A method, apparatus, terminal equipment, and storage medium for wildfire spread risk assessment based on Bayesian networks.

Abstract

This invention discloses a method, apparatus, terminal device, and storage medium for wildfire spread risk assessment based on Bayesian networks, relating to the field of power system disaster prevention technology. The method includes: acquiring the latitude and longitude coordinates of the fire point, determining the prediction range, and dividing it into several grids; acquiring the vegetation distribution of the grids and determining the ignition coefficient; repeating the ignition simulation process: acquiring the daily meteorological parameters, real-time wind speed, and real-time wind direction, determining the relative wind direction spread probability of the grids, and converting it into four-directional spread probabilities (up, down, left, and right); using the ignition probability of the previous period as a priori, and with the four-directional spread probabilities and ignition coefficient as conditions, determining the current ignition probability; outputting the time series sequence of ignition probabilities for all grids when the current spread time is not less than the total spread duration; and determining the wildfire spread risk assessment result. By implementing this invention, the problem of low computational efficiency caused by the need to solve complex differential equations in existing physical models can be solved, improving the timeliness of wildfire spread risk assessment.

Description

Technical Field

[0001] This invention relates to the field of disaster prevention technology for power systems, and in particular to a method, apparatus, terminal equipment, and storage medium for assessing the risk of wildfire spread based on Bayesian networks. Background Technology

[0002] Currently, global warming continues to intensify, extreme weather events are occurring more frequently, and the frequency, scale, and destructive intensity of wildfires are all significantly increasing. These disasters not only disrupt regional ecological balance but also seriously threaten human life and property. Wildfire spread prediction is a core component of the wildfire prevention and mitigation system.

[0003] Current wildfire spread predictions mainly rely on physical models based on heat conduction. These physical models start from the physical nature of fire spread, using the three major heat transfer methods—conduction, convection, and radiation—as their logic to quantify the energy exchange process between flames and vegetation, thereby deriving the rate, direction, and extent of wildfire spread and revealing the physical laws governing wildfire spread.

[0004] However, physical models have a large number of key physical parameters, and complex physical processes require solving multivariable differential equations. Simulating a fire covering several square kilometers may take several hours or even days, far exceeding the spread rate of wildfires, which is several meters per minute, and cannot meet the needs of real-time prediction. Summary of the Invention

[0005] This invention provides a method, apparatus, terminal device, and storage medium for wildfire spread risk assessment based on Bayesian networks. It can solve the problem of low computational efficiency caused by the need to solve complex differential equations in existing physical models, and improve the timeliness of wildfire spread risk assessment.

[0006] One embodiment of the present invention provides a wildfire spread risk assessment method based on Bayesian networks, comprising: Obtain the latitude and longitude coordinates of the wildfire's ignition point; The predicted range of wildfire spread is determined based on latitude and longitude coordinates, and the predicted range is divided into several grids; Obtain the vegetation distribution of each grid cell and determine the initial ignition probability and ignition coefficient of the corresponding grid cell; Repeat the fire simulation process to obtain the fire probability time series of all grids; Based on the time series of ignition probabilities of all grids, the results of the wildfire spread risk assessment are determined; The fire simulation process includes: obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid. The relative wind direction spread probability of each grid is converted into the absolute spread probability in the four directions of up, down, left, and right in the grid coordinate system, so as to obtain the four-way spread probability of each grid. Using the fire probability of each grid in the previous period as a priori, and the four-way spread probability and fire coefficient of each grid as conditions, the current fire probability of the corresponding grid is determined; where the fire probability of the previous period is the initial fire probability. Update the current spread time, and if the current spread time is less than the preset total spread time, use the current ignition probability of all grids as the ignition probability of the corresponding grid in the previous period when the next ignition simulation process is executed. Otherwise, integrate the current ignition probability of each grid when the ignition simulation process is executed into the ignition probability time sequence of the corresponding grid.

[0007] Furthermore, the fire probability time series includes: several wildfire fire probabilities ordered in chronological order; After obtaining the ignition probability time series of all grids, the following is also included: For each time period, based on the probability of wildfire ignition in each grid and the correspondence between the preset ignition probability range and the color code, each grid is filled with color to obtain the wildfire spread prediction map for the corresponding time period. Based on wildfire spread prediction maps for all time periods, generate dynamic visualization results of wildfire spread.

[0008] Furthermore, the vegetation distribution of each grid cell is obtained to determine the initial ignition probability and ignition coefficient of the corresponding grid cell, including: Obtain the vegetation distribution for each grid cell; Based on the vegetation distribution of each grid and the preset correspondence between vegetation type and ignition coefficient, the ignition coefficient of the corresponding grid is determined. The initial fire probability of the corresponding grid is determined based on the vegetation distribution of each grid and the preset correspondence between vegetation type and initial fire probability.

