Forest steppe fire situation real-time correction and simulation prediction method

Abstract

The invention provides a forest steppe fire situation real-time correction and simulation prediction method, which comprises the steps that a three-dimensional grid map of a fire scene is constructed, and three-dimensional fire scene data comprises grid terrain data and grid combustible data of each grid; the method comprises the following steps: acquiring meteorological data and a live wire range in real time, and executing a prediction process: correcting the meteorological data through an interpolation algorithm or a CFD model, correcting the live wire range through a historical prediction live wire, and inputting the corrected live wire range, the corrected meteorological data and three-dimensional fire scene data into a live wire prediction model, and when new meteorological data and the live wire range are obtained next time, the predicted live wire range is not output, and the prediction process is executed again. According to the application, correction is carried out every time the new meteorological data and the live wire range are obtained, so that the obtained live wire range is closer to the real evolution process of the live wire range when the live wire prediction model is applied to carry out live wire range prediction in a plurality of time periods.

Description

Technical Field

[0001] This specification relates to the field of fire simulation technology, and more specifically, this application relates to a method for real-time correction and simulation prediction of forest and grassland fire situations. Background Art

[0002] Forest fires are sudden, destructive, difficult to control, and challenging to extinguish. Because fire spread is influenced by factors like the environment and wind speed, it can be challenging for rescue workers to predict fire spread, control firefighting, and plan evacuations. Therefore, developing efficient and accurate fire situation prediction and simulation technologies is crucial for fire prevention, control, and emergency response. Summary of the Invention

[0003] The Summary of the Invention introduces a series of simplified concepts that will be further described in the Detailed Description of the Invention. The Summary of the Invention of this application is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0004] This application proposes a real-time correction and simulation prediction method for forest and grassland fire situations, including: Constructing a three-dimensional grid map of the target fire scene, wherein the three-dimensional grid map of the target fire scene includes three-dimensional fire scene data of the target fire scene, and the three-dimensional fire scene data includes grid terrain data and grid combustible material data of each grid in the three-dimensional grid map; Real-time acquisition of regional meteorological data and fire front range of the target fire scene; Each time the regional meteorological data and the fire line range of the target fire scene are obtained, a preset prediction process is executed to update the fire line range of the target fire scene in real time for multiple time periods in the future; The prediction process is as follows: Apply a preset interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data of the current time period to obtain the current meteorological data of each grid; input the target fire line range, the current meteorological data and the three-dimensional fire scene data into a pre-trained fire line prediction model to obtain the predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model; when the regional meteorological data and fire line range of the target fire scene are obtained next time, re-execute the prediction process; wherein, the target fire line range is the current fire line range of the current time period or the fire line range after the current fire line range is corrected based on the historical predicted fire line range, the historical predicted fire line range is the fire line range of the current time period obtained by the prediction process executed last time, and the fire line prediction model is a combination model of the CNN model and the LSTM model; The training process of the fire prediction model is as follows: Obtain historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges in multiple historical time periods, and historical meteorological data; input the historical three-dimensional fire scene data of the historical fire scenes, the historical fire line ranges and historical meteorological data of each historical time period into the fire line prediction model, obtain the prediction results of each historical time period output by the fire line prediction model, calculate the loss function between the prediction results of each historical time period and the historical fire line range, and adjust the model parameters of the fire line prediction model based on the loss function until the loss function currently calculated converges, thereby obtaining the trained fire line prediction model.

[0005] Optionally, in the above method, constructing a three-dimensional grid map of the target fire scene includes: Using a preset data acquisition device to collect regional terrain data and regional combustible material data of the target fire scene; Based on the regional terrain data and regional combustible material data, construct a three-dimensional BIM model corresponding to the target fire scene; The three-dimensional BIM model is divided into multiple grids to obtain a three-dimensional grid map of the target fire scene.

[0006] In the above method, optionally, dividing the three-dimensional BIM model into a plurality of grids includes: Dividing the three-dimensional BIM model into a plurality of grids according to a preset grid size; or, Based on the regional terrain data and the regional combustible material data, each terrain type and / or combustible material type in the three-dimensional BIM model is determined; and according to each terrain type and / or each combustible material type, the three-dimensional BIM model is divided into multiple grids.

[0007] Optionally, the method of applying a preset interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data for the current time period includes: determining a terrain type for each of the grid cells; When the terrain type of the grid is flat land, the interpolation algorithm is applied to calculate the regional meteorological data, the grid terrain data, and the grid combustible data for the current time period to obtain the current meteorological data of the grid for the current time period; When the terrain type of the grid is not flat land, the regional meteorological data, the grid terrain data and the grid combustible data of the current time period are input into the CFD model to obtain the current meteorological data of the grid in the current time period output by the CFD model.

[0008] In the above method, optionally, obtaining the fire line range of the target fire scene includes: Applying a preset image recognition method to identify the initial fire line range of the target fire scene; The initial fire line range is processed using a preset edge algorithm to obtain the fire line range of the target fire scene.

[0009] Optionally, the method includes inputting the target fire line range, the current meteorological data, and the three-dimensional fire scene data into a pre-trained fire line prediction model to obtain predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model, including: Inputting the target fire line range, the current meteorological data, and the three-dimensional fire scene data into a CNN model of the fire line prediction model, and obtaining fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics output by the CNN model; The LSTM model is applied to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics and combustible material characteristics to obtain the fire spread velocity vector and the fire spread direction vector corresponding to the target fire line range; based on the fire spread direction vector and the fire spread velocity vector, the predicted fire line range corresponding to each of the time periods is predicted.

[0010] The above method optionally includes applying the LSTM model to calculate the fire range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread velocity vector and fire spread direction vector corresponding to the target fire range, including: Divide a plurality of target ranges based on the fire line range characteristics, and input the flame intensity, heat transfer rate, crown moisture content and first branch height of each target range into the LSTM model; Calculating the surface moisture content within the target range based on the terrain characteristics and meteorological characteristics of the target grid, wherein the target grid is a grid that intersects with the target range; Calculating the flame intensity, the heat transfer rate, the surface moisture content, and the combustible material characteristics of the target grid within the target range to obtain a surface fire spread rate; determining whether tree crowns within the target range will be ignited based on the tree crown moisture content and the first branch height; If it will not be ignited, determining the surface fire spread speed as the fire spread speed vector of the target range; If it will be ignited, calculating the terrain characteristics, meteorological characteristics and combustible characteristics of the target grid, as well as the heat transfer rate, the flame intensity and the crown moisture content, to obtain the crown fire spread speed; A fire spread velocity vector corresponding to the target range is obtained based on the surface fire spread velocity and the crown fire spread velocity.

