Method and system for predicting fire spreading in chemical industry park

Abstract

The invention provides a chemical industrial park fire spreading prediction method and system. The method comprises the following steps: dividing a grid chart structure of a target chemical industrial park space structure into a plurality of function partition sub-graphs; determining a fire propagation flux of a shared node in an overlapping region between every two adjacent function zone sub-graphs, converting each function zone sub-graph into a macro-scale sub-graph, and determining a fire propagation macro-situation of the target chemical industrial park; according to the fire propagation flux and the fire propagation macroscopic situation, determining a fire front edge envelope line of each area, thereby constructing a fire propagation prediction model of the target chemical industry park, inputting fire initial environment data into the fire propagation prediction model, and outputting a fire propagation prediction result. And automatically triggering a fire alarm signal of the target park according to a fire spreading prediction result. By adopting the scheme of the application, the fire spreading of the target chemical industrial park can be predicted by combining the dynamic fire interaction among the areas in the target chemical industrial park.

Description

Technical Field

[0001] This application relates to the field of fire spread prediction technology, and more specifically, to a method and system for predicting the spread of fires in chemical industrial parks. Background Technology

[0002] Fire spread prediction is one of the core technologies in the field of fire prevention and emergency management. It uses scientific methods and technical means to predict the spread path, spread speed, scope of impact and intensity change of fire in specific environments (such as forests, buildings, urban areas, etc.), providing key basis for fire prevention, emergency decision-making and rescue command.

[0003] Chemical industrial parks, as high-risk industrial clusters, store and process large quantities of flammable and explosive hazardous chemicals. Fires in these parks are characterized by their suddenness, rapid spread, strong coupling, and wide destructive range. Traditional fire prediction methods based on single physical models or empirical criteria are insufficient to cope with their complex spatial structures, multi-material interactions, and dynamically changing environmental conditions. Existing technologies typically rely on macroscopic fluid simulations or simplified regional divisions, which suffer from high computational resource consumption, poor real-time performance, and insufficient description of the correlation between local fire behavior and the global situation. In particular, they struggle to achieve efficient and accurate simulations of dynamic fire propagation at multiple scales (from equipment level to park level). Furthermore, the fire interaction mechanisms between different areas within the park are complex, involving various fire transmission forms such as thermal radiation, flying sparks, and the diffusion of combustible vapor clouds. Traditional methods often ignore the influence of topological relationships and heterogeneity of spatial structures on fire evolution, leading to prediction results that deviate from the actual disaster situation and failing to provide a reliable basis for emergency evacuation and rescue decisions. Therefore, how to predict the fire spread of a target chemical industrial park by combining the dynamic fire interactions between different areas has become a problem facing the industry. Summary of the Invention

[0004] This application provides a method and system for predicting the spread of fires in chemical industrial parks, which can predict the spread of fires in the target chemical industrial park by combining the dynamic fire interactions between various areas in the target chemical industrial park.

[0005] Firstly, this application provides a method for predicting the spread of fires in a chemical industrial park, wherein the target chemical industrial park comprises multiple areas, and the method includes the following steps: A mesh diagram structure representing the spatial structure of the target chemical industrial park is constructed based on a three-dimensional geometric model of the target chemical industrial park. The grid structure is divided into multiple functional sub-graphs according to the fire dynamics characteristics of each region; By identifying the fire propagation characteristics between nodes within each functional sub-map, the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined. Each functional sub-map is then converted into a macro-scale sub-map under the macro-analysis scale of fire. Fire information is then transmitted between the macro-scale sub-maps, and the macro-spreading trend of the fire in the target chemical industrial park is determined. Based on all fire propagation fluxes and the macroscopic fire spread situation, the fire spread trend of each area in the target chemical industrial park is dynamically predicted when a fire occurs, and the fire front envelope of each area is obtained. A fire spread prediction model for the target chemical industrial park is constructed based on all fire front envelopes. Initial environmental data of the target chemical industrial park is input into the fire spread prediction model, and the fire spread prediction result for the target chemical industrial park is output. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

[0006] In some embodiments, dividing the grid structure into multiple functional sub-graphs according to the fire dynamics characteristics of each region specifically includes: The division window for the target chemical industrial park is determined based on the fire dynamics characteristics of each region; Determine the constraints when dividing the mesh graph structure; Based on the partitioning window and the constraints, the mesh graph structure is divided into multiple functional subgraphs.

[0007] In some embodiments, determining the fire propagation flux of shared nodes in overlapping areas between adjacent functional sub-maps based on fire propagation characteristics between nodes within each functional sub-map specifically includes: Determine the fire transmission characteristics between nodes within each functional sub-map; Extract shared nodes in the overlapping areas between adjacent functional sub-maps; Based on the fire transmission characteristics, determine the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps; Determine the ventilation characteristics of the target chemical industrial park; The fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined based on the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps and the ventilation characteristics.

[0008] In some embodiments, the transfer of fire information between various macro-scale sub-maps and the determination of the macro-level fire spread situation in the target chemical industrial park specifically include: Fire information is transmitted using shared nodes between various macro-scale subgraphs as information interaction interfaces, and fire energy exchange data is determined during the fire information transmission process. The fire spread status of each area in the target chemical industrial park is determined based on the fire energy exchange data. Obtain the time step for fire monitoring; The macro-level fire spread situation of the target chemical industrial park is determined based on all fire spread statuses and the time steps described.

[0009] In some embodiments, the fire spread trend of a fire occurring in each area of ​​the target chemical industrial park is dynamically predicted based on all fire propagation fluxes and the macroscopic fire spread situation, resulting in the fire front envelope of each area, specifically including: Select one area in the target chemical industrial park as the selected area, and extract the fire propagation flux corresponding to the selected area from all fire propagation fluxes. When a fire occurs in a selected area, the dynamic boundary of fire spread in the selected area is predicted based on all extracted fire propagation fluxes. The fire front envelope of the selected area is determined by the macroscopic fire spread situation and the dynamic boundary of fire spread. Continue to determine the fire front envelope for the remaining areas.

