A forest fire spread risk identification system based on artificial intelligence

Abstract

The present application relates to the technical field of risk identification, and particularly relates to a forest fire spreading risk identification system based on artificial intelligence, which comprises a hidden heat channel network construction module: penetrating thermal sensing data and materialization parameters of a target forest area are acquired, an area with high temperature but no combustion under the ground is deduced through an inversion model, and a three-dimensional hidden heat source network is constructed; a surface breakthrough risk calculation module: for each node in the three-dimensional hidden heat source network, the heat conduction and combustible gas penetration comprehensive action intensity on the surface combustible material are calculated, and potential breakthrough points exceeding an ignition threshold are identified; a discrete fire risk generation module: a heat energy redistribution process along the network is simulated, the ignition probability and ignition time of each potential breakthrough point are dynamically predicted, and a discrete fire risk map is generated. The present application effectively improves the early intervention accuracy and spatial deployment decision-making ability of forest fires, and is suitable for coping with forest undergrowth fire scenes with strong concealment and complex propagation path.

Description

Technical Field

[0001] This invention relates to the field of risk identification technology, and in particular to an artificial intelligence-based forest fire spread risk identification system. Background Technology

[0002] The current forest fire prevention technology system mainly relies on satellite remote sensing, infrared detection, and visible light video monitoring for early identification and alarm of forest fires. These methods primarily use abnormal surface temperature or smoke image characteristics as the main criteria, and are suitable for identifying fires at the open flame stage or when the fire point has broken through the surface. However, existing technologies are ineffective in identifying hidden heat sources that have not yet reached the ignition point, are underground, or are covered by vegetation. These hidden heat sources are very likely to break through the surface and form new fire points under windy and dry conditions, often becoming sudden triggers for the spread of forest fires and seriously threatening the safety of forest areas. Therefore, there is an urgent need for a new technology that can detect and identify potential underground fire sources in advance to fill the blind spots in early detection of traditional forest fire monitoring.

[0003] On the other hand, the formation and ignition path of underground heat sources are subject to a variety of complex factors, including soil thermal conductivity, fissure channel structure, combustible gas seepage, surface combustibles and their humidity, wind speed and direction, etc. These factors exhibit nonlinear coupling relationships in time and space, making it difficult to accurately model them with a single sensor or static model. Existing studies mostly focus on large-scale fire propagation simulation or fire point identification, and rarely involve modeling and early warning work on the complete chain from heat source network structure to heat propagation evolution to dynamic assessment of ignition risk. Summary of the Invention

[0004] This invention provides an artificial intelligence-based forest fire spread risk identification system that systematically depicts the formation process of a fire from an underground heat source to a surface fire point, thereby improving the precision and intelligence of forest fire prediction.

[0005] An artificial intelligence-based forest fire spread risk identification system includes the following modules: Concealed thermal channel network construction module: acquire penetrating thermal sensing data and physical parameters of the target forest area, deduce underground unburned but continuously high-temperature areas through inversion model, and connect spatially continuous thermal anomaly areas to construct a three-dimensional concealed heat source network composed of nodes and edges representing thermal connectivity; Surface breach risk calculation module: For each node in the three-dimensional concealed heat source network, calculate the combined effect of heat conduction and combustible gas penetration on surface combustibles, and identify potential breach points that exceed the ignition threshold; Discrete fire risk generation module: Based on the three-dimensional concealed heat source network structure and node thermal state, it simulates the redistribution process of heat energy along the network, dynamically predicts the ignition probability and ignition time of each potential breakthrough point, and generates a discrete fire risk map that characterizes the spatial location and hazard level of new fire points that may break out in the future.

[0006] Optionally, the concealed hot aisle network construction module specifically includes: Data acquisition unit: acquires penetrating thermal sensing data and physicochemical parameters of the target forest area from the multi-source sensing platform; wherein, the penetrating thermal sensing data includes electromagnetic radiation data that has the ability to penetrate vegetation and topsoil; Heat source inversion unit: The penetrating thermal sensing data and physicochemical parameters are input into an inversion model that couples physical conduction equations and data-driven models to deduce the unburned but persistently high-temperature thermal anomaly region and temperature distribution in the underground three-dimensional space.