[0011] The above method optionally includes applying the LSTM model to calculate the fire range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread direction vector corresponding to the target fire range, including: Based on the terrain characteristics, meteorological characteristics and combustible material characteristics of the target grid, obtaining a three-dimensional slope vector, a three-dimensional wind direction vector and a combustible material position in the target grid corresponding to the target grid; constructing an initial direction vector corresponding to the target range based on the three-dimensional slope vector and the three-dimensional wind direction vector; Based on the position of the combustible material, the initial direction vector is adjusted to obtain a fire spread direction vector corresponding to the target range.

[0012] Optionally, in the above method, predicting the predicted fire line range corresponding to each time period based on the fire spread direction vector and the fire spread speed vector includes: Based on the fire spread direction vector and the fire spread speed vector corresponding to the target range, drawing an elliptical circular wave corresponding to each time period of the target range on the three-dimensional grid map; Based on the elliptical circular wave corresponding to each target range in each time period, a predicted fire line range corresponding to the target fire scene in each time period is generated.

[0013] Optionally, in the above method, the correction of the current fire line range based on the historically predicted fire line range includes: Obtaining an observed fire line range in a current time period, wherein the observed fire line range is the fire line range in the current time period obtained by correcting the historical predicted fire line range before obtaining the current fire line range; Get the Kalman gain coefficient for the current time period; determining a difference parameter between the observed fire line range and the current fire line range; The target fire line range is obtained based on the Kalman gain coefficient, the difference parameter and the historical predicted fire line range.

[0014] The present application provides a storage medium, which includes stored instructions, wherein when the instructions are executed, the device where the storage medium is located is controlled to execute the above-mentioned forest and grassland fire situation real-time correction and simulation prediction method.

[0015] The present application provides an electronic device comprising a memory and one or more instructions, wherein the one or more instructions are stored in the memory and configured to be executed by one or more processors to perform the above-mentioned real-time correction and simulation prediction method for forest and grassland fire situations.

[0016] In summary, the present application proposes a method for real-time correction and simulation prediction of forest and grassland fire situations, including: constructing a three-dimensional grid map of the target fire scene, the three-dimensional grid map contains three-dimensional fire scene data of the target fire scene, and the three-dimensional fire scene data contains grid terrain data and grid combustible data of each grid; obtaining regional meteorological data and fire line range of the target fire scene in real time; executing a prediction process each time the regional meteorological data and fire line range of the target fire scene are obtained: applying an interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data of the current time period to obtain the current meteorological data of each grid; and The fire line prediction model inputs the target fire line range, current meteorological data, and three-dimensional fire scene data to obtain the predicted fire line range corresponding to multiple time periods output by the fire line prediction model; the prediction process is re-executed when the regional meteorological data and fire line range of the target fire scene are obtained next time; wherein, the target fire line range is the current fire line range of the current time period or the fire line range after the current fire line range is corrected based on the historical predicted fire line range, the historical predicted fire line range is the fire line range of the current time period obtained by the last executed prediction process, and the fire line prediction model is a combination model of the CNN model and the LSTM model. Using the method provided in this application, high-precision simulation and real-time correction of the fire situation can be achieved through the steps of constructing a three-dimensional grid map, refining meteorological data, denoising and correcting the fire line range, simulating the spread of three-dimensional fire, and real-time correction and iterative calculation.

[0017] The method proposed in this application, and other advantages, objectives and features of this application will be reflected in part through the following description, and in part will be understood by those skilled in the art through research and practice of this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present description. The same reference symbols are used throughout the drawings to represent the same components. In the drawings: Figure 1 A flowchart of a method for real-time correction and simulation prediction of forest and grassland fire situations provided in an embodiment of the present application; Figure 2 A flow chart of a method for obtaining grid meteorological data provided in an embodiment of the present application; Figure 3Flowchart of the method for correcting the range of the line of fire provided in the embodiment of the present application Figure 4 A flow chart of a method for calculating a fire spread velocity vector within a target range provided in an embodiment of the present application; Figure 5 A schematic diagram of the current firing range provided in an embodiment of the present application; Figure 6 A flow chart of a method for calculating a fire spread direction vector within a target range provided in an embodiment of the present application; Figure 7 A schematic diagram of the change process of the predicted firing range provided in an embodiment of the present application; Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the embodiments described are only some embodiments of the present application, not all embodiments.

[0020] See also Figure 1 , which is a flow chart of a method for real-time correction and simulation prediction of forest and grassland fire situations provided in an embodiment of the present application, which may specifically include: S1: Construct a 3D grid map of the target fire scene.

[0021] The three-dimensional grid map includes three-dimensional fire scene data of the target fire scene, and the three-dimensional fire scene data includes grid terrain data and grid combustible material data of each grid in the three-dimensional grid map.

[0022] Raster terrain data includes height, slope and aspect, and raster fuel data includes vegetation type, fuel moisture, fuel density, fuel location and fuel load.

[0023] In one embodiment of the present application, the target fire scene is divided into a three-dimensional grid map including multiple grids. The three-dimensional grid map can be visualized to three-dimensionally display the changes in terrain and the distribution of combustibles in the target fire scene area.

[0024] The process of constructing a 3D grid map of the target fire area is as follows: Use preset data acquisition equipment to collect regional terrain data and regional combustible material data of the target fire scene; based on the regional terrain data and regional combustible material data, construct a three-dimensional BIM model corresponding to the target fire scene; divide the three-dimensional BIM model into multiple grids to obtain a three-dimensional grid map of the target fire scene.

[0025] Real-time regional terrain data and regional combustible material data within the area are acquired through data acquisition equipment (e.g., remote sensing satellites, drones, etc.). After obtaining the regional terrain data and regional combustible material data, the acquired data can also be preprocessed, including data format conversion, coordinate unification, and data cleaning. The regional terrain data and regional combustible material data are input into BIM (Building Information Modeling) modeling software, which then constructs a three-dimensional BIM model of the target fire area based on the input data. Furthermore, during the construction of the corresponding three-dimensional BIM model of the target fire area, images of the various regions of the target fire area can be input into the BIM modeling software. Combining the regional images, regional terrain data, and regional combustible material data, a three-dimensional BIM model of the target fire area can be constructed.