[0010] In some embodiments, constructing a fire spread prediction model for a target chemical industrial park based on all fire front envelopes specifically includes: Determine the boundary coordinates and corresponding fire intensity parameters of each fire front envelope; Graph neural networks were chosen as the basic framework. Based on the boundary coordinates of each fire front envelope and the corresponding fire intensity parameters, a fire spread prediction model for the target chemical industrial park is constructed using the aforementioned basic framework.

[0011] In some embodiments, if the fire spread prediction result indicates that the fire will spread to a preset safety limit within the danger warning time, automatically triggering the fire alarm signal of the target park further includes: If the fire spread prediction results show that the fire has not spread to the preset safety limit within the danger warning time, then the fire warning signal of the target park will be triggered.

[0012] In some embodiments, a three-dimensional geometric model of the target chemical industrial park is constructed using a three-dimensional modeling tool.

[0013] In some embodiments, the various areas include a production facility area, a utilities area, a loading and unloading and transportation area, an auxiliary production area, an environmental treatment area, and a fire and emergency area.

[0014] Secondly, this application provides a fire spread prediction system for chemical industrial parks, comprising: The module is used to construct a mesh structure representing the spatial structure of the target chemical industrial park based on the three-dimensional geometric model of the target chemical industrial park. The processing module is used to divide the grid structure into multiple functional sub-graphs according to the fire dynamics characteristics of each region; The processing module is also used to determine the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps by using the fire propagation characteristics between nodes within each functional sub-map, convert each functional sub-map into a macro-scale sub-map under the macro-analysis scale of fire, transmit fire information between each macro-scale sub-map, and determine the macro-proliferation trend of fire in the target chemical industrial park. The processing module is also used to dynamically predict the fire spread trend when a fire occurs in each area of ​​the target chemical industrial park based on all fire propagation fluxes and the macroscopic fire spread situation, and to obtain the fire front envelope of each area. The execution module is used to construct a fire spread prediction model for the target chemical industrial park based on all fire front envelopes, input the initial environmental data of the target chemical industrial park into the fire spread prediction model, and then output the fire spread prediction result of the target chemical industrial park. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The fire spread prediction method and system for chemical industrial parks provided in this application firstly constructs a grid structure representing the spatial structure of the target chemical industrial park based on a three-dimensional geometric model of the target chemical industrial park; the grid structure is divided into multiple functional sub-maps according to the fire dynamic characteristics of each region; the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined by the fire transmission characteristics between nodes within each functional sub-map, and each functional sub-map is converted into a macro-scale sub-map under the macro-analysis scale of fire; fire information is transmitted between each macro-scale sub-map, and the macro-fire spread situation of the target chemical industrial park is determined; the fire spread trend of each region in the target chemical industrial park is dynamically predicted based on all fire propagation fluxes and the macro-fire spread situation, and the fire front envelope of each region is obtained; a fire spread prediction model for the target chemical industrial park is constructed based on all fire front envelopes, the initial environmental data of the target chemical industrial park is input into the fire spread prediction model, and the fire spread prediction result of the target chemical industrial park is output; if the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

[0016] Therefore, in the fire spread prediction process, this application first constructs a grid structure representing the spatial structure based on a three-dimensional geometric model, providing a precise spatial digital foundation for capturing dynamic fire interactions between regions. This transforms the complex spatial layout of the park into quantifiable and analyzable basic data, laying a solid foundation for subsequent fire propagation research. Secondly, it divides the area into multiple functional sub-maps according to fire dynamics characteristics, achieving precise differentiation of fire characteristics in different regions and avoiding biases in fire interaction analysis caused by mixed regional characteristics, making the study of fire transmission characteristics within each zone more targeted. Thirdly, it determines the fire propagation flux of shared nodes and converts it into a macro-scale sub-map, accurately quantifying the intensity of dynamic fire interactions between adjacent regions. Simultaneously, through macro-scale conversion... Balancing comprehensiveness and efficiency in analysis, this approach provides crucial data for accurately assessing the overall macro-level fire spread situation. By dynamically predicting fire spread trends based on fire propagation flux and macro-level conditions, and obtaining the leading-edge envelope, it enables real-time tracking and boundary clarification of the spread process caused by dynamic fire interactions between regions. This provides a clear visual representation of the specific paths and extent of fire spread, offering dynamic data support for predictive model construction. The approach constructs a fire spread prediction model and outputs prediction results based on initial environmental data. It automatically triggers alarms when the fire is about to exceed safety limits, achieving full automation from dynamic fire interaction analysis to risk warning. This allows for rapid response to initial fire changes and early risk warnings, effectively improving the timeliness and accuracy of fire emergency response in the industrial park. Using this approach, the fire spread of a target chemical industrial park can be predicted by combining the dynamic fire interactions between different regions within the target park. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of a method for predicting the spread of fires in chemical industrial parks, as shown in some embodiments of this application. Figure 2 This is an exemplary flowchart illustrating the determination of the macroscopic situation of fire spread according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the determination of a fire spread prediction model according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a fire spread prediction system for chemical industrial parks, as shown in some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a method for predicting the spread of fires in a chemical industrial park, according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The figure is an exemplary flowchart of a method for predicting the spread of a fire in a chemical industrial park according to some embodiments of this application. The method for predicting the spread of a fire in a chemical industrial park mainly includes the following steps: The target chemical industrial park comprises several areas with varying functions and risk characteristics. These include the core production unit area (containing reactors, synthesis units, etc., which are high-risk areas), the centralized material storage area (divided into raw material storage tank areas, solid raw material warehouses, and finished product storage areas, with a high risk of fire spread), the public works area ensuring operation (containing substations, boiler rooms, etc., affecting fire protection capabilities), the loading and unloading and transportation area involving material transfer (including docks, loading and unloading stations, etc., with leakage and fire risks), the auxiliary production area containing small combustibles and hot work operations (such as laboratories and maintenance workshops), the densely populated office and living area (office buildings, dormitories, etc., requiring key protection), the environmental protection treatment area for handling pollutants (including sewage treatment plants, hazardous waste temporary storage rooms, etc., which are prone to generating toxic fumes), and the fire protection and emergency response area, which serves as the core of fire prevention and control (including fire pump rooms, fire stations, etc.). Each area is separated by roads, firewalls, etc., and needs to be analyzed separately in fire prediction.

