Data-driven fire early warning and decision support method and system

Abstract

The invention discloses a data-driven fire early warning and decision support method and system, and relates to the field of mountain photovoltaic power station safety. The method comprises the following steps: constructing a multi-source data acquisition system, and directionally acquiring equipment operation data, environment perception data, terrain space data and video image data of a mountain photovoltaic power station; the method comprises the following steps: constructing a knowledge graph in the field of mountain photovoltaic fire, performing entity matching, attribute alignment and graph reasoning on collected multi-source data based on the knowledge graph, realizing semantic level fusion of multi-source heterogeneous data, and generating a structured fusion data set; constructing a fire early warning model based on the structured fusion data set, mining fire risk features in the data, and outputting a fire risk level; and triggering decision support according to the fire hazard grade, calling a geographic information system, integrating real-time fire behavior with environment and topographic data in combination with the entity association relationship of the knowledge graph, and generating a fire extinguishing and rescue scheme. The scientificity of fire prevention and control of the mountain photovoltaic power station is improved.

Description

Technical Field

[0001] This invention relates to the field of safety in mountain photovoltaic power stations, and more specifically to a data-driven method and system for fire early warning and decision support. Background Technology

[0002] Mountainous photovoltaic power stations are located in complex terrains such as hills and mountains, with photovoltaic modules scattered along slopes. The surrounding area is often accompanied by flammable vegetation such as shrubs and weeds, making fire prevention and control significantly more difficult than in plains power stations. The causes of fires are diverse. In addition to conventional factors such as high temperatures and exposure to sunlight, aging equipment and short circuits, and inverter failures, lightning strikes caused by severe convective weather in mountainous areas, module damage and leakage due to melting ice in winter, and secondary fires caused by spontaneous combustion of surrounding forest vegetation can all rapidly ignite fires. Furthermore, the elevation differences and ravines in mountainous terrain allow fires to spread rapidly along slopes and vegetation belts, and rescue vehicles and personnel have difficulty quickly reaching the core area, easily causing severe losses such as mass damage to photovoltaic modules and power grid outages.

[0003] Existing fire early warning technologies mostly rely on single-point sensors such as temperature and smoke detectors, resulting in warning lags. Furthermore, mountainous terrain can cause unstable sensor signal transmission, leading to high false alarm and missed alarm rates. In terms of decision support, traditional methods are largely based on pre-set plans and lack comprehensive analysis of real-time fire data, terrain data, and meteorological data. This makes it difficult to provide dynamic and accurate rescue decision-making suggestions for complex mountainous scenarios. Key decisions such as fire extinguishing route planning and resource allocation optimization lack data support.

[0004] With the large-scale development of the photovoltaic industry, the need for fire prevention and control in mountainous photovoltaic power stations is becoming increasingly urgent. There is a pressing need for an early warning and decision support technology that integrates multi-source data and has real-time analysis capabilities to improve the timeliness and accuracy of fire early warning, while providing scientific data support for rescue decisions and reducing fire losses. Summary of the Invention

[0005] In view of this, the present invention provides a data-driven fire early warning and decision support method and system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A data-driven fire early warning and decision support method includes the following steps: Construct a distributed, multi-source data acquisition system covering equipment, environment, terrain, and images; collect equipment operation data, environmental perception data, terrain spatial data, and video image data from mountain photovoltaic power stations in a targeted manner to achieve comprehensive acquisition of disaster monitoring data. A knowledge graph for the field of mountain photovoltaic fires is constructed. The knowledge graph includes core entities of photovoltaic equipment, meteorological factors, terrain units, and fire events, as well as the relationships between entities. Based on the knowledge graph, entity matching, attribute alignment, and graph reasoning are performed on the collected multi-source data to achieve semantic-level fusion of multi-source heterogeneous data and generate a structured fusion dataset. A fire early warning model is constructed based on a structured fusion dataset. The fire risk characteristics in the data are mined through the fire early warning model, and four levels of fire risk are output: no risk, low risk, medium risk, and high risk, so as to achieve accurate monitoring and early warning of disaster risks. When the fire risk level is medium or high, decision support is triggered, which calls the geographic information system and combines the entity association relationship of the knowledge graph to integrate real-time fire situation, environmental and terrain data to generate a fire fighting and rescue plan.

[0007] Optionally, it also includes pushing out early warning information and the fire fighting and rescue plan in real time, and triggering alarm devices to issue warning signals.