[0026] After obtaining the 3D BIM model, the model is divided into multiple grids according to the preset grid size or the terrain type and combustible type of the target fire scene. After the grids are divided, the grid terrain data and grid combustible data of each grid are determined based on the regional terrain data and regional combustible data of the target fire scene to obtain the 3D fire scene data of each grid.

[0027] In one possible implementation, the method for dividing the three-dimensional BIM model into a plurality of grids may be: dividing the three-dimensional BIM model into a plurality of grids according to a preset grid size. After dividing the three-dimensional BIM model into a plurality of grids, each grid in the three-dimensional grid map obtained by dividing the three-dimensional BIM model into the plurality of grids has a uniform grid size.

[0028] In another possible implementation, the method of dividing the three-dimensional BIM model into multiple grids may be: based on regional terrain data and regional combustible material data, determining each terrain type and / or combustible material type in the three-dimensional BIM model; according to each terrain type and / or each combustible material type, dividing the three-dimensional BIM model into multiple grids, and the size and shape of each grid are not necessarily consistent.

[0029] It is understandable that, due to the influence of different climate, geographical or human factors, there may be different terrain types (such as mountains, basins, slopes and flat lands) and different combustible types (such as moss, grassland, shrubs and trees) within the fire area. Different terrains or combustibles may have different corresponding burning rates. Therefore, when dividing the three-dimensional BIM model into multiple grids, the division can be based on terrain type and / or combustible type. By dividing the grids according to different terrain types and / or combustible types, when calculating the fire spread speed or fire spread direction for a single grid, only the influence of a single terrain and / or a single combustible needs to be considered, thereby accelerating the efficiency of the simulation and prediction of forest and grassland fire spread.

[0030] S2: Obtain regional meteorological data and fire line range of the target fire site in real time.

[0031] In an optional embodiment, regional meteorological data and fire line range of the target fire scene can be obtained according to a preset time period.

[0032] During a fire's spread, wind direction and speed often fluctuate dramatically due to the terrain and distribution of combustibles within the area. Furthermore, sudden meteorological changes are a significant uncontrollable factor after a fire occurs. When a fire source is present in the target fire area, it is difficult to obtain meteorological data for each grid within the target fire area. Therefore, meteorological data from multiple sample points within the target fire area can be collected through weather stations or satellites, and statistically calculated to obtain regional meteorological data for the entire target fire area. Alternatively, meteorological data from only the center of the target fire area can be collected as regional meteorological data for the entire target fire area.

[0033] The fire line range of the target fire scene is the regional boundary of the target fire scene. The fire line range can be the regional boundary of the target fire scene area obtained based on the target fire scene area identified on the three-dimensional raster map, or it can be the regional image of the target fire scene area collected by a drone or remote sensing satellite, and the regional boundary of the target fire scene area in the regional image is identified by an image recognition method.

[0034] The current fire line range of the target fire area is determined by: identifying the initial fire line range of the target fire area; applying a preset edge algorithm to process the initial fire line range to obtain the fire line range of the target fire area.

[0035] The edge algorithm is a Sobel algorithm, a Laplacian algorithm, or a Canny algorithm. The edge algorithm can include noise reduction processing. For example, the Canny algorithm can be applied to the initial fire line range to reduce noise, calculate the gradient of the noise-reduced fire line range, perform non-maximum suppression and threshold screening, and thus extract a clearer fire line range.

[0036] Images of most fire areas may contain areas covered in thick smoke, which can cause the initial fire boundary to be too blurry during image recognition. Therefore, an edge algorithm is used to remove noise from the image to extract a clearer fire boundary.

[0037] Optionally, after applying the edge algorithm to process the initial fire line range, the fire line range can be further refined. For example, through denoising and correction, noise interference in the fire line range can be eliminated, and the clarity and continuity of the fire line boundary can be improved.

[0038] In the embodiment of the present application, after each acquisition of regional meteorological data and fire scene range, the prediction process corresponding to S3 to S10 described below is executed.

[0039] S3: Apply an interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data of the current time period to obtain the current meteorological data of each grid.

[0040] Due to differences in terrain and vegetation, grid meteorological data within each grid may vary. Grid meteorological data includes wind speed, wind direction, temperature, and humidity within the grid. To obtain the current meteorological data for each grid, interpolation algorithms or algorithms corresponding to CFD models are required to calculate and correct the meteorological data for each grid based on the regional meteorological data due to the influence of terrain.

[0041] Due to the change of time, the wind field, temperature, humidity, etc. around the target fire site will change. Therefore, while obtaining the regional meteorological data of the target fire site in real time, it is also necessary to process the regional meteorological data obtained each time to correct the current meteorological data of the grid in different time periods.

[0042] refer to Figure 2 The specific steps for correcting the regional meteorological data for the current time period using the interpolation algorithm or CFD model are as follows: S301: Determine the terrain type of each grid.

[0043] It is understandable that if, in the process of constructing a three-dimensional grid map, the grid is divided according to terrain type as mentioned in the above embodiment, there is only one terrain type; if the grid is not divided according to terrain type, there is at least one terrain type.

[0044] S302: When the terrain type of the grid is flat land, an interpolation algorithm is applied to calculate the regional meteorological data, grid terrain data, and grid combustible data of the current time period to obtain the current meteorological data of the grid in the current time period.

[0045] For each grid in the three-dimensional grid map, if the terrain type within the grid is only flat land, an interpolation algorithm can be applied to calculate grid meteorological data. Specifically, the application of the interpolation algorithm to calculate grid meteorological data is to calculate the wind speed and wind direction within each grid.

[0046] S303: When the terrain type of the grid is not flat, input regional meteorological data, grid terrain data and grid combustible data of the current time period into the CFD model to obtain the current meteorological data of the grid in the current time period output by the CFD model.