[0020] In step 101, a mesh diagram structure representing the spatial structure of the target chemical industrial park is constructed based on the three-dimensional geometric model of the target chemical industrial park.

[0021] In specific implementation, the three-dimensional coordinates and attribute information (such as material and combustible type) of all physical entities within the target chemical industrial park, including buildings, equipment, pipelines, and roads, are acquired. This can be achieved through methods such as LiDAR scanning or UAV oblique photography. Based on the acquired information, a three-dimensional geometric model that closely matches the actual park is constructed, for example, using 3D modeling tools (such as Maya Modeling Software or Brand Modeling Software). On this basis, the three-dimensional geometric model is discretized into grid cells of different sizes (0.5-1 meter side length in the core area and 2-5 meters in non-critical areas). Each grid cell serves as a node in the graph structure, with node attributes including three-dimensional coordinates, grid size, combustible parameters, and information on surrounding obstacles. These attributes can be determined based on the importance of the area. Simultaneously, edges are established between adjacent grid cells based on spatial adjacency relationships (such as 6-neighbor or 18-neighbor connections), with the edge weight initialized to spatial distance. This constructs a grid graph structure that accurately represents the spatial structure of the target chemical industrial park. Other methods can also be used in other embodiments, which are not limited here.

[0022] It should be noted that the mesh graph structure in this application represents a digital abstraction of the three-dimensional space of a chemical industrial park. It uses a combination of "nodes" and "edges" to characterize the spatial features and relationships of the park. Each mesh cell, as a "node," not only contains spatial information such as three-dimensional coordinates and dimensions, but also integrates the physical attributes of the area (such as building materials, combustible material types and parameters, and obstacle distribution). The "edges" are used to connect adjacent nodes in space (such as mesh cells above, below, left, right, and front and back). The weight of the edges is initially set based on spatial distance and can be dynamically updated later based on fire propagation characteristics (such as heat transfer efficiency and the impact of obstacle shading). This graph structure not only preserves the spatial topology of the park, but also carries various physical parameters required for fire spread calculation, providing a structured data foundation for subsequent regional simulation, cross-regional information transmission, and fire situation prediction.

[0023] In step 102, the grid structure is divided into multiple functional sub-graphs according to the fire dynamics characteristics of each region.

[0024] In some embodiments, dividing the grid structure into multiple functional sub-graphs according to the fire dynamics characteristics of each region can be achieved using the following steps: The division window for the target chemical industrial park is determined based on the fire dynamics characteristics of each region; Determine the constraints when dividing the mesh graph structure; Based on the partitioning window and the constraints, the mesh graph structure is divided into multiple functional subgraphs.

[0025] In practice, the specific implementation method for defining the division windows of the target chemical industrial park based on the functions of each area is as follows: First, analyze the fire dynamics characteristics of each area in the chemical industrial park, classifying and labeling them according to production type (e.g., tank area, reaction unit area), fire risk level (high-risk flammable area, low-risk auxiliary area), facility density (dense unit area, open storage area), and process flow correlation (upstream and downstream unit clusters), and establish a function-risk correlation database; set the core parameters of the division windows, which can be set based on differences in characteristic attributes. For example, a small window size (e.g., 5m×5m) is used for high-risk and densely equipped areas (e.g., liquefied petroleum gas tank area) to improve local detail accuracy, while a large window size (e.g., 20m×20m) is used for low-risk open areas (e.g., administrative office area) to balance calculation efficiency. At the same time, set window edge alignment rules according to the functional area boundaries (e.g., walls, roads), load the 3D model of the park (which can be loaded from a geographic information system tool), and automatically generate the initial network according to the preset parameters. The grid window is adjusted by modifying its boundaries, which can be done manually to ensure that the window division conforms to the functional characteristics of each area while avoiding cross-critical functional units (such as dividing a single storage tank by a window). The final result is a window division that matches the function of the area, providing a spatial basis for subsequent functional sub-graph division. The window edge should cover a buffer zone of 2-3 grid units to ensure that adjacent windows have overlapping areas to retain shared nodes. Secondly, the constraints for dividing the grid graph structure are determined: a threshold for the number of sub-graph nodes is set (e.g., a single sub-graph contains 500-2000 grid nodes) to ensure balanced computational load; the shared node ratio requirement is clarified (the number of shared nodes in the overlapping area of ​​adjacent sub-graphs should not be less than 5% of the total number of nodes in the sub-graph) to ensure the accuracy of information transmission across sub-graphs; simultaneously, the consistency of physical properties within the sub-graph is constrained (e.g., the difference in combustible material type and ventilation conditions within the same sub-graph should be ≤ a preset threshold) to reduce computational complexity within the sub-graph. The above determinations are used as constraints for dividing the grid graph structure.

[0026] In addition, in specific implementation, the grid graph structure is divided into multiple functional sub-graphs based on the partitioning window and the constraints: the partitioning window slides or is fixedly covered on the grid graph structure, and the grid nodes within the partitioning window are clustered to ensure that each sub-graph meets the constraints of the number of nodes, the proportion of shared nodes, and the consistency of physical attributes; sub-graphs that exceed the constraints are further divided or merged and adjusted to finally form multiple functional sub-graphs that cover the entire park and have overlapping boundaries. The above steps can be implemented by a graph segmentation algorithm (as shown in the graph segmentation tool); other methods can also be used for partitioning in other embodiments, which are not limited here.