[0007] Optionally, the concealed hot aisle network construction module further includes: Network construction unit: Based on the preset temperature connectivity threshold and spatial proximity rules, thermal anomaly regions with continuous temperature and physical properties in three-dimensional space are aggregated and connected. The thermal anomaly regions are used as nodes and the thermal connectivity between regions are used as edges to construct a three-dimensional hidden heat source network.

[0008] Optionally, the surface breach risk calculation module specifically includes: Risk intensity calculation unit: For each node in the three-dimensional hidden heat source network, the unit couples and calculates the intensity of heat conduction from the upper medium to the surface and the intensity of combustible gas generated by pyrolysis permeating upward along cracks or pores, and generates the comprehensive effect intensity value of the corresponding node through a weighted fusion model. Breakthrough point identification unit: compares the combined effect intensity value of each node with a dynamic ignition threshold; all nodes whose combined effect intensity value exceeds the corresponding dynamic ignition threshold are marked as potential breakthrough points.

[0009] Optionally, the risk intensity calculation unit specifically includes: For each node in the three-dimensional concealed heat source network, the heat conduction intensity transferred from the node to the surface through the upper medium and the permeation intensity of the combustible gas generated by the pyrolysis of the corresponding node permeating upward along the cracks or pores are calculated respectively. The thermal conductivity and the combustible gas permeability are weighted and fused to generate a comprehensive action intensity value for the corresponding node.

[0010] Optionally, the dynamic ignition threshold is determined in real time based on the type of surface combustibles directly above the node, ambient humidity, and airflow conditions.

[0011] Optionally, the breakthrough point identification unit specifically includes: Based on the type of combustible material on the ground directly above the node, the ambient humidity and airflow conditions, the dynamic ignition threshold corresponding to each node is determined in real time, and the dynamic ignition threshold is compared with the comprehensive effect intensity value of the corresponding node. Nodes whose combined effect intensity value exceeds the corresponding dynamic ignition threshold are identified and marked as potential breakthrough points.

[0012] Optionally, the discrete fire risk generation module specifically includes: Network heat redistribution simulation unit: The three-dimensional hidden heat source network is regarded as a dynamic graph. Based on the thermodynamic model on the dynamic graph, the redistribution process of heat energy along the network edge between nodes is iteratively simulated, and the network thermal state at future time points is updated. Discrete Risk Prediction Unit: For each potential breakthrough point, based on the simulated future thermal state time series and the surface environmental parameters directly above, a pre-trained ignition risk mapping model is used to dynamically calculate the ignition probability and estimated ignition time of the potential breakthrough point in multiple future time windows. Risk map synthesis unit: Integrates the spatial location, ignition probability, estimated ignition time, and hazard level of all potential breakthrough points to generate a discrete fire risk map that includes the time dimension.

[0013] Optionally, the node attributes of the dynamic graph include heat capacity and current temperature, and the edge attributes include thermal conductivity.

[0014] Optionally, the discrete risk prediction unit specifically includes: For each potential breakthrough point, the future thermal state time series obtained through network thermal redistribution simulation is acquired, and the corresponding surface environment parameters directly above the potential breakthrough point are also acquired. The future thermal state time series and the surface environmental parameters are input into a pre-trained ignition risk mapping model to dynamically calculate the ignition probability of potential breakthrough points in multiple future time windows and the corresponding estimated ignition time.

[0015] The beneficial effects of this invention are: This invention reconstructs persistently high-temperature underground areas that are difficult to capture with traditional remote sensing into a three-dimensional graph network by constructing a hidden thermal channel network. It introduces an inversion algorithm that couples a heat conduction model and a data-driven model, and integrates penetrating thermal sensing data with multi-source physical and chemical parameters to identify and connect unburned but ignition-risk thermal anomaly areas. This technology breaks through the limitation of existing fire sensing methods that can only observe visible hotspots on the ground, and has stronger precursor sensing capabilities and thermal propagation structure expression capabilities, providing a fundamental support for building an early warning system for latent fire points.