[0047] Specifically, regional meteorological data, grid terrain data, and grid combustible material data are input into the CFD model. Based on this data, the CFD model analyzes the wind field characteristics of the grid and simulates the three-dimensional distribution of the wind field within the grid to obtain the wind speed and direction within the grid. Alternatively, regional meteorological data, grid terrain data, and grid combustible material data are input into the CFD model. Based on this data, the CFD model identifies changes in wind fields under different terrains and outputs meteorological data such as wind speed, direction, temperature, and humidity within each grid.

[0048] After executing step S3, data needs to be input into the fire line prediction model to predict the predicted fire line range in multiple time periods in the future. However, the data input into the fire line prediction model includes the target fire line range of the target fire scene, and the target fire line range is obtained through the current fire line range. If the current fire line range is the fire line range obtained for the first time, then the target fire line range is the fire line range obtained for the first time; if the current fire line range is not the fire line range obtained for the first time, then the historical predicted fire line range of the current time period predicted by the last executed fire line prediction model is obtained, and the fire line range obtained in the current time period is corrected using the historical predicted fire line range to obtain the target fire line range.

[0049] In the real-time correction and simulation prediction method for forest and grassland fire situations provided in the embodiment of the present application, during the prediction process, it is also necessary to determine the target fire line range to be input into the fire line prediction model. Therefore, after executing the above S3 and before inputting data into the fire line prediction model, the present application may further perform the following steps: S4: Determine whether the current firing line range is the first-acquired firing line range.

[0050] If yes, execute S5; otherwise, execute S6.

[0051] S5: Determine that the current firing line range is the target firing line range, and continue to execute S7.

[0052] S6: Obtain the historical predicted fire line range of the current time period obtained by the last execution of the prediction model, and correct the current fire line range based on the historical fire line range to obtain the target fire line range, and continue to execute S7.

[0053] The fire front range obtained for the current time period, after processing using image recognition and edge algorithms, can largely eliminate the influence of various factors (such as smoke, clouds, or shrubs) on the target fire. However, the obtained fire front range may still differ from the actual fire front range. If the fire front range obtained for the current time period is not the first time, the current fire front range can be corrected based on the historical fire front range predicted for the current time period obtained in the previous prediction process.

[0054] Specifically, refer to Figure 3 , based on the historical predicted fire line range, the current fire line range is corrected as follows: S601: Obtain the observed fire line range of the current time period.

[0055] It should be noted that after obtaining the historical predicted fire line range in the previous time period, before waiting for the current time period and obtaining the current fire line range of the current time period, the historical predicted fire line range is pre-corrected to obtain the observed fire line range. The specific correction formula is: Where H is the observation matrix, The historical forecast fireline range for the current time period predicted for the previous time period.

[0056] S602: Obtain the Kalman gain coefficient of the current time period.

[0057] The calculation formula of Kalman gain coefficient is: in, is the prediction error variance of the current time period, H is the observation matrix, S is the observation error variance of the current time period, , is the covariance matrix of the observation noise.

[0058] S603: Determine a difference parameter between the historically predicted fire line range and the current fire line range.

[0059] The difference parameter between the historical predicted fire line range and the current fire line range is: in, is the current firing range, For historical predictions of the firing range.

[0060] S604: Correct the current fire line range based on the Kalman gain coefficient and the difference parameter to obtain the target fire line range.

[0061] The target line of fire is calculated as follows: In this application, through Kalman filtering, the predicted fire line range and the observed fire line range can be used to correct the current fire line range actually obtained in the current time period, so that the fire line prediction model and actual data can be dynamically combined for processing, reducing the impact of noise on the fire line range and improving the accuracy of the fire line range.

[0062] S7: Input the target fire line range, current meteorological data, and three-dimensional fire scene data into the pre-trained fire line prediction model to obtain the predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model.

[0063] The fire front prediction model is a combination of a CNN (convolutional neural network) and an LSTM (long short-term memory) model. The CNN extracts data features, specifically the spatial characteristics of the target fire front, current meteorological data, and three-dimensional fire scene data. The LSTM processes time series data, capturing the temporal dependencies and dynamic changes of a fire's spread.

[0064] Specifically, in addition to inputting the target fire line range, current meteorological data and three-dimensional fire scene data into the fire line prediction model, data such as flame intensity, heat transfer rate, crown moisture content and height under the first branch in the target fire line range can also be input into the fire line prediction model. The CNN model and LSTM model in the fire line prediction model process the input data to obtain the predicted fire line ranges corresponding to multiple future time periods output by the fire line prediction model.

[0065] Therefore, the method of predicting the fire line range corresponding to multiple time periods through the CNN model and LSTM model is as follows: The target fire line range, current meteorological data, and three-dimensional fire scene data are input into the CNN model of the fire line prediction model to obtain the fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics output by the CNN model; the LSTM model is used to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread velocity vector and fire spread direction vector corresponding to the target fire line range; based on the fire spread direction vector and fire spread velocity vector, the predicted fire line range corresponding to each time period is predicted.

[0066] The LSTM model integrates the calculation formulas corresponding to the fire spread velocity vector and the fire spread direction vector. The LSTM model is used to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics. The process of obtaining the fire spread velocity vector and fire spread direction vector corresponding to the target fire line range is actually: applying the calculation formula corresponding to the fire spread velocity vector in the LSTM model to calculate the fire spread velocity vector corresponding to the target fire line range, and applying the calculation formula corresponding to the fire spread direction vector in the LSTM model to calculate the fire spread direction vector corresponding to the target fire line range.

[0067] Specifically, refer to Figure 4 The specific process of applying the LSTM model to calculate the fire range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread velocity vector corresponding to the target fire range is as follows: S701: Divide multiple target ranges based on the fire line range characteristics, and input the flame intensity, heat transfer rate, crown moisture content and first branch height of each target range into the LSTM model.

[0068] The target range falls within the current fire front range and is divided into multiple target ranges based on its characteristics. An infrared imager can be used to detect flame intensity within the target range, while a heat flux meter can be used to detect the heat transfer rate. Pre-set detection equipment can be used to measure the crown moisture content of trees within the target range. The first branch height is the height of the branch closest to the ground in the target range.

[0069] refer to Figure 5 ,exist Figure 5 In [1], the target fire scene is the area enclosed by curves ABCD, where curves ABCD are the current fire front range, and curves AB, BC, CD, and DA are all target ranges within the current fire front range. Each target range intersects with at least one grid.