[0027] It should be noted that the functional partition subgraph in this application represents a sub-unit with the topological relationship and attribute characteristics of a local area of ​​the target chemical industrial park. Each functional partition subgraph contains an independent set of nodes, internal connection relationships, and shared node information with adjacent subgraphs.

[0028] In step 103, the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined by the fire propagation characteristics between nodes within each functional sub-map. Each functional sub-map is then converted into a macro-scale sub-map under the macro-analysis scale of fire. Fire information is transmitted between the macro-scale sub-maps, and the macro-spreading trend of the fire in the target chemical industrial park is determined.

[0029] In some embodiments, determining the fire propagation flux of shared nodes in overlapping areas between adjacent functional sub-maps based on fire propagation characteristics between nodes within each functional sub-map can be achieved using the following steps: Determine the fire transmission characteristics between nodes within each functional sub-map; Extract shared nodes in the overlapping areas between adjacent functional sub-maps; Based on the fire transmission characteristics, determine the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps; Determine the ventilation characteristics of the target chemical industrial park; The fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined based on the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps and the ventilation characteristics.

[0030] In specific implementation, the fire transmission characteristics between nodes within each functional sub-graph are determined: For each functional sub-graph, the heat conduction efficiency of adjacent nodes is calculated based on the physical properties of its internal nodes (such as combustible type, material, and temperature). This can be calculated using fire dynamics principles combined with Fourier's law. The heat radiation intensity of non-contact nodes is derived, for example, through the Stefan-Boltzmann law. The quantitative characteristics of flame spread resistance and heat transfer rate between nodes are determined, which can be determined by combining the combustion characteristics of combustibles (such as the fire spread rate in liquid pools and the ignition point of solid combustibles) and a modified Rosemer model. Finally, the fire transmission characteristics between nodes within the functional sub-graph are formed. The fire transmission characteristics are a set of quantitative attributes characterizing the propagation law of fire between different locations (grid nodes) in space, reflecting the intensity and efficiency of the physical processes of heat transfer, flame spread, and changes in combustion state. Other methods can also be used to determine these characteristics in other embodiments, which are not limited here.

[0031] Among them, the shared nodes in the overlapping area between adjacent functional sub-graphs are extracted: by comparing the node coordinates and index information of adjacent functional sub-graphs, the grid nodes that belong to both sub-graphs are selected and marked as shared nodes in the overlapping area between the corresponding adjacent functional sub-graphs. Shared nodes represent nodes that coexist in the overlapping area between adjacent functional sub-graphs.

[0032] In addition, in specific implementation, the internal transmission characteristics of shared nodes in the overlapping area between adjacent functional sub-graphs are determined based on the fire transmission characteristics: For each shared node, its fire transmission characteristics in its two sub-graphs (such as the thermal conductivity coefficient with surrounding nodes in sub-graph A and the thermal radiation reception intensity in sub-graph B) are extracted from the fire transmission characteristics, and integrated into the internal transmission characteristics of the node by weighted averaging (the weights are based on the connection density of the node in the two sub-graphs), thereby obtaining the internal transmission characteristics of shared nodes in the overlapping area between adjacent functional sub-graphs. The internal transmission characteristics represent the inherent quantitative attributes exhibited by the shared node when fire-related energy transmission (such as heat conduction, radiation, and flame spread) occurs between it and surrounding nodes in its respective sub-graphs. Other methods can also be used to determine this in other embodiments, which are not limited here.

[0033] In addition, in specific implementation, the ventilation characteristics of the target chemical industrial park are determined: wind speed, wind direction, and air pressure data are collected in real time through a meteorological sensor array deployed in the park, and the airflow field distribution in the park is simulated by a computational fluid dynamics model. The ventilation data is mapped to each grid node to obtain the ventilation parameters (such as wind speed vector and air convection coefficient) of the area where the shared node is located. The wind direction determines the fire spread direction correction coefficient, and the wind speed affects the heat diffusion rate. The above airflow field distribution and ventilation parameters are used as the ventilation characteristics of the target chemical industrial park. The ventilation characteristics refer to the comprehensive characterization of the air flow state and related environmental parameters in the target chemical industrial park.

[0034] In addition, in specific implementation, the fire propagation flux of the shared nodes in the overlapping areas between adjacent functional sub-graphs is determined based on the internal transmission characteristics of the shared nodes in the overlapping areas between adjacent functional sub-graphs, combined with the ventilation characteristics. Based on the internal transmission characteristics of the shared nodes, a ventilation correction coefficient is introduced from the ventilation characteristics (the higher the wind speed, the larger the correction coefficient; transmission is enhanced in the downwind direction and weakened in the upwind direction). It can be calculated by the formula "fire propagation flux = equivalent heat transfer coefficient × temperature difference × ventilation correction coefficient ÷ distance between nodes". The distance between nodes refers to the straight-line distance between the shared node and its associated node in the adjacent functional sub-graph. The greater the distance, the lower the energy flux per unit area. The fire propagation flux per unit time transmitted between adjacent functional sub-graphs through the shared nodes is obtained. Other methods can also be used to determine this in other embodiments, which are not limited here.

[0035] It should be noted that the fire propagation flux in this application represents the physical quantity of the intensity of fire energy (or combustion intensity) in the target chemical industrial park being transferred between adjacent subgraphs through shared nodes. It reflects the energy scale and efficiency of fire propagation from one subgraph to adjacent subgraphs and is a core quantitative indicator for assessing the risk of cross-regional fire spread.