[0009] Furthermore, the meteorological parameters for the day include: the highest daily temperature, the average wind speed at noon, and the lowest daily humidity; Obtain the daily meteorological parameters and real-time wind speed for the forecast area, as well as the real-time wind direction for each grid cell within the forecast area, and determine the relative wind direction spread probability for each grid cell, including: Obtain the daily maximum temperature, midday average wind speed, daily minimum humidity, and real-time wind speed for the forecast range; The initial spread rate of the predicted area is calculated based on the daily maximum temperature, the average wind speed at noon, and the daily minimum humidity. Calculate the wind speed adjustment coefficient for the forecast range based on real-time wind speed; The downwind spread rate of each grid is calculated based on the initial spread rate and wind speed adjustment coefficient of the predicted range, as well as the combustible coefficient of each grid. The combustible coefficient is determined based on the vegetation distribution of each grid and the pre-defined correspondence between vegetation type and combustible coefficient. Obtain the real-time wind direction for each grid cell within the forecast range; For each grid cell, the downwind spread rate is taken as the mean of the normal distribution, and the square root of the downwind spread rate is taken as the standard deviation of the normal distribution to obtain the normal distribution of the downwind spread rate. The relative wind direction spread probability is determined based on the normal distribution of real-time wind direction and downwind spread rate.

[0010] Furthermore, the relative wind direction spread probability includes: basic wind direction spread probability or intermediate wind direction spread probability; the basic wind direction spread probability includes: downwind direction spread probability, first downwind left and right side spread probability and headwind direction spread probability; the intermediate wind direction spread probability includes: second downwind left and right side spread probability and headwind left and right side spread probability. Based on the normal distribution of real-time wind direction and downwind spread rate, the relative wind direction spread probability is determined, including: When the real-time wind direction is within the basic wind direction, the probability of spreading in the downwind direction, the probability of spreading to the left and right sides of the first downwind direction, and the probability of spreading in the headwind direction are calculated based on the normal distribution of the downwind spread rate and the preset grid length. The basic wind direction includes: east, south, west or north. When the real-time wind direction is in the middle wind direction, the spread probability in the left and right directions of the second downwind and the spread probability in the left and right directions of the headwind are calculated based on the normal distribution of the downwind spread rate and the preset grid length; the middle wind direction includes: northeast, southeast, southwest or northwest wind.

[0011] Furthermore, the four-way spread probability includes: downward spread probability, upward spread probability, leftward spread probability, and rightward spread probability; Using the fire probability of each grid cell in the previous time period as a priori, and the four-way spread probability and fire coefficient of each grid cell as conditions, the current fire probability of the corresponding grid cell is determined, including: The probability of downward spread, upward spread, leftward spread, and rightward spread of each grid cell is summed and then weighted and corrected in conjunction with the ignition coefficient of the corresponding grid cell to obtain the neighborhood spread contribution probability of the corresponding grid cell. The neighborhood spread contribution probability of each grid cell is integrated with the fire probability of the corresponding grid cell in the previous period to obtain the current fire probability of the corresponding grid cell.

[0012] Furthermore, the results of the wildfire spread risk assessment include: the predicted area is in a state of potential fire threat or the predicted area is in a state of no significant fire threat; Based on the time series of ignition probabilities for all grid cells, the results of the wildfire spread risk assessment are determined, including: If the probability of a wildfire ignition exceeds a preset ignition probability threshold, the predicted area is determined to be in a state of potential fire threat; otherwise, the predicted area is determined to be in a state of no significant fire threat.

[0013] Based on the above method embodiments, the present invention provides corresponding device embodiments, including: a fire point coordinate acquisition module, a prediction range determination module, a vegetation parameter configuration module, a fire simulation module, and a spread risk assessment module; The fire point coordinate acquisition module is used to obtain the latitude and longitude coordinates of the wildfire fire point; The prediction range determination module is used to determine the predicted range of wildfire spread based on latitude and longitude coordinates, and divide the predicted range into several grids; The vegetation parameter configuration module is used to obtain the vegetation distribution of each grid and determine the initial ignition probability and ignition coefficient of the corresponding grid. The fire simulation module is used to repeatedly execute the fire simulation process to obtain the fire probability time series sequence of all grids. The fire simulation process includes: obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid; converting the relative wind direction spread probability of each grid into the absolute spread probabilities in the four directions (up, down, left, and right) in the grid coordinate system, to obtain the four-directional spread probability of each grid; using the fire probability of each grid in the previous period as a priori, and using the four-directional spread probability and fire coefficient of each grid as conditions, determining the current fire probability of the corresponding grid; wherein, the fire probability of the previous period is the initial fire probability; updating the current spread time, and if the current spread time is less than the preset total spread time, using the current fire probability of all grids as the fire probability of the corresponding grid in the previous period when the fire simulation process is executed next time; otherwise, integrating the current fire probability of each grid in each execution of the fire simulation process into the fire probability time series sequence of the corresponding grid. The spread risk assessment module is used to determine the wildfire spread risk assessment results based on the time series of ignition probabilities of all grids.

[0014] Based on the above method embodiments, the present invention provides a corresponding terminal device embodiment, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps of the wildfire spread risk assessment method based on Bayesian networks as described in the present invention.

[0015] Based on the above method embodiments, the present invention provides a corresponding computer-readable storage medium embodiment, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to execute the steps of the wildfire spread risk assessment method based on Bayesian networks as described in the present invention.

[0016] Compared with the prior art, the beneficial effects of this embodiment are as follows