[0070] Optionally, the size of the target range can be determined based on the terrain type. For example, within the current fire front, if there is a continuous stretch of flat terrain, this stretch is considered a target range. After the current fire front is divided into multiple target ranges, the fire spread direction vector and fire spread velocity vector within the target range are calculated based on the grid terrain data, grid combustible material data, and meteorological data of each grid that intersects with each target range.

[0071] For each target range, the specific process of calculating the fire spread speed vector of each target range is the process corresponding to S702-S707 described below.

[0072] S702: Calculate the surface moisture content within the target range based on the terrain characteristics and meteorological characteristics of the target grid.

[0073] Among them, the target grid is the grid that intersects with the target range, such as Figure 5 The shaded area shown, Figure 5 The grid with shadows in the middle is the target grid, for example, the grid with shadows between curves AB is the target grid.

[0074] Specifically, the Nelson model is applied to calculate the grid terrain data and grid meteorological data of the benchmark grid to obtain the surface moisture content within the target range.

[0075] S703: Calculate the flame intensity, heat transfer rate, surface moisture content, and combustible material characteristics of the target grid within the target range to obtain the surface fire spread rate.

[0076] Specifically, the Rothermel principle is applied to calculate the flame intensity, heat transfer rate, surface moisture content and combustible characteristics of the target grid within the target range to obtain the surface fire spread rate.

[0077] S704: Based on the tree crown moisture content and the height below the first branch, determine whether the tree crown within the target range will be ignited.

[0078] Specifically, the crown flame intensity required for the crown to ignite is calculated based on the crown moisture content and the height below the first branch. Based on the known surface flame intensity and the crown flame intensity, it is determined whether the crown will be ignited by the surface fire. If the crown will not ignite, execute S705; if the crown will ignite, execute S706 and S707.

[0079] The formula for calculating the crown flame intensity required for the crown to be ignited is: in, represents the crown flame intensity, CHB represents the height below the first branch, and FMC represents the crown moisture content.

[0080] S705: Determine the surface fire spread speed as the fire spread speed vector of the target range.

[0081] S706: Calculate the terrain characteristics, meteorological characteristics, and combustible characteristics of the target grid, as well as the heat transfer rate, flame intensity, and crown moisture content to obtain the crown fire spread rate.

[0082] It should be noted that the calculation process of the crown fire spread rate also applies the Rothermel principle, which will not be repeated here.

[0083] S707: Based on the surface fire spread speed and the crown fire spread speed, obtain a fire spread speed vector corresponding to the target range.

[0084] The surface fire spread velocity and the crown fire spread velocity are synthesized to obtain the fire spread velocity vector corresponding to the target range.

[0085] In the method provided in the embodiment of the present application, when calculating the fire spread velocity vector of the target range, it is necessary to comprehensively consider the surface fire spread velocity and the crown fire spread velocity within the target range. For a target range without vegetation, only the surface fire spread velocity needs to be considered.

[0086] When calculating the fire spread velocity vector, the fire spread direction within the target range must also be calculated. Figure 6 The specific implementation process of applying the LSTM model to calculate the fire range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread direction vector corresponding to the target fire range is as follows: S708: Based on the terrain characteristics, meteorological characteristics, and combustible material characteristics of the target grid, obtain the three-dimensional slope vector, three-dimensional wind direction vector, and combustible material position in the target grid corresponding to the target grid.

[0087] The 3D slope vector is a vector representation of the slope and aspect of the terrain surface within the target grid. The 3D wind direction vector is a vector representation of the direction and speed of the wind in 3D space.

[0088] Because the target range is within the boundary of the target fire area, when the target range intersects the target grid, the combustibles within the target grid may be in a combustible state or a non-combustible state. Combustibles in a combustible state include both burning combustibles and unburned combustibles. In this application, the combustible position obtained in S708 is the position of the combustibles in a combustible state.

[0089] S709: Constructing an initial direction vector corresponding to the target range based on the three-dimensional slope vector and the three-dimensional wind direction vector.

[0090] In the present application, the three-dimensional slope vector and the three-dimensional wind direction vector are synthesized to obtain a synthetic vector of the two, namely, the initial direction vector.

[0091] Since the three-dimensional slope vector and the three-dimensional wind direction vector have corresponding components in each direction in the three-dimensional coordinates, in the process of synthesizing the three-dimensional slope vector and the three-dimensional wind direction vector to obtain a synthetic vector, the initial direction vector is determined based on the synthesis of the sizes of the components in different dimensions.

[0092] S710: Based on the position of the combustible material, the initial direction vector is adjusted to obtain the fire spread direction vector corresponding to the target range.

[0093] It should be noted that if the distribution of combustibles in the target grid is uneven, the locations containing combustibles are more likely to catch fire. Therefore, the direction vectors of multiple combustible locations are synthesized, and the initial direction vector is adjusted based on the synthesized direction vector of the combustibles to more accurately identify the fire spread direction vector.

[0094] Furthermore, if the target grid contains burning combustibles and unburned combustibles, the vector magnitude of the direction vector corresponding to the burning combustibles is smaller than the vector magnitude of the direction vector of the unburned combustibles.

[0095] By applying the method provided in the embodiment of the present application, the three-dimensional wind direction, three-dimensional slope and position of combustibles in each target grid within the target range are comprehensively considered to obtain a more accurate fire spread direction vector within the target range.

[0096] In this application, after obtaining the fire spread direction vector and fire spread speed vector corresponding to the target range, the process of predicting the predicted fire line range corresponding to each time period is as follows using the fire spread direction vector and fire spread speed vector: Based on the fire spread direction vector and fire spread speed vector corresponding to the target range, an elliptical circular wave corresponding to the target range in each time period is drawn on the three-dimensional grid map; based on the elliptical circular wave corresponding to each target range in each time period, the predicted fire line range corresponding to the target fire scene in each time period is generated.

[0097] In this application, after obtaining a 3D grid map, the map is visualized, and the current fire range and the target range within the current fire range are also marked. While calculating the fire spread direction vector and fire spread velocity vector corresponding to the target range, three-dimensional elliptical circular waves are generated for multiple time periods within the 3D grid map based on the fire spread direction vector and fire spread velocity vector. By displaying the dynamic changes of the three-dimensional elliptical circular waves, the predicted fire ranges for multiple time periods are predicted.