[0008] Optionally, the information pushed in real time includes the latitude and longitude coordinates of the fire location, the fire risk level, the expected direction of spread, the optimal rescue route, and a list of equipment that needs to be prioritized for protection; the warning signal intensity of the audible and visual alarm device is adjusted according to the fire risk level, emitting a low-frequency audible and visual signal when the risk is medium and a high-frequency strong light signal when the risk is high.

[0009] Optionally, the knowledge graph includes an entity layer, a relationship layer, and an attribute layer; the entity layer includes photovoltaic modules, inverters, meteorological factors, terrain units, and fire events; the relationship layer defines the associations between entities, including the deployment of photovoltaic modules in terrain units and the influence of meteorological factors on fire events; the attribute layer records the characteristic parameters of each entity, including the rated temperature of photovoltaic modules and the slope value of terrain units.

[0010] Optionally, the knowledge graph construction process includes: collecting industry standards, historical accident reports and equipment manual data in the field of mountain photovoltaic fires; extracting core entities through entity recognition algorithms; mining the relationships between entities using relation extraction algorithms; constructing an initial graph through ontology modeling tools; and finally optimizing the graph structure by combining manual review and real-time data feedback to ensure the accuracy and completeness of the graph.

[0011] Optionally, during the training process of the fire early warning model, an attention mechanism is introduced to optimize the LSTM neural network, so that the model focuses on entity attribute data in the knowledge graph that are strongly associated with "fire events". The reasoning rules in the knowledge graph are used as constraints during the training of the fire early warning model to improve the model's adaptability to mountain photovoltaic fire scenarios. The threshold for determining the fire risk level is jointly calibrated by historical fire data and graph reasoning results.

[0012] Optionally, the decision support calls upon a geographic information system, combines entity relationships in a knowledge graph, integrates real-time fire data, environmental perception data, and terrain spatial data, performs path planning calculations and resource allocation analysis, and generates the optimal fire-fighting and rescue path, priority protection areas for photovoltaic equipment, and fire resource allocation scheme.

[0013] Optionally, the path planning operation uses A The algorithm, combining the relationships between road networks, terrain units, and photovoltaic module arrays in the knowledge graph, generates multiple candidate paths under constraints such as the distance from the fire rescue starting point to the fire occurrence point, the path slope, and the road capacity. The resource allocation analysis is based on the proximity and matching relationships between fire stations, fire extinguishing equipment, and photovoltaic module arrays in the knowledge graph, and allocates suitable fire-fighting resources based on the fire spread speed prediction results.

[0014] A data-driven fire early warning and decision support system, comprising: Disaster monitoring data acquisition module: used to build a distributed multi-source data acquisition system covering equipment, environment, terrain and images, and to collect equipment operation data, environmental perception data, terrain spatial data and video image data of mountain photovoltaic power stations in a targeted manner, so as to achieve comprehensive acquisition of disaster monitoring data; Fire domain knowledge graph construction module: used to construct a knowledge graph for the mountain photovoltaic fire domain. The knowledge graph includes core entities of photovoltaic equipment, meteorological factors, terrain units, fire events and the relationships between entities. Based on the knowledge graph, entity matching, attribute alignment and graph reasoning are performed on the collected multi-source data to achieve semantic-level fusion of multi-source heterogeneous data and generate a structured fusion dataset. Fire risk early warning module: It is used to build a fire early warning model based on a structured fusion dataset, mine fire risk characteristics in the data through the fire early warning model, and output four levels of fire risk: no risk, low risk, medium risk, and high risk, so as to achieve accurate monitoring and early warning of disaster risks. Fire Decision Support Module: When the fire risk level is medium or high, it triggers decision support, calls the geographic information system and combines the entity association relationship of the knowledge graph, integrates real-time fire situation and environmental and terrain data, and generates a fire fighting and rescue plan.

[0015] Optionally, it also includes a rescue push plan push module: used to push early warning information and the fire fighting and rescue plan in real time, and to link the alarm device to issue a warning signal.