[0016] After identifying potential breakthrough points, this invention introduces a graph thermodynamics model to simulate the propagation process of underground thermal energy in the network over time, obtaining the future thermal state evolution trend of each node. Combined with environmental parameters such as surface combustible material type, humidity, and airflow, a pre-trained ignition risk model is used to dynamically calculate the ignition probability and estimated ignition time for multiple time periods. Finally, these are integrated into a discrete fire risk map with spatial location, time label, and risk level. This mechanism can achieve a complete closed loop from static structure identification to dynamic risk prediction, effectively improving the accuracy of early intervention and spatial deployment decision-making capabilities for forest fires. It is suitable for dealing with forest fires that are highly concealed and have complex propagation paths. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this 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 for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system modules according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating risk generation in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] like Figures 1-2 As shown, an artificial intelligence-based forest fire spread risk identification system includes the following modules: Concealed thermal channel network construction module: acquire penetrating thermal sensing data and physical parameters of the target forest area, deduce underground unburned but continuously high-temperature areas through inversion model, and connect spatially continuous thermal anomaly areas to construct a three-dimensional concealed heat source network composed of nodes and edges representing thermal connectivity; The concealed hot aisle network construction module specifically includes: Data acquisition unit: Acquires penetrating thermal sensing data and physicochemical parameters of the target forest area from the multi-source sensing platform.

[0021] Among them, penetrating thermal sensing data is denoted as This includes electromagnetic radiation data that can penetrate vegetation and topsoil, possessing a certain ability to respond to deep thermal anomalies; microwaves can penetrate topsoil and respond to shallow thermal anomalies, while infrared provides high-resolution thermal information from shallower layers; the combination of both enhances the ability to identify concealed heat sources; the set of soil and land cover physicochemical parameters is denoted as... Including soil thermal conductivity Thermal diffusivity Moisture content Values ​​of land feature attributes as a function of spatial distribution; in, The thermal conductivity ranges from 0.2 to 3.0 W / m·K, and the significant differences in thermal conductivity among different soil types affect the accuracy of heat conduction simulations. for It characterizes the ability of heat to diffuse in soil and controls the response rate to temperature changes. The value ranges from 0.05 to 0.45. Moisture content affects soil thermal conductivity and heat capacity, which in turn affects heat transfer capacity and the accuracy of inversion results. Heat source inversion unit: and As input, a coupled inversion model is constructed to deduce the three-dimensional distribution and temperature state of the underground thermal anomaly, represented as: ; in, For the temperature distribution of the underground thermal field, The density of the soil varies significantly among different soil types, affecting the calculation of heat capacity. Specific heat capacity is a core parameter in the energy conservation equation; different land features respond differently to thermal changes. It is a time- and space-variable heat source term, characterizing the localized sustained heat release capacity. It is a core indicator for identifying hidden fire sources. Combined with a data-driven model for dynamic estimation, it is expressed as: ; It is based on the data-driven inversion function Conv-LSTM trained by a deep neural network, which improves the prediction ability of unsteady heat sources and fits nonlinear thermal perturbation processes. The parameter set is as follows: It is used to predict local latent heat accumulation behavior from observational data; The inversion result is a set of thermal anomaly regions in three-dimensional space that satisfy the following conditions, represented as: ; in, The threshold for detecting underground high-temperature anomalies is 60-80. Areas exceeding this temperature and remaining stable thereafter can be considered potential unburned open flame points or underground heat sources. The core of the heat source inversion unit is to integrate a physical model of heat conduction with a machine learning model to achieve high-precision, spatially continuous extrapolation of hidden underground heat sources. The entire model takes penetrating thermal sensing data and ground object physical parameters as input and outputs a three-dimensional underground temperature field and its anomalous areas.

[0022] In terms of thermophysics, the evolution trend of temperature with time and space is described by the heat conduction control equation, which can reflect the physical mechanism of soil heat diffusion. In terms of data-driven approach, the nonlinear behavior of local heat sources is learned from observation data by the Conv-LSTM neural network, which enhances the ability to identify hidden heat sources, especially suitable for scenarios with complex morphology and unsteady changes of anomalous sources.