[0098] refer to Figure 7The dynamic change process of the three-dimensional elliptical wave shown in the figure is as follows. Before the dynamic change, a three-dimensional elliptical wave corresponding to the target range AB, BC, CD, and DA has been generated for a certain future time period. After the dynamic change process, the fire line range of the time period is predicted to change from the current fire line range ABCD to the predicted fire line range of Figure 7 The area formed by A1B1C1D1 is shown in the figure. Given the fire spread direction vector and fire spread velocity vector, the target 3D circular wave's major and minor axis lengths are simulated and updated based on the fire spread velocity. Simultaneously, the center position of the target 3D circular wave is simulated and updated based on the fire spread velocity and fire spread direction. Dynamic changes in the major and minor axes, as well as the center position of the target 3D circular wave, are used to achieve dynamic changes. Based on the dynamic changes in the target 3D circular wave, a new fire area is regenerated, with the boundary of this fire area serving as the predicted fire front range for the next time period.

[0099] In this application, after obtaining the predicted fire line range for multiple time periods in the future through the fire line prediction model, it is necessary to wait for the next time period to receive the new fire line range and meteorological data of the target fire scene to further determine whether each predicted fire line range needs to be output to the client.

[0100] S8: Real-time detection of whether the regional meteorological data and fire line range of the target fire scene in the next time period are obtained.

[0101] If the regional meteorological data and the fire line range of the target fire scene in the next time period are obtained, S9 is executed; otherwise, S10 is executed.

[0102] S9: Do not output the predicted fire line ranges of the current time period, and return to execute S3.

[0103] Among them, when the real fire line range of the next time period is obtained, the fire line range predicted for the current time period may be different from the actual fire line range. At this time, it is necessary to reuse the fire line prediction model to update the fire line ranges of multiple future time periods to ensure that the predicted fire line ranges of multiple future time periods are closer to the evolution process of the real fire line range.

[0104] Since the real fire line range for the next time period has been obtained, the prediction can be continued based on the new real fire line range to update the prediction result of the latest fire line range. Therefore, when the real fire line range for the next time period is obtained, the predicted fire line range for the current time period is not output.

[0105] Optionally, without outputting each predicted fire line range, if a query request for each predicted fire line range in the current time period is received from the client, each predicted fire line range may be output to the client.

[0106] S10: Output each predicted fire line range in the current time period.

[0107] Before obtaining the fire line range for the next period, the predicted fire line range is output, and fire prevention can be carried out based on the predicted fire line range, thereby improving data support for forest and grassland fire safety.

[0108] In this embodiment of the present application, after obtaining predicted fire ranges for multiple future time periods, if the actual fire range for the next time period is obtained, the actual fire range is compared with the predicted fire range for that time period; based on the comparison result, the fire prediction model is updated. If the prediction process corresponding to S3-S10 is executed again, the updated fire prediction model is used to predict the fire range again, ensuring that the predicted fire range is closer to the actual result.

[0109] Optionally, during the execution of steps S1-S10 above, the regional meteorological data and the current fire front range obtained by the present application are both dynamic data. The present application can correct the current fire front range to respond to changes in the fire front range in real time. Similarly, correction parameters for meteorological data can also be set. The correction of this meteorological data is mainly implemented through a CFD model to address the differences in meteorology in each grid under different terrain conditions. By correcting the fire front range and the meteorological data, the present application can further improve the accuracy of subsequent fire spread predictions.

[0110] In the method provided in the embodiment of the present application, a three-dimensional grid map of the target fire area is constructed. The grid terrain data, grid combustible material data, grid meteorological data, and the current fire line range of the target fire area in the three-dimensional grid map are combined to calculate the fire spread speed vector and fire spread direction vector for each target range in the current fire line range. Based on the fire spread speed vector and fire spread direction vector, the change process of the current fire line range in the three-dimensional grid map is simulated to determine the fire line range predicted at the next moment.

[0111] The method for real-time correction and simulation prediction of forest and grassland fire situations provided in this application includes key steps such as data processing, construction of a three-dimensional grid map, correction of fire front ranges, three-dimensional fire spread simulation, and real-time correction and iterative calculation. The three-dimensional grid map accurately reflects the spatial distribution of regional terrain and combustibles, providing basic data support for subsequent wind field calculations and fire spread simulations. Regional terrain data (such as elevation, slope, and aspect) and combustible data (such as vegetation type, moisture, and fuel load) are acquired using remote sensing satellites, drones, and other technologies. The acquired data undergoes preprocessing, including data format conversion and coordinate unification. The processed regional terrain and combustible data are then imported into a geographic information system (GIS) to construct a three-dimensional grid map. The size of each grid cell can be customized (e.g., 10m×10m×1m). The grid contains grid terrain and combustible data, including elevation, slope, aspect, and combustible type.

[0112] This application also uses refined wind field calculations to accurately reflect the impact of local meteorological conditions on fire spread, especially in areas with complex terrain and vegetation distribution. Real-time meteorological data (such as wind speed, wind direction, temperature, humidity, etc.) in the region is obtained through weather stations, satellites, and other equipment. Based on the terrain and combustible material distribution, computational fluid dynamics (CFD) models or wind field interpolation algorithms are used to calculate the wind speed and direction vectors for each grid. For example, in areas with complex terrain, CFD models are used to simulate the three-dimensional distribution of the wind field; in flat areas, interpolation algorithms are used to calculate wind speed and direction. Using real-time regional meteorological data, the wind field of the meteorological data is dynamically corrected to adapt to sudden changes in meteorological conditions.

[0113] To determine the current fire front extent of a target fire area, image data of the fire area can be acquired through drone or satellite photography. Noise reduction processing is performed on the image data, image gradients are calculated, and non-maximum suppression and threshold filtering are performed to determine the current fire front extent. A thinning algorithm can also be used to further process the fire front boundary, eliminating noise interference and improving the clarity and continuity of the current fire front extent. Correction parameters are then applied to the current fire front extent to ensure data accuracy. Finally, the current fire front extent is segmented into multiple target areas. To calculate the fire spread direction, a three-dimensional elliptical circular wave is defined, with the center of each wave being a point on the fire front boundary. Using the Rothermel principle, the fire spread velocity at each point is calculated by integrating information about combustible materials, terrain, and wind speed within the area.