[0036] In practice, converting each functional zone subgraph into a macro-scale subgraph under the fire macro-analysis scale can be achieved in the following way: Following a preset dimensionality reduction ratio (e.g., merging a 10×10 grid into one supernode), adjacent grid nodes within the functional zone subgraph are clustered and merged to form fewer "supernodes." This can be achieved using grid aggregation algorithms (e.g., regular grid merging, K-clustering) or graph simplification tools. The attributes of the supernodes are obtained by weighting the attributes of the original nodes they contain. For example, the average temperature is the weighted average of the original node temperatures (weights are based on the original node's combustible density or area percentage). The total combustible material quantity is the sum of the combustible material quantities of the original nodes. At the same time, the edge connection relationship between super nodes is redefined, and the edge weight is updated to the average or sum of the fire propagation flux between the original adjacent nodes. Redundant local connection details in the atomic graph are removed. Finally, the shared nodes between the subgraph and adjacent subgraphs are retained as the boundary nodes of the macro-scale subgraph. This can be achieved through spatial indexing technology to ensure that the interface for cross-subgraph information transmission is not lost, forming a macro-scale subgraph with a simplified structure but retained core features. Thus, a macro-scale subgraph under the macro-analysis scale of fire is obtained. Other conversion methods can also be used in other embodiments, which are not limited here.

[0037] It should be noted that the macro-scale subgraph in this application represents a simplified map structure at the macro-analysis scale of fire, formed after the original functional zoning subgraph has been reduced in dimension. This subgraph can accurately represent the macro-laws of regional fire propagation while significantly reducing computational complexity, providing an efficient structured data foundation for subsequent macro-situation analysis of the spread of fires at the park level.

[0038] In some embodiments, reference Figure 2 As shown, this figure is an exemplary flowchart for determining the macroscopic situation of fire spread in some embodiments of this application. In this embodiment, the transfer of fire information between various macroscopic scale sub-graphs and the determination of the macroscopic situation of fire spread in the target chemical industrial park can be achieved by the following steps: First, in step 1031, fire information is transmitted using the shared nodes between the various macro-scale subgraphs as the information interaction interface, and the fire energy exchange data in the fire information transmission process is determined. Secondly, in step 1032, the fire spread status of each area in the target chemical industrial park is determined based on the fire energy exchange data; Then, in step 1033, the time step for monitoring the fire is obtained; Finally, in step 1034, the macroscopic fire spread situation of the target chemical industrial park is determined based on all fire spread states and the time step.

[0039] In this method, fire information is transmitted through shared nodes between various macro-scale subgraphs, and the fire energy exchange data during the transmission of fire information is specifically implemented as follows: Shared nodes serve as information hubs for adjacent macro-scale subgraphs. These nodes receive fire state parameters (such as average temperature and heat release rate) and edge weights (propagation flux) from nodes within their respective subgraphs and synchronize this information to associated nodes in adjacent subgraphs. This can be achieved through the message passing mechanism of a graph neural network. Simultaneously, the energy transfer value (such as heat exchange) per unit time through the shared nodes is calculated, which can be done using a propagation flux formula. This forms cross-subgraph fire energy exchange data, where the fire energy exchange data represents the core data set of fire-related energy transmitted between adjacent macro-scale subgraphs through shared nodes. Other methods can be used to determine this in other embodiments, which are not limited here.

[0040] Furthermore, determining the fire spread status of each area in the target chemical industrial park based on the fire energy exchange data is specifically implemented as follows: For each functional area of ​​the park (such as the production unit area and the storage area), the corresponding macro-scale sub-map is analyzed. Combining the temperature rise rate of the nodes in the macro-scale sub-map, the expansion range of the combustion area (determined by the combustion status mark of the super node), and the energy exchange intensity transmitted between sub-maps through shared nodes, the fire intensity distribution heat map, the spread direction vector (determined by the dominant direction of energy exchange), and the high-risk boundary (formed by the connection of super nodes whose temperature exceeds the ignition point of the combustible material) of each area are obtained. The results of the analysis are used as the fire spread status of the corresponding area in the target chemical industrial park. The fire spread status represents the spread status of each functional area of ​​the target chemical industrial park during the fire development process.

[0041] Specifically, the time step for obtaining the fire monitoring time step is implemented as follows: based on the balance between the accuracy requirements and computing power of the fire response in the target chemical industrial park, a fixed time interval (such as 5 seconds, 10 seconds or 30 seconds) is preset as the fire monitoring time step, where the time step represents the fixed time interval set in the fire monitoring and simulation process.

[0042] Specifically, determining the macro-level fire spread situation of the target chemical industrial park based on all fire spread states and the time step is achieved as follows: Fire spread status data for each region is sequentially integrated according to the time step. By dynamically tracking the fire intensity distribution, spread direction changes, and high-risk area expansion at different times, the overall fire spread rate of the park (the increase in the area of ​​high-risk areas per unit time) and the predicted time of impact on key facilities (such as tank areas and control rooms) are calculated. This can be achieved through a spatial analysis engine and numerical calculation library. The resulting macro-level fire spread situation at the park level, covering the global spread trend, risk sequence of key areas, and the range of thermal radiation impact, can be output using visualization tools (such as heat map animations and vector arrow layers). Other methods can also be used in other embodiments, which are not limited here.

[0043] It should be noted that the macro-level trend of fire spread in this application refers to the overall fire development trend and impact range of the target chemical industrial park. It can reflect the overall spread pattern of fire at the park scale, and also clarify the expansion path of high-risk areas and the degree of threat to key protected targets, providing a global and dynamic panoramic reference for emergency decision-making regarding the fire development.

[0044] In step 104, the fire spread trend of each area in the target chemical industrial park when a fire occurs is dynamically predicted based on all fire propagation fluxes and the macroscopic fire spread situation, and the fire front envelope of each area is obtained.

[0045] In some embodiments, dynamically predicting the fire spread trend when a fire occurs in each area of ​​the target chemical industrial park based on all fire propagation fluxes and the macroscopic fire spread situation, and obtaining the fire front envelope of each area, can be achieved by the following steps: Select one area in the target chemical industrial park as the selected area, and extract the fire propagation flux corresponding to the selected area from all fire propagation fluxes. When a fire occurs in a selected area, the dynamic boundary of fire spread in the selected area is predicted based on all extracted fire propagation fluxes. The fire front envelope of the selected area is determined by the macroscopic fire spread situation and the dynamic boundary of fire spread. Continue to determine the fire front envelope for the remaining areas.

