of key risk points and monitoring locations are extracted. A spatial feature fitting algorithm is used to dynamically determine and update the boundary parameters of the fire-affected area, defining the dynamically updated fire-affected area, including: By querying the structured knowledge base, the spatial coordinates and layout information of the identified key risk points and their surrounding monitoring locations are extracted; Based on spatial coordinates and layout information, a topological network is constructed that reflects the spatial connections and proximity relationships between key risk points, monitoring locations, and surrounding equipment and facilities. The real-time fire parameter data is combined with the topology network and processed by a spatial feature fitting algorithm. Based on the spatial distance between the monitoring point and the key risk point, the real-time fire parameter values ​​and their changing trends, weights are assigned to the monitoring data at different locations. An interpolation calculation method is used to dynamically calculate the fire intensity distribution of the entire area and calculate the boundary parameters of the fire-affected area. Based on the obtained boundary parameters, the geometric range of the fire-affected area is defined and updated in real time. The geometric range is represented in the form of a set of boundary coordinates or a region outline.

5. The power fire emergency decision-making method based on knowledge graphs according to claim 4, characterized in that, Based on the fire-affected area, the fire-affected area is regularly divided into multiple analysis units. Real-time fire parameters are collected within each analysis unit, and a comprehensive adjustment parameter is calculated based on the parameter change characteristics, including: Based on the determined, dynamically updated fire impact area, a gridding method based on geographic coordinates or predefined dimensions is used to regularly divide the area into several independent spatial analysis units. Based on each spatial analysis unit, corresponding monitoring equipment is associated or deployed according to the geographical location, and real-time fire parameters within the coverage area of ​​that unit are collected. The time-series change characteristics of real-time fire parameters collected in each spatial analysis unit are analyzed. The rate of change, extreme values ​​and fluctuation amplitude of each parameter per unit time are calculated by sliding window, and characteristic indicators that characterize the dynamic evolution of the fire situation in each unit are extracted. Based on the characteristic indicators of each unit, weighting factors are assigned to them according to the degree of contribution of the characteristic indicators to the overall fire situation. A multi-indicator weighted fusion algorithm is used to normalize and aggregate the characteristic indicators of all units to obtain a comprehensive adjustment parameter for quantitative assessment and dynamic reflection of the current overall fire situation.

6. The power fire emergency decision-making method based on knowledge graphs according to claim 5, characterized in that, Based on comprehensive adjustment parameters, the initial differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operational instruction set for different execution roles is generated, including: The comprehensive adjustment parameters are compared with the preset adjustment thresholds of the contingency plan, and the optimization direction and intensity of the preliminary differentiated emergency response plan are determined based on the comparison results. Based on the optimization direction and intensity, the emergency response rules and historical handling cases in the structured knowledge base are invoked to dynamically adjust the initial plan's handling measures, resource allocation plan, and execution priority, thereby obtaining the latest handling strategy adapted to the current fire situation; The latest handling strategy is broken down into specific, actionable tasks, and based on the responsibilities and permissions of the execution roles and the terminal type, operation instruction sets are generated and packaged for different execution roles.

7. The power fire emergency decision-making method based on knowledge graphs according to claim 6, characterized in that, The multi-role operation instruction set is adapted to the functional authority of the executing entity and the terminal's perception requirements, and then processed for multi-modal information adaptation and visualization rendering. It is then distributed to the corresponding role's terminal devices via a secure communication protocol, driving multi-terminal collaborative execution of emergency response operations, including: Receive operation instruction sets for different execution roles, and parse the instruction content and its corresponding target terminal attributes; Based on the obtained instruction content and target terminal attributes, combined with the functional authority of the executing entity and the terminal's perception requirements, the instruction content is adapted and converted into a multimodal information format; Visualize and render multimodal information to generate an interactive command interface that includes graphics, text, voice, and augmented reality elements; The multimodal instruction information generated by rendering is distributed to the terminal devices of the corresponding user roles through an encrypted secure communication protocol; Based on the instruction information, the system drives each terminal device to receive and display the corresponding instructions, and coordinates to trigger and execute the corresponding emergency response operations.

8. A knowledge graph-based power fire emergency decision-making system, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to build a power fire protection knowledge graph. By integrating substation equipment ledger data, real-time monitoring data, external environmental data and historical case data, and establishing the relationship between multi-source data, a structured knowledge base that supports emergency decision-making is obtained. The calculation module is used to perform dynamic risk reasoning based on a structured knowledge base and real-time fire alarm data, identify the fire impact range and key risk points, and obtain preliminary differentiated emergency response plans based on the reasoning results; extract the spatial features of key risk points and monitoring locations, and dynamically determine and update the boundary parameters of the fire impact area through a spatial feature fitting algorithm, thereby defining the dynamically updated fire impact area. The adjustment module is used to divide the fire-affected area into multiple analysis units based on the fire-affected area, collect real-time fire parameters in each analysis unit, and calculate a comprehensive adjustment parameter based on the parameter change characteristics. Based on the comprehensive adjustment parameters, the initial differentiated emergency response plan is dynamically optimized to obtain the latest response strategy adapted to the current fire situation, and an operation instruction set for different execution roles is generated. The processing module is used to adapt and visualize the multi-role operation instruction set according to the functional authority of the executing subject and the terminal perception requirements, and distribute it to the corresponding role terminal devices through a secure communication protocol to drive multi-terminal collaborative execution of emergency response operations.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images