[0016] As can be seen from the above technical solutions, compared with the prior art, the present invention provides a data-driven fire early warning and decision support method and system. It strengthens the data foundation through comprehensive multi-source data collection, achieves efficient semantic fusion of multi-source heterogeneous data with the help of domain knowledge graphs, realizes accurate and forward-looking management of fire risks with four-level graded early warning, and finally combines geographic information systems and knowledge graphs to generate fire fighting and rescue plans that fit actual combat, thereby comprehensively improving the scientific nature, accuracy and emergency response efficiency of fire prevention and control in mountain photovoltaic power stations. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the method flow provided by the present invention; Figure 2 This is a schematic diagram of the system structure provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention discloses a data-driven fire early warning and decision support method, such as... Figure 1 As shown, it includes the following steps: S1: Construct a distributed multi-source data acquisition system covering equipment, environment, terrain and images, and collect equipment operation data, environmental perception data, terrain spatial data and video image data of mountain photovoltaic power stations in a targeted manner to achieve comprehensive acquisition of disaster monitoring data; S2: Construct a knowledge graph for the field of mountain photovoltaic fires. The knowledge graph includes core entities of photovoltaic equipment, meteorological factors, terrain units, and fire events, as well as the relationships between entities. Based on the knowledge graph, perform entity matching, attribute alignment, and graph reasoning on the collected multi-source data to achieve semantic-level fusion of multi-source heterogeneous data and generate a structured fusion dataset. S3: Construct a fire early warning model based on a structured fusion dataset, mine the fire risk characteristics in the data through the fire early warning model, and output four levels of fire risk: no risk, low risk, medium risk, and high risk, so as to achieve accurate monitoring and early warning of disaster risks. S4: When the fire risk level is medium or high, trigger decision support, call the geographic information system and combine the entity association relationship of the knowledge graph to integrate real-time fire situation and environmental and terrain data to generate a fire fighting and rescue plan.

[0021] Furthermore, in S1, the multi-source data acquisition system acquires data through distributed acquisition nodes. The distributed acquisition nodes include wireless temperature sensors deployed on the backsheet of photovoltaic modules, current and voltage sensors installed in the inverter cabinet, a meteorological station installed at the highest point of the site, and a mobile monitoring terminal equipped with a GPS positioning module. Each acquisition node transmits data to the data acquisition gateway through 5G or LoRa wireless communication technology, and the acquisition gateway has a data storage function to avoid data loss caused by signal interruption in mountainous areas.

[0022] Furthermore, this embodiment also includes pushing the early warning information and the fire fighting and rescue plan in real time, and linking the alarm device to issue a warning signal.

[0023] The real-time information push includes the latitude and longitude coordinates of the fire location, the fire risk level, the expected direction of spread, the optimal rescue route, and a list of equipment that needs priority protection; the warning signal intensity of the audible and visual alarm device is adjusted according to the fire risk level, emitting a low-frequency audible and visual signal when the risk is medium and a high-frequency strong light signal when the risk is high.

[0024] Furthermore, in S2, the knowledge graph includes an entity layer, a relation layer, and an attribute layer; the entity layer includes photovoltaic modules, inverters, meteorological factors, terrain units, and fire events; the relation layer defines the relationships between entities, including the deployment of photovoltaic modules in terrain units and the influence of meteorological factors on fire events; the attribute layer records the characteristic parameters of each entity, including the rated temperature of photovoltaic modules and the slope value of terrain units.

[0025] The knowledge graph construction process includes: collecting industry standards, historical accident reports, and equipment manual data in the field of mountain photovoltaic fires; extracting core entities through entity recognition algorithms; mining the relationships between entities using relation extraction algorithms; constructing an initial graph using ontology modeling tools; and finally optimizing the graph structure by combining manual review and real-time data feedback to ensure the accuracy and completeness of the graph.

[0026] Furthermore, the data and knowledge graph fusion process includes: using an entity matching algorithm based on semantic similarity to match information such as "component number" and "temperature value" in the collected data with the attributes of the "photovoltaic module" entity in the graph; correcting data differences through attribute alignment rules (such as unit unification and format standardization); and using rule reasoning in the graph (such as "when the component temperature is >85℃ and the wind speed is <2m / s, the fire risk increases") to complete the missing risk-related data.

[0027] Furthermore, in S3, during the training process of the fire early warning model, an attention mechanism is introduced to optimize the LSTM neural network, so that the model focuses on entity attribute data in the knowledge graph that are strongly associated with "fire events". The reasoning rules in the knowledge graph are used as constraints during the training of the fire early warning model to improve the model's adaptability to mountain photovoltaic fire scenarios. The threshold for determining the fire risk level is jointly calibrated by historical fire data and graph reasoning results.