[0046] Specifically, the fire propagation flux corresponding to the selected area is extracted from all fire propagation fluxes: the selected area corresponds to multiple functional sub-maps, and the fire propagation flux corresponding to each functional sub-map is used as the fire propagation flux corresponding to the selected area.

[0047] Furthermore, when a fire occurs in a selected area, the dynamic boundary of fire spread in the selected area is predicted based on all extracted fire propagation fluxes. Specifically, this is achieved by taking the initial fire location in the selected area as the starting point, using the extracted fire propagation fluxes as the basis for energy transfer intensity, and simulating fire spread by combining the critical ignition point and combustion rate of the combustibles in the area. This can be simulated using an improved Huygens principle model. Within each time step, only the node connections where the propagation flux exceeds the critical value (such as the minimum energy threshold that can ignite adjacent nodes) are retained. The boundaries of newly ignited nodes are marked, forming a dynamic boundary profile that expands continuously over time. The coordinates and timestamps of the boundary nodes are recorded simultaneously, ultimately yielding the dynamic boundary of fire spread in the selected area. Here, the dynamic boundary represents the boundary of fire spread in the selected area that changes over time. Other prediction methods can also be used in other embodiments, which are not limited here.

[0048] Furthermore, determining the fire front envelope of a selected area based on the macroscopic fire spread situation and the dynamic boundary of fire spread is specifically implemented as follows: Global parameters from the macroscopic fire spread situation of the target chemical industrial park (such as the dominant spread direction and wind speed correction coefficient for the area) are used to trend-correct the dynamic boundary of fire spread (e.g., the downstream boundary expands faster according to the spread rate in the macroscopic situation, while the upstream boundary contracts according to the flux decay law). For example, when the dynamic boundary is downstream, it expands faster according to the formula "corrected expansion distance = original predicted distance × wind speed correction coefficient × proportion of macroscopic spread rate," while the upstream boundary expands faster according to the formula "corrected expansion distance = original predicted distance × (1 - flux decay coefficient) × wind speed correction coefficient ... The formula for "flux attenuation coefficient" is contracted, where the flux attenuation coefficient is calculated by a negative correlation model between headwind speed and propagation flux. By correcting the rule of acceleration with the wind and contraction with the wind, the dynamic boundary can reflect both the energy transfer characteristics between local nodes and the macroscopic evolution law of fire under the influence of environmental conditions and regional functional layout at the park scale. Ultimately, the obtained fire front envelope is closer to the actual fire spread scenario. The corrected dynamic boundaries of each time step are connected according to the time series, and the boundary curve is smoothed. This can be achieved by using an interpolation algorithm (such as cubic spline interpolation) to form a closed fire front envelope. Other methods can also be used to determine the fire front in other embodiments, which are not limited here.

[0049] It should be noted that the fire front envelope in this application represents the fire advance boundary formed by energy transfer between local nodes. The area inside the envelope is the already spread area, and the area outside is the unaffected area. It can be used to dynamically track the expansion trajectory, speed and shape changes of the fire front, providing an intuitive and reliable spatial boundary basis for accurately judging the fire spread range, assessing the degree of threat to the area and formulating emergency evacuation and fire fighting plans.

[0050] In step 105, a fire spread prediction model for the target chemical industrial park is constructed based on all fire front envelopes. Initial environmental data of the target chemical industrial park is input into the fire spread prediction model, and the fire spread prediction result for the target chemical industrial park is output. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

[0051] In some embodiments, reference Figure 3 As shown, this figure is an exemplary flowchart for determining a fire spread prediction model in some embodiments of this application. In this embodiment, the fire spread prediction model for the target chemical industrial park can be constructed based on all fire front envelopes using the following steps: First, in step 1051, the boundary coordinates of each fire front envelope and the corresponding fire intensity parameters are determined; Secondly, in step 1052, a graph neural network is selected as the basic framework; Finally, in step 1053, a fire spread prediction model for the target chemical industrial park is constructed based on the boundary coordinates of each fire front envelope and the corresponding fire intensity parameters, combined with the aforementioned basic framework.

[0052] In practical implementation, firstly, the boundary coordinates of each fire front envelope, the corresponding fire intensity parameters (such as average temperature and heat release rate), and related environmental variables (ventilation characteristics and combustible material distribution) are extracted to form a timestamped multi-dimensional training dataset. A graph neural network is selected as the basic framework, or a spatiotemporal sequence model can be chosen. The multi-dimensional training dataset and the boundary nodes of the fire front envelope are used as the model input features of the basic framework. The input layer is constructed by combining key variables such as fire propagation flux and macroscopic situation parameters, with the expansion trajectory and morphological changes of the fire front envelope at future moments as the output target. The basic framework is trained using historical fire simulation data, the multi-dimensional training dataset, and actual case data. For example, mini-batch gradient descent is used during the training phase. The method divides the data into training and validation sets proportionally. Based on a graph neural network framework, each iteration inputs a feature vector containing envelope boundaries, propagation flux, and environmental variables. The model's predicted future envelope output is calculated through forward propagation. Loss functions (such as the sum of squared errors of boundary coordinates) can be used to optimize model parameters, enabling the model to learn the evolution of the envelope over time, environmental factors, and energy transfer. Finally, the model's prediction accuracy is tested using a validation set, and the model structure is adjusted (e.g., by adding an attention mechanism to strengthen the weights of key areas). Ultimately, a fire spread prediction model is formed that can dynamically output the overall fire spread range and boundaries of the park over multiple time steps based on the current fire front envelope and real-time environmental data. Other construction methods can also be used in other embodiments, which are not limited here.

[0053] It should be noted that the fire spread prediction model in this application represents a model for dynamically predicting the spatial range and evolution trend of fire development in a target chemical industrial park. It can be used to calculate the boundary expansion trajectory, morphological changes, and high-risk area range of the overall fire spread in the park over multiple time steps in the future, providing scientific and accurate predictive support for fire emergency decision-making, risk assessment, and the formulation of prevention and control measures.