[0023] Network building blocks: sets of thermal anomaly regions obtained through inversion. The construction of the three-dimensional heat source network is based on the following two rules, specifically including: (1) Spatial proximity rule: If the Euclidean distance between the center points of two thermal anomaly regions satisfies If they are adjacent units, a potential connection edge is established. (2) Temperature connectivity threshold: If the average temperature gradient between two regions satisfies If so, then it is considered that there is hot connectivity, and an edge connection can be established; The final constructed three-dimensional hidden heat source network is represented as follows: ; in, A three-dimensional hidden heat source network, It is a node The three-dimensional coordinates, in meters, are used for spatial distance determination and connectivity analysis. The distance threshold for spatial proximity determination has a range of 10-30m. This is the temperature gradient threshold, with a value ranging from 1 to 5°C / m. It is the average temperature of adjacent heat source nodes, used to calculate the temperature gradient, and the node set. Corresponding to the thermal anomaly region, edge set This indicates the thermal connectivity between nodes.

[0024] The network construction unit treats underground thermal anomaly regions as several spatial units with physical properties. By judging their spatial proximity and temperature connectivity, a connection relationship is established. Each thermal anomaly region is defined as a node, and edges are established between node pairs that have the conditions for heat transfer. Finally, a three-dimensional hidden heat source network composed of nodes and edges is formed. This network can be used to simulate the spread trend of heat underground and identify paths and key points that may lead to surface breaches.

[0025] During heat conduction and combustible gas diffusion, the closer the regions are, the higher the likelihood of energy exchange. Therefore, this unit sets a maximum spatial distance threshold. Only when the distance between the center points of two thermal anomaly regions is less than this maximum spatial distance threshold is a connection possible. This rule can exclude long-distance regions that do not have a physical propagation path, ensuring the physical rationality of the network structure.

[0026] Surface breach risk calculation module: For each node in the three-dimensional hidden heat source network, calculate the combined effect of heat conduction and combustible gas penetration on surface combustibles, and identify potential breach points that exceed the ignition threshold; The surface breach risk calculation module specifically includes: Risk intensity calculation unit: for each node in the three-dimensional hidden heat source network. The intensity of heat conduction and the intensity of combustible gas penetration through the upper medium to the surface are calculated and then weighted to form a comprehensive action intensity. .

[0027] This step addresses the question of how concealed underground heat sources can penetrate the surface and ignite new fire points. It breaks down the risk into two physical processes: the upward movement of heat and the upward movement of combustible gases, and performs a coupled assessment at the nodal scale. By separately calculating the heat conduction intensity and gas permeability intensity, and then weighted and fused, a more realistic characterization of the comprehensive ignition capability of underground heat sources on surface combustibles can be achieved.

[0028] (1) Calculation of thermal conductivity: thermal conductivity By node The vertical heat flux density estimated towards the Earth's surface directly above it is expressed as: ; in, For nodes The intensity of heat conduction to the Earth's surface The average thermal conductivity of the medium above the node, expressed in W / m·K, ranges from 0.2 to 3.0 W / m·K, covering dry soil, wet soil, humus layer, and rock-fissured media, accurately reflecting the differences in heat conduction in the underground media of forest land. For nodes Temperature, in units The value range is 60-150. , The temperature of the Earth's surface directly above the node, in units of... The value range is 10-60. This covers conditions ranging from normal surface temperature to high-temperature solar radiation, serving as a reasonable upper boundary for thermal gradient calculations. The vertical distance from the node to the ground surface is measured in meters (m) and ranges from 0.3 to 5 meters. It reflects the main depth of shallow underground heat sources affecting the ground surface, beyond which heat attenuation becomes significant.

[0029] The direct threat of high-temperature underground regions to the surface is primarily reflected in whether heat can be continuously and effectively transferred to the surface. This step adopts a simplified one-dimensional steady-state heat conduction model because, in shallow underground layers and on a short time scale, the dominant direction of heat propagation is usually vertically upward. The ratio of temperature difference to burial depth can intuitively reflect the magnitude of the thermal driving force. At the same time, the thermal conductivity of the medium is introduced to reflect the thermal attenuation characteristics under different soil or rock strata conditions.