[0028] Furthermore, in S4, the decision support calls upon the geographic information system, combines entity relationships in the knowledge graph, integrates real-time fire data, environmental perception data, and terrain spatial data, performs path planning calculations and resource allocation analysis, and generates the optimal fire-fighting and rescue path, priority protection areas for photovoltaic equipment, and fire resource allocation scheme.

[0029] Furthermore, the path planning operation uses A The algorithm, combining the relationships between road networks, terrain units, and photovoltaic module arrays in the knowledge graph, generates multiple candidate paths under constraints such as the distance from the fire rescue starting point to the fire occurrence point, the path slope, and the road capacity. The resource allocation analysis is based on the proximity and matching relationships between fire stations, fire extinguishing equipment, and photovoltaic module arrays in the knowledge graph, and allocates suitable fire-fighting resources based on the fire spread speed prediction results.

[0030] This embodiment also includes iterative optimization of the knowledge graph and model: historical fire case data, early warning records and decision execution effect data of mountain photovoltaic power stations are periodically input into the knowledge graph, and the entity attributes and relationships of the graph are updated through incremental learning algorithms; at the same time, the updated structured fusion dataset is input into the fire early warning model and decision support module to optimize model parameters and decision rules, so as to continuously improve the accuracy of model early warning and the rationality of decision schemes.

[0031] and Figure 1 Corresponding to the method shown, the present invention also discloses a data-driven fire early warning and decision support system for... Figure 1 The implementation of the method, specifically its structure, is as follows: Figure 2 As shown, it includes: Disaster monitoring data acquisition module: used to build a distributed multi-source data acquisition system covering equipment, environment, terrain and images, and to collect equipment operation data, environmental perception data, terrain spatial data and video image data of mountain photovoltaic power stations in a targeted manner, so as to achieve comprehensive acquisition of disaster monitoring data; Fire domain knowledge graph construction module: used to construct a knowledge graph for the mountain photovoltaic fire domain. The knowledge graph includes core entities of photovoltaic equipment, meteorological factors, terrain units, fire events and the relationships between entities. Based on the knowledge graph, entity matching, attribute alignment and graph reasoning are performed on the collected multi-source data to achieve semantic-level fusion of multi-source heterogeneous data and generate a structured fusion dataset. Fire risk early warning module: It is used to build a fire early warning model based on a structured fusion dataset, mine fire risk characteristics in the data through the fire early warning model, and output four levels of fire risk: no risk, low risk, medium risk, and high risk, so as to achieve accurate monitoring and early warning of disaster risks. Fire Decision Support Module: When the fire risk level is medium or high, it triggers decision support, calls the geographic information system and combines the entity association relationship of the knowledge graph, integrates real-time fire situation and environmental and terrain data, and generates a fire fighting and rescue plan.

[0032] Furthermore, this embodiment also includes a rescue push plan push module: used to push early warning information and the fire fighting and rescue plan in real time, and to link the alarm device to issue a warning signal.

[0033] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0034] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data-driven fire early warning and decision support method, characterized in that, The method comprises the following steps: A distributed multi-source data collection system covering equipment, environment, terrain and images is constructed, equipment operation data, environment perception data, terrain spatial data and video image data of the mountain photovoltaic power station are collected, and all-around acquisition of disaster monitoring data is realized; A mountain photovoltaic fire field knowledge graph is constructed, the knowledge graph comprises core entities of photovoltaic equipment, meteorological factors, terrain units and fire events and the correlation between the entities, entity matching, attribute alignment and graph reasoning are performed on the collected multi-source data based on the knowledge graph, semantic-level fusion of the multi-source heterogeneous data is realized, and a structured fusion data set is generated; A fire warning model is constructed based on the structured fusion data set, fire risk features in the data are mined through the fire warning model, four levels of fire risk levels, i.e. no risk, low risk, medium risk and high risk, are output, and accurate monitoring and early warning of disaster risks are realized; When the fire risk level is medium or high, decision support is triggered, a geographic information system is called, the entity correlation of the knowledge graph is combined, real-time fire information and environment and terrain data are integrated, and a fire extinguishing and rescue plan is generated.

2. A data-driven fire warning and decision support method according to claim 1, characterized in that, The method further comprises real-time pushing of warning information and the fire extinguishing and rescue plan and linkage of an alarm device to send a warning signal.