[0054] Furthermore, the specific implementation of inputting initial environmental data of the target chemical industrial park into the fire spread prediction model and then outputting the fire spread prediction results of the target chemical industrial park is as follows: First, the initial environmental data of the target chemical industrial park is standardized and preprocessed. The collected initial environmental data (such as the initial fire location coordinates, ventilation speed and direction at the time of ignition, combustible material types and distribution density in each area, initial temperature and heat release rate, etc.) is converted according to the feature dimension format during model training, missing values ​​are supplemented (such as obtaining local environmental parameters through interpolation of adjacent areas), and normalization is performed to ensure that the data format is consistent with the requirements of the model input layer; the preprocessed initial environmental data is compared with the currently monitored fire... The initial boundary data of the fire front envelope are integrated into an input feature vector, which is then input into a trained fire spread prediction model. The model performs time-series extrapolation on the input data through forward propagation calculations. It can use the learned fire evolution patterns to extrapolate and generate intermediate results such as the fire front envelope expansion trajectory, high-risk area boundaries, and fire intensity distribution heatmaps at preset time steps (e.g., 10 minutes or 30 minutes in the future). Finally, the intermediate results output by the model are post-processed to remove outliers and integrate them into a prediction report that includes the fire spread range at multiple future time points, the timing of threats to key facilities, and the areas affected by thermal radiation. This report can be integrated using visualization tools to complete the output of the fire spread prediction results.

[0055] It should be noted that the fire spread prediction results in this application represent the extent, speed, intensity and potential impact of the fire within the park over a future period of time. This provides a scientific and dynamic basis for emergency command departments to formulate evacuation plans, allocate firefighting resources and assess risk levels.

[0056] Furthermore, if the fire spread prediction result indicates that the fire will spread to the preset safety boundary within the danger warning time, the specific steps for automatically triggering the fire alarm signal of the target park are as follows: First, preset the safety boundary coordinate threshold (such as the protection boundary of critical facilities and the personnel evacuation safety line) and the danger warning time threshold for triggering the alarm (such as within the next 10 minutes) for each area of ​​the park in the system, and establish a real-time comparison mechanism between the fire spread prediction result and the safety boundary; after the model outputs the prediction result, the system automatically extracts the fire front envelope expansion trajectory and high-risk area boundary data for each time step, and determines the future risk through spatial coordinate calculation. The system records whether the fire boundary at each moment intersects with the preset safety limit, and also records the expected time point when the fire spreads to the safety limit. If a certain time point is detected to meet the triggering conditions of "expected spread time ≤ threshold of danger warning time" and "overlap between fire boundary and safety limit", the system immediately sends a trigger command to the park's fire alarm control system through the linkage interface, activates the corresponding area's audible and visual alarms, emergency broadcasts, and linkage equipment (such as automatic access control and smoke exhaust system activation), and simultaneously pushes the alarm triggering reason (including predicted time and area involving safety limit) to the emergency command platform, completing the automatic triggering of alarm signals and information synchronization.

[0057] If the fire spread prediction results show that the fire has not spread to the preset safety boundary within the danger warning time, a fire warning signal is triggered in the target park. First, a "fire controllable trend report" is automatically generated, which clarifies the current distance margin between the fire front envelope and the safety boundary, the expected extension of the danger warning time, and key influencing factors (such as headwind environment and the effectiveness of fire extinguishing measures), and is pushed to the emergency command platform. Based on the report, a graded response adjustment mechanism is activated to reduce the alarm level of non-essential areas (such as turning off some area audible and visual alarms, while maintaining the emergency broadcast standby status). At the same time, resource allocation is optimized, fire extinguishing forces redundantly deployed outside the safety boundary are suspended and redirected to the core fire area for enhanced suppression, and the fire situation is continuously monitored. Parameters such as wind direction and temperature are collected in real time through an environmental sensor network, and updated data is input into the fire spread prediction model for secondary simulation every half time step to verify the stable trend of the fire. If two consecutive simulations confirm that the safety boundary is not threatened, the system can trigger the "dynamic prevention and control downgrade" process to gradually restore the normal production order of unaffected areas of the park, while maintaining the core monitoring and emergency preparedness status until the fire is completely extinguished.

[0058] In another aspect, in some embodiments, this application provides a fire spread prediction system for chemical industrial parks, with reference to... Figure 4 The figure is a schematic diagram of the structure of a chemical industrial park fire spread prediction system 400 according to some embodiments of this application. The chemical industrial park fire spread prediction system 400 includes: a construction module 401, a processing module 402, and an execution module 403, which are described below: Construction module 401, in this application, is mainly used to construct a mesh diagram structure representing the spatial structure of the target chemical industrial park based on the three-dimensional geometric model of the target chemical industrial park; Processing module 402, in this application, is used to divide the grid structure into multiple functional sub-graphs according to the fire dynamics characteristics of each region; It should be noted that the processing module 402 in this application is also used to determine the fire propagation flux of shared nodes in the overlapping area between adjacent functional sub-maps by the fire propagation characteristics between nodes within each functional sub-map, convert each functional sub-map into a macro-scale sub-map under the fire macro-analysis scale, transmit fire information between each macro-scale sub-map, and determine the macro-scale fire spread situation of the target chemical industrial park. In addition, it should be noted that the processing module 402 in this application is also used to dynamically predict the fire spread trend when a fire occurs in each area of ​​the target chemical industrial park based on all fire propagation fluxes and the macroscopic fire spread situation, so as to obtain the fire front envelope of each area. The execution module 403 in this application is mainly used to construct a fire spread prediction model of the target chemical industrial park based on all fire front envelopes, input the initial environmental data of the target chemical industrial park into the fire spread prediction model, and then output the fire spread prediction result of the target chemical industrial park. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

[0059] In addition, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described method for predicting the spread of fires in chemical industrial parks.

[0060] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing a method for predicting the spread of fires in chemical industrial parks, according to some embodiments of this application. The method for predicting the spread of fires in chemical industrial parks described in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0061] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0062] The communication bus 502 can be used to transmit information between the aforementioned components.