[0114] For example, the fire spread rate is calculated based on the moisture and load of the combustible material, and corrected by wind speed and slope.

[0115] Using the corrected wind direction and terrain vector, the fire spread direction is calculated for each point. For example, the shape and direction of the elliptical wave are determined based on the wind-slope vector. The calculated speed and direction are then combined to form a three-dimensional expansion of the elliptical wave at each point, simulating the dynamic process of fire spread.

[0116] The specific implementation of this application achieves high-precision simulation and real-time correction of forest and grassland fire situations through steps such as constructing a three-dimensional grid map, refining meteorological data, denoising and correcting the fire line range, three-dimensional fire spread simulation, and real-time correction and iterative calculation.

[0117] In an embodiment of the present application, before predicting the fire line range of the target fire scene, it is necessary to obtain a pre-trained fire line prediction model, which is a combination model of a CNN model and an LSTM model. After multiple iterative training, the fire line prediction model is applied to the above-mentioned prediction process of S3-S10.

[0118] Specifically, the training process of the fire prediction model of this application is as follows: Obtain historical fire front ranges and historical meteorological data for multiple historical time periods. This data can be obtained through web crawlers or a pre-set knowledge base. This includes the evolution of the fire front range and the historical meteorological data associated with each change. The fire front ranges for multiple historical time periods can be obtained based on the evolution of the fire front range. Furthermore, historical fire data also includes historical 3D fire data, including terrain data and combustible material data.

[0119] The fire prediction model is fed with historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges for each historical time period, and historical meteorological data to obtain a prediction result for each historical time period output by the fire prediction model. It is understood that in the process of feeding the fire prediction model with historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges for each historical time period, and historical meteorological data, the data can be fed sequentially in chronological order of the historical time periods. Each time data is fed, in addition to the historical three-dimensional fire scene data, only the historical fire line ranges and historical meteorological data for a certain historical time period are fed. After obtaining prediction results for other historical time periods later than the historical time period, the historical fire line ranges and historical meteorological data for the next historical time period are fed to continue the prediction.

[0120] Calculate the loss function between the prediction results for each historical time period and the historical fire front range. During the fire front range prediction process, the fire front prediction model compares the prediction results for each historical time period with the historical fire front range after each prediction output and calculates the loss function between the two. The loss function obtained after multiple predictions is used to determine whether the loss function has converged. The smaller the loss function, the better the training effect of the fire front prediction model. After multiple iterative training, the loss function obtained each time approaches stability, indicating that the prediction results output by the fire front prediction model are infinitely close to the true value (historical fire front range).

[0121] The model parameters of the fire prediction model are adjusted based on the loss function until the currently calculated loss function converges, thereby obtaining the trained fire prediction model. If the loss function does not converge, the model parameters of the fire prediction model are adjusted based on the loss function. After adjusting the model parameters, the fire prediction model is fed with historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges for each historical time period, and historical meteorological data. The above-mentioned processes (1) to (4) are re-executed until the currently calculated loss function converges, thereby terminating the training of the fire prediction model.

[0122] In the embodiment of the present application, after predicting the fire range for the next time period, the actual fire range at the next moment is obtained. Once the actual fire range for the next time period is obtained, the actual fire range is compared with the predicted fire range. Based on the comparison result, the fire prediction model is updated. If the fire range needs to be predicted again, the updated fire correction model is used to correct the fire range again, and the fire prediction model is used to predict the fire range for the next time period again, ensuring that the predicted fire range is closer to the actual result.

[0123] An embodiment of the present application also provides a storage medium, which includes stored instructions, wherein when the instructions are executed, the device where the storage medium is located is controlled to execute the above-mentioned forest and grassland fire situation real-time correction and simulation prediction processing method.

[0124] The present application also provides an electronic device, the structure of which is shown in FIG. Figure 8 As shown, the system specifically includes a memory 801 and one or more instructions 802, wherein the one or more instructions 802 are stored in the memory 801 and are configured to be executed by one or more processors 803 to perform the following operations: Constructing a three-dimensional grid map of the target fire scene, wherein the three-dimensional grid map of the target fire scene includes three-dimensional fire scene data of the target fire scene, and the three-dimensional fire scene data includes grid terrain data and grid combustible material data of each grid in the three-dimensional grid map; Real-time acquisition of regional meteorological data and fire front range of the target fire scene; Each time the regional meteorological data and the fire line range of the target fire scene are obtained, a preset prediction process is executed to update the fire line range of the target fire scene in real time for multiple time periods in the future; The prediction process is as follows: Apply a preset interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data of the current time period to obtain the current meteorological data of each grid; input the target fire line range, the current meteorological data and the three-dimensional fire scene data into a pre-trained fire line prediction model to obtain the predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model; when the regional meteorological data and fire line range of the target fire scene are obtained next time, re-execute the prediction process; wherein, the target fire line range is the current fire line range of the current time period or the fire line range after the current fire line range is corrected based on the historical predicted fire line range, the historical predicted fire line range is the fire line range of the current time period obtained by the prediction process executed last time, and the fire line prediction model is a combination model of the CNN model and the LSTM model; The training process of the fire prediction model is as follows: Obtain historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges in multiple historical time periods, and historical meteorological data; input the historical three-dimensional fire scene data of the historical fire scenes, the historical fire line ranges and historical meteorological data of each historical time period into the fire line prediction model, obtain the prediction results of each historical time period output by the fire line prediction model, calculate the loss function between the prediction results of each historical time period and the historical fire line range, and adjust the model parameters of the fire line prediction model based on the loss function until the loss function currently calculated converges, thereby obtaining the trained fire line prediction model.