[0030] (2) Calculation of combustible gas permeability: Combustible gas permeability Estimated based on the gas seepage formula in fractures or pores, expressed as: ; in, It represents the penetration intensity of combustible gases from nodes into the Earth's surface, reflecting the contribution of pyrolysis gas production to the risk of surface combustion. It emphasizes the trend rather than the absolute amount. Porosity or fracture openness of the seepage path, in dimensionless units, ranging from 0.05 to 0.6, covering dense soil to fractured rock and soil masses, determines whether gas has an upward flow path. This is the diffusion coefficient of combustible gas, with units of m² / s, and a range of values ​​of [value missing]. This represents the effective diffusion capacity of gases such as methane and carbon monoxide in soil and fissures. This represents the concentration of combustible gas at the node, expressed in mol / m³, ranging from 0.1 to 10 mol / m³, covering gas release levels from mild to severe pyrolysis. This represents the background gas concentration in the atmosphere, expressed in mol / m³. It is negligible compared to underground pyrolysis gases and is used as a concentration gradient reference. To be consistent with heat conduction, the value is the length of the upward permeation path, in meters, ranging from 0.3 to 5 meters, to ensure physical consistency with the heat conduction path. Combustible gases produced by underground pyrolysis accumulate upwards in fissures or pores, which is a significant contributing factor to sudden surface fires, especially in forest areas with thick humus layers and loose structures. Therefore, this step introduces a separate gas permeability index to assess the potential for gas to reach and participate in combustion through pore conditions, diffusion capacity, and concentration gradients, rather than simply assuming uniform gas diffusion.

[0031] (3) Integration of comprehensive action intensity: comprehensive action intensity Representation Nodes The overall activating effect on surface combustibles is expressed as: ; in, To determine the intensity of the combined effects of nodal interactions, a unified characterization of the coupling effect of heat and gas on surface ignition is established. The heat conduction weighting factor ranges from 0.4 to 0.7. In most forest areas, heat flux remains the dominant ignition factor, so its weighting is slightly higher. This is a weighting factor for gas permeability, ranging from 0.3 to 0.6. In areas with well-developed fractures or thick humus layers, the effect of gas permeability cannot be ignored, satisfying the following conditions: Its value can be determined by fitting environmental models or historical data.

[0032] Heat and combustible gases do not act in isolation, but rather are coupled and jointly determine whether a surface breach occurs. This step generates a comprehensive effect intensity through a weighted fusion method. On the one hand, this avoids a single factor from overly dominating the judgment, and on the other hand, it provides room for flexible adjustment of the model under different forest areas and seasonal conditions. This comprehensive index can be directly used for comparison with the dynamic ignition threshold to form a clear and interpretable risk criterion.

[0033] Breakthrough Point Identification Unit: The identification process is based on the combined effect strength of each node. The dynamic ignition threshold corresponding to its location For comparison, it can be represented as: ; ; in, It is a set of potential surface breach points. For nodes The corresponding ignition threshold is dynamically generated based on the surface properties directly above it, with a value ranging from 0 to 1. It represents the node's overall ability to ignite surface ignition points and is derived from the weighted superposition of two factors: heat conduction and gas permeability. These are ignition parameters for surface combustibles, ranging from 0.2 to 0.6, reflecting the ignition energy and dryness of different surface vegetation / organic matter types. The values ​​are derived from combustible material classification databases or remote sensing classifications. This is a dimensionless surface moisture correction factor; the wetter the surface, the higher the flame retardancy. The value ranges from 0.1 to 0.8; the wetter the surface, the harder it is to ignite. It is estimated using real-time meteorological or soil moisture information. This is an airflow condition correction factor that considers the promoting or inhibiting effect of wind speed and direction on ignition. Its value ranges from -0.2 to 0.5, characterizing the effect of wind. A positive value indicates strong winds that promote combustion, while a negative or zero value indicates weak winds or reverse airflow. This is an empirical calibration coefficient, which depends on the combustion model and ranges from 0.5 to 1.5. It controls the overall threshold level and, combined with historical fire point calibration, can adjust the system sensitivity.