[0063] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0064] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The method used in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0065] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0066] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0067] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0068] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting the spread of fires in chemical industrial parks.

[0069] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0070] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for predicting the spread of fires in chemical industrial parks, wherein, The target chemical industrial park comprises multiple areas, and the method is characterized by the following steps: A mesh diagram structure representing the spatial structure of the target chemical industrial park is constructed based on a three-dimensional geometric model of the target chemical industrial park. The grid structure is divided into multiple functional sub-graphs according to the fire dynamics characteristics of each region; By identifying the fire propagation characteristics between nodes within each functional sub-map, the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined. Each functional sub-map is then converted into a macro-scale sub-map under the macro-analysis scale of fire. Fire information is then transmitted between the macro-scale sub-maps, and the macro-spreading trend of the fire in the target chemical industrial park is determined. Based on all fire propagation fluxes and the macroscopic fire spread situation, the fire spread trend of each area in the target chemical industrial park is dynamically predicted when a fire occurs, and the fire front envelope of each area is obtained. A fire spread prediction model for the target chemical industrial park is constructed based on all fire front envelopes. Initial environmental data of the target chemical industrial park is input into the fire spread prediction model, and the fire spread prediction result for the target chemical industrial park is output. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

2. The method as described in claim 1, characterized in that, The grid structure is divided into multiple functional sub-maps based on the fire dynamics characteristics of each region, specifically including: The division window for the target chemical industrial park is determined based on the fire dynamics characteristics of each region; Determine the constraints when dividing the mesh graph structure; Based on the partitioning window and the constraints, the mesh graph structure is divided into multiple functional subgraphs.

3. The method as described in claim 1, characterized in that, The fire propagation flux of shared nodes in overlapping areas between adjacent functional sub-maps is determined by analyzing the fire propagation characteristics between nodes within each functional sub-map. Specifically, this includes: Determine the fire transmission characteristics between nodes within each functional sub-map; Extract shared nodes in the overlapping areas between adjacent functional sub-maps; Based on the fire transmission characteristics, determine the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps; Determine the ventilation characteristics of the target chemical industrial park; The fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps is determined based on the internal transmission characteristics of shared nodes in the overlapping areas between adjacent functional sub-maps and the ventilation characteristics.

4. The method as described in claim 1, characterized in that, The transmission of fire information between various macro-scale sub-maps and the determination of the macro-level fire spread situation in the target chemical industrial park specifically include: Fire information is transmitted using shared nodes between various macro-scale subgraphs as information interaction interfaces, and fire energy exchange data is determined during the fire information transmission process. The fire spread status of each area in the target chemical industrial park is determined based on the fire energy exchange data. Obtain the time step for fire monitoring; The macro-level fire spread situation of the target chemical industrial park is determined based on all fire spread statuses and the time steps described.

5. The method as described in claim 1, characterized in that, Based on all fire propagation fluxes and the macroscopic fire spread situation, the fire spread trend in each area of ​​the target chemical industrial park is dynamically predicted when a fire occurs, resulting in the fire front envelope of each area, specifically including: Select one area in the target chemical industrial park as the selected area, and extract the fire propagation flux corresponding to the selected area from all fire propagation fluxes. When a fire occurs in a selected area, the dynamic boundary of fire spread in the selected area is predicted based on all extracted fire propagation fluxes. The fire front envelope of the selected area is determined by the macroscopic fire spread situation and the dynamic boundary of fire spread. Continue to determine the fire front envelope for the remaining areas.

6. The method as described in claim 1, characterized in that, The fire spread prediction model for the target chemical industrial park is constructed based on all fire front envelopes, specifically including: Determine the boundary coordinates and corresponding fire intensity parameters of each fire front envelope; Graph neural networks were chosen as the basic framework. Based on the boundary coordinates of each fire front envelope and the corresponding fire intensity parameters, a fire spread prediction model for the target chemical industrial park is constructed using the aforementioned basic framework.

7. The method as described in claim 1, characterized in that, If the fire spread prediction results indicate that the fire will spread to the preset safety limit within the danger warning time, the automatic triggering of the fire alarm signal in the target park also includes: If the fire spread prediction results show that the fire has not spread to the preset safety limit within the danger warning time, then the fire warning signal of the target park will be triggered.

8. The method as described in claim 1, characterized in that, A three-dimensional geometric model of the target chemical industrial park is constructed using 3D modeling tools.

9. The method as described in claim 1, characterized in that, The various areas include the production unit area, utilities area, loading and unloading and transportation area, auxiliary production area, environmental protection treatment area, and fire protection and emergency response area.

10. A fire spread prediction system for chemical industrial parks, characterized in that, include: The module is used to construct a mesh structure representing the spatial structure of the target chemical industrial park based on the three-dimensional geometric model of the target chemical industrial park. The processing module is used to divide the grid structure into multiple functional sub-graphs according to the fire dynamics characteristics of each region; The processing module is also used to determine the fire propagation flux of shared nodes in the overlapping areas between adjacent functional sub-maps by using the fire propagation characteristics between nodes within each functional sub-map, convert each functional sub-map into a macro-scale sub-map under the macro-analysis scale of fire, transmit fire information between each macro-scale sub-map, and determine the macro-proliferation trend of fire in the target chemical industrial park. The processing module is also used to dynamically predict the fire spread trend when a fire occurs in each area of ​​the target chemical industrial park based on all fire propagation fluxes and the macroscopic fire spread situation, and to obtain the fire front envelope of each area. The execution module is used to construct a fire spread prediction model for the target chemical industrial park based on all fire front envelopes, input the initial environmental data of the target chemical industrial park into the fire spread prediction model, and then output the fire spread prediction result of the target chemical industrial park. If the fire spread prediction result shows that the fire will spread to the preset safety limit within the danger warning time, the fire alarm signal of the target industrial park is automatically triggered.

Images