3. A data-driven fire warning and decision support method according to claim 2, characterized in that, The information pushed in real time comprises longitude and latitude coordinates of a fire occurrence position, a fire risk level, an expected spreading direction, an optimal rescue path and a list of equipment that needs to be protected in priority; and the warning signal strength of the sound and light alarm device is adjusted according to the fire risk level, a low-frequency sound and light signal is sent when the risk is medium, and a high-frequency strong light signal is sent when the risk is high.

4. The data-driven fire warning and decision support method of claim 1, wherein, The knowledge graph comprises an entity layer, a relationship layer and an attribute layer; the entity layer comprises photovoltaic modules, inverters, meteorological factors, terrain units and fire events; the relationship layer defines the correlation between the entities, and the correlation comprises that the photovoltaic modules are arranged in the terrain units and the meteorological factors affect the fire events; and the attribute layer records characteristic parameters of the entities, and the characteristic parameters comprise a rated temperature of the photovoltaic modules and a slope value of the terrain units.

5. The data-driven fire warning and decision support method of claim 1, wherein, The construction process of the knowledge graph comprises the following steps: collecting industry standards, historical accident reports and equipment manual data in the field of mountain photovoltaic fires, extracting core entities through an entity recognition algorithm, mining the correlation between the entities by using a relationship extraction algorithm, constructing an initial graph through an ontology modeling tool, and finally optimizing the graph structure by combining manual review and real-time data feedback to ensure the accuracy and integrity of the graph.

6. The data-driven fire warning and decision support method of claim 1, wherein, In the training process of the fire warning model, an attention mechanism is introduced to optimize an LSTM neural network, so that the model focuses on entity attribute data that are strongly correlated with "fire events" in the knowledge graph, and the reasoning rules in the knowledge graph are used as constraint conditions during training of the fire warning model to improve the adaptability of the model to the mountain photovoltaic fire scene, and the determination threshold of the fire risk level is calibrated by combining historical fire data and graph reasoning results.

7. The data driven fire warning and decision support method of claim 1, wherein, The decision support calls a geographic information system, integrates real-time fire data, environmental perception data and terrain spatial data in combination with entity association relationships in the knowledge graph, performs path planning calculation and resource allocation analysis, and generates an optimal fire extinguishing and rescue path, a priority protected photovoltaic equipment area and a fire fighting resource allocation scheme.

8. A data-driven fire warning and decision support method according to claim 7, characterized in that, The path planning operation adopts A Algorithm, combined with the association relationship of road network, ground element and photovoltaic module array in the knowledge graph, to generate multiple candidate paths with the distance from the fire rescue starting point to the fire occurrence point, the path slope and the road traffic capacity as the constraint conditions; the resource allocation analysis is based on the proximity relationship and matching relationship of fire station, fire extinguishing equipment and photovoltaic module array in the knowledge graph, combined with the fire spread speed prediction result to distribute the adaptive fire resources.

9. A data driven fire warning and decision support system, characterized by It comprises: A disaster monitoring data acquisition module: used for constructing a distributed multi-source data collection system covering equipment, environment, terrain and images, collecting equipment operation data, environmental perception data, terrain spatial data and video image data of mountain photovoltaic power stations, and realizing all-around acquisition of disaster monitoring data; A fire field knowledge graph construction module: used for constructing a mountain photovoltaic fire field knowledge graph, the knowledge graph containing core entities and entity association relationships of photovoltaic equipment, meteorological factors, terrain units and fire events, performing entity matching, attribute alignment and graph reasoning on the collected multi-source data based on the knowledge graph, realizing semantic-level fusion of multi-source heterogeneous data, and generating a structured fusion data set; A fire risk early warning module: used for constructing a fire early warning model based on the structured fusion data set, mining fire risk features in the data through the fire early warning model, outputting four levels of fire risk levels of no risk, low risk, medium risk and high risk, and realizing accurate monitoring and early warning of disaster risks; A fire decision support module: used for triggering decision support when the fire risk level is medium or high, calling a geographic information system in combination with entity association relationships in the knowledge graph, integrating real-time fire and environmental and terrain data, and generating a fire extinguishing and rescue scheme.

10. A data-driven fire warning and decision support method according to claim 9, characterized in that, It also comprises a rescue push scheme push module: used for pushing early warning information and the fire extinguishing and rescue scheme in real time, and issuing warning signals through a warning device.

Images