[0034] Discrete fire risk generation module: Based on the three-dimensional hidden heat source network structure and node thermal state, it simulates the redistribution process of heat energy along the network, dynamically predicts the ignition probability and ignition time of each potential breakthrough point, and generates a discrete fire risk map that characterizes the spatial location and hazard level of new fire points that may break out in the future. The discrete fire risk generation module specifically includes: Networked heat redistribution simulation unit: a three-dimensional hidden heat source network It can be viewed as a thermodynamic diagram structure; Each node represents a heat source unit, and its attribute is the current temperature. With heat capacity Each edge represents the thermal connectivity between nodes, and its property is thermal conductivity. ; Based on the thermodynamic model uploaded from the dynamic graph, the heat energy redistribution satisfies the following discrete iterative formula: ; in, For nodes The set of adjacent nodes determines the basic graph structure of the heat conduction path. For nodes and The thermal conductivity between nodes, ranging from 0.1 to 2.0 W / (m·K), characterizes the ability to conduct heat between nodes and is set according to the properties of the medium. For nodes The heat capacity is used to control the temperature rise response rate, with a value ranging from 100 to 10000 J / K. The time step, ranging from 60 to 300 seconds, controls the simulation's timing accuracy and efficiency, matching the sampling period of the measured data. For nodes At any moment The temperature range is 50-150°C. .

[0035] By iterating the above formula over multiple time steps, we can update the future time steps. The predicted thermal state time series.

[0036] Discrete risk prediction unit: targeting all potential breakthroughs identified in the preceding steps. Using it to predict thermal state sequences The above-ground surface environmental parameters are input into a pre-trained ignition risk mapping model. The following output is obtained, which is represented as: ; in, For the future hot state time series of node 𝑖, In the first The probability of ignition within a future time window. This refers to the estimated ignition time for the corresponding time window. Ignition parameters for the type of surface combustibles. Humidity correction factor This is the airflow correction factor. The trained nonlinear mapping function is used to map thermal state and environmental factors into fire ignition probability and time, supporting LSTM and Transformer models.

[0037] Risk map synthesis unit: Integrates information from all potential breach points to generate a discrete fire risk map with spatial location and temporal dimensions, represented as: ; in, For nodes Projected coordinates on a two-dimensional Earth surface For ignition probability and ignition time series. For nodes In the The probability of ignition within a future time window, ranging from 0 to 1, represents the intensity of fire risk. For nodes In the The estimated ignition time within a future time window. The danger levels are categorized into 1-5. The resulting discrete fire risk map, used by the dispatch and monitoring platform and compatible with GIS systems, is a dynamic spatial layer with temporal attributes and probability labels. Each potential breach point is represented as an independent risk unit with dynamic attributes in this map, facilitating precise early warning and response deployment by regulatory authorities in a spatiotemporal dimension.

[0038] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0039] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A forest fire spread risk identification system based on artificial intelligence, characterized in that, Includes the following modules: Concealed thermal channel network construction module: acquire penetrating thermal sensing data and physical parameters of the target forest area, deduce underground unburned but continuously high-temperature areas through inversion model, and connect spatially continuous thermal anomaly areas to construct a three-dimensional concealed heat source network composed of nodes and edges representing thermal connectivity; Surface breach risk calculation module: For each node in the three-dimensional concealed heat source network, calculate the combined effect of heat conduction and combustible gas penetration on surface combustibles, and identify potential breach points that exceed the ignition threshold; Discrete fire risk generation module: Based on the three-dimensional concealed heat source network structure and node thermal state, it simulates the redistribution process of heat energy along the network, dynamically predicts the ignition probability and ignition time of each potential breakthrough point, and generates a discrete fire risk map that characterizes the spatial location and hazard level of new fire points that may break out in the future.

2. The forest fire spread risk identification system based on artificial intelligence according to claim 1, characterized in that, The concealed hot aisle network construction module specifically includes: Data acquisition unit: acquires penetrating thermal sensing data and physicochemical parameters of the target forest area from the multi-source sensing platform; wherein, the penetrating thermal sensing data includes electromagnetic radiation data that has the ability to penetrate vegetation and topsoil; Heat source inversion unit: The penetrating thermal sensing data and physicochemical parameters are input into an inversion model that couples physical conduction equations and data-driven models to deduce the unburned but persistently high-temperature thermal anomaly region and temperature distribution in the underground three-dimensional space.