[0125] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for real-time correction and simulation prediction of forest and grassland fire situation, characterized in that: include: Constructing a three-dimensional grid map of the target fire scene, wherein the three-dimensional grid map of the target fire scene includes three-dimensional fire scene data of the target fire scene, and the three-dimensional fire scene data includes grid terrain data and grid combustible material data of each grid in the three-dimensional grid map; Real-time acquisition of regional meteorological data and fire front range of the target fire scene; Each time the regional meteorological data and the fire line range of the target fire scene are obtained, a preset prediction process is executed to update the fire line range of the target fire scene in real time for multiple time periods in the future; The prediction process is as follows: Apply a preset interpolation algorithm or a fluid dynamics CFD model to correct the regional meteorological data of the current time period to obtain the current meteorological data of each grid; input the target fire line range, the current meteorological data and the three-dimensional fire scene data into a pre-trained fire line prediction model to obtain the predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model; when the regional meteorological data and fire line range of the target fire scene are obtained next time, re-execute the prediction process; wherein, the target fire line range is the current fire line range of the current time period or the fire line range after the current fire line range is corrected based on the historical predicted fire line range, the historical predicted fire line range is the fire line range of the current time period obtained by the prediction process executed last time, and the fire line prediction model is a combination model of the CNN model and the LSTM model; The training process of the fire prediction model is as follows: Obtain historical three-dimensional fire scene data of historical fire scenes, historical fire line ranges in multiple historical time periods, and historical meteorological data; input the historical three-dimensional fire scene data of the historical fire scenes, the historical fire line ranges and historical meteorological data of each historical time period into the fire line prediction model, obtain the prediction results of each historical time period output by the fire line prediction model, calculate the loss function between the prediction results of each historical time period and the historical fire line range, and adjust the model parameters of the fire line prediction model based on the loss function until the loss function currently calculated converges, thereby obtaining the trained fire line prediction model.

2. The method according to claim 1, characterized in that The three-dimensional grid map of the target fire scene is constructed, including: Using a preset data acquisition device to collect regional terrain data and regional combustible material data of the target fire scene; Based on the regional terrain data and regional combustible material data, construct a three-dimensional BIM model corresponding to the target fire scene; The three-dimensional BIM model is divided into multiple grids to obtain a three-dimensional grid map of the target fire scene.

3. The method according to claim 2, characterized in that Dividing the three-dimensional BIM model into a plurality of grids includes: Dividing the three-dimensional BIM model into a plurality of grids according to a preset grid size; or, Based on the regional terrain data and the regional combustible material data, each terrain type and / or combustible material type in the three-dimensional BIM model is determined; and according to each terrain type and / or each combustible material type, the three-dimensional BIM model is divided into multiple grids.

4. The method according to claim 2, characterized in that The application of a preset interpolation algorithm or a fluid dynamics CFD model to correct regional meteorological data for the current time period includes: determining a terrain type for each of the grid cells; When the terrain type of the grid is flat land, the interpolation algorithm is applied to calculate the regional meteorological data, the grid terrain data, and the grid combustible data for the current time period to obtain the current meteorological data of the grid for the current time period; When the terrain type of the grid is not flat land, the regional meteorological data, the grid terrain data and the grid combustible data of the current time period are input into the CFD model to obtain the current meteorological data of the grid in the current time period output by the CFD model.

5. The method according to claim 1, wherein The obtaining of the fire line range of the target fire scene includes: Applying a preset image recognition method to identify the initial fire line range of the target fire scene; The initial fire line range is processed using a preset edge algorithm to obtain the fire line range of the target fire scene.

6. The method according to claim 1, wherein The step of inputting the target fire line range, the current meteorological data, and the three-dimensional fire scene data into the pre-trained fire line prediction model to obtain predicted fire line ranges corresponding to multiple time periods output by the fire line prediction model includes: Inputting the target fire line range, the current meteorological data, and the three-dimensional fire scene data into a CNN model of the fire line prediction model, and obtaining fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics output by the CNN model; The LSTM model is applied to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics and combustible material characteristics to obtain the fire spread velocity vector and the fire spread direction vector corresponding to the target fire line range; based on the fire spread direction vector and the fire spread velocity vector, the predicted fire line range corresponding to each of the time periods is predicted.

7. The method according to claim 6, characterized in that The applying of the LSTM model to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain a fire spread velocity vector and a fire spread direction vector corresponding to the target fire line range includes: Divide a plurality of target ranges based on the fire line range characteristics, and input the flame intensity, heat transfer rate, crown moisture content and first branch height of each target range into the LSTM model; Calculating the surface moisture content within the target range based on the terrain characteristics and meteorological characteristics of the target grid, wherein the target grid is a grid that intersects with the target range; Calculating the flame intensity, the heat transfer rate, the surface moisture content, and the combustible material characteristics of the target grid within the target range to obtain a surface fire spread rate; determining whether tree crowns within the target range will be ignited based on the tree crown moisture content and the first branch height; If it will not be ignited, determining the surface fire spread speed as the fire spread speed vector of the target range; If it will be ignited, calculating the terrain characteristics, meteorological characteristics and combustible characteristics of the target grid, as well as the heat transfer rate, the flame intensity and the crown moisture content, to obtain the crown fire spread speed; A fire spread velocity vector corresponding to the target range is obtained based on the surface fire spread velocity and the crown fire spread velocity.

8. The method according to claim 7, characterized in that The applying of the LSTM model to calculate the fire line range characteristics, meteorological characteristics, terrain characteristics, and combustible material characteristics to obtain the fire spread direction vector corresponding to the target fire line range includes: Based on the terrain characteristics, meteorological characteristics and combustible material characteristics of the target grid, obtaining a three-dimensional slope vector, a three-dimensional wind direction vector and a combustible material position in the target grid corresponding to the target grid; constructing an initial direction vector corresponding to the target range based on the three-dimensional slope vector and the three-dimensional wind direction vector; Based on the position of the combustible material, the initial direction vector is adjusted to obtain a fire spread direction vector corresponding to the target range.

9. The method according to claim 8, characterized in that The predicting of the fire line range corresponding to each time period based on the fire spread direction vector and the fire spread speed vector includes: Based on the fire spread direction vector and the fire spread speed vector corresponding to the target range, drawing an elliptical circular wave corresponding to each time period of the target range on the three-dimensional grid map; Based on the elliptical circular wave corresponding to each target range in each time period, a predicted fire line range corresponding to the target fire scene in each time period is generated.

10. The method according to claim 1, characterized in that The correcting the current fire line range based on the historical predicted fire line range includes: Obtaining an observed fire line range in a current time period, wherein the observed fire line range is the fire line range in the current time period obtained by correcting the historical predicted fire line range before obtaining the current fire line range; Get the Kalman gain coefficient for the current time period; determining a difference parameter between the observed fire line range and the current fire line range; The target fire line range is obtained based on the Kalman gain coefficient, the difference parameter and the historical predicted fire line range.

Images