3. The forest fire spread risk identification system based on artificial intelligence according to claim 1, characterized in that, The concealed hot aisle network construction module also includes: Network construction unit: Based on the preset temperature connectivity threshold and spatial proximity rules, thermal anomaly regions with continuous temperature and physical properties in three-dimensional space are aggregated and connected. The thermal anomaly regions are used as nodes and the thermal connectivity between regions are used as edges to construct a three-dimensional hidden heat source network.

4. The forest fire spread risk identification system based on artificial intelligence according to claim 1, characterized in that, The surface breach risk calculation module specifically includes: Risk intensity calculation unit: For each node in the three-dimensional hidden heat source network, the unit couples and calculates the intensity of heat conduction from the upper medium to the surface and the intensity of combustible gas generated by pyrolysis permeating upward along cracks or pores, and generates the comprehensive effect intensity value of the corresponding node through a weighted fusion model. Breakthrough point identification unit: compares the combined effect intensity value of each node with a dynamic ignition threshold; all nodes whose combined effect intensity value exceeds the corresponding dynamic ignition threshold are marked as potential breakthrough points.

5. The forest fire spread risk identification system based on artificial intelligence according to claim 4, characterized in that, The risk intensity calculation unit specifically includes: For each node in the three-dimensional concealed heat source network, the heat conduction intensity transferred from the node to the surface through the upper medium and the permeation intensity of the combustible gas generated by the pyrolysis of the corresponding node permeating upward along the cracks or pores are calculated respectively. The thermal conductivity and the combustible gas permeability are weighted and fused to generate a comprehensive action intensity value for the corresponding node.

6. The forest fire spread risk identification system based on artificial intelligence according to claim 4, characterized in that, The dynamic ignition threshold is determined in real time based on the type of surface combustibles directly above the node, ambient humidity, and airflow conditions.

7. The forest fire spread risk identification system based on artificial intelligence according to claim 4, characterized in that, The breakthrough point identification unit specifically includes: Based on the type of combustible material on the ground directly above the node, the ambient humidity and airflow conditions, the dynamic ignition threshold corresponding to each node is determined in real time, and the dynamic ignition threshold is compared with the comprehensive effect intensity value of the corresponding node. Nodes whose combined effect intensity value exceeds the corresponding dynamic ignition threshold are identified and marked as potential breakthrough points.

8. The forest fire spread risk identification system based on artificial intelligence according to claim 1, characterized in that, The discrete fire risk generation module specifically includes: Network heat redistribution simulation unit: The three-dimensional hidden heat source network is regarded as a dynamic graph. Based on the thermodynamic model on the dynamic graph, the redistribution process of heat energy along the network edge between nodes is iteratively simulated, and the network thermal state at future time points is updated. Discrete Risk Prediction Unit: For each potential breakthrough point, based on the simulated future thermal state time series and the surface environmental parameters directly above, a pre-trained ignition risk mapping model is used to dynamically calculate the ignition probability and estimated ignition time of the potential breakthrough point in multiple future time windows. Risk map synthesis unit: Integrates the spatial location, ignition probability, estimated ignition time, and hazard level of all potential breakthrough points to generate a discrete fire risk map that includes the time dimension.

9. A forest fire spread risk identification system based on artificial intelligence according to claim 8, characterized in that, The node attributes of the dynamic graph include heat capacity and current temperature, and the edge attributes include thermal conductivity.

10. A forest fire spread risk identification system based on artificial intelligence according to claim 8, characterized in that, The discrete risk prediction unit specifically includes: For each potential breakthrough point, the future thermal state time series obtained through network thermal redistribution simulation is acquired, and the corresponding surface environment parameters directly above the potential breakthrough point are also acquired. The future thermal state time series and the surface environmental parameters are input into a pre-trained ignition risk mapping model to dynamically calculate the ignition probability of potential breakthrough points in multiple future time windows and the corresponding estimated ignition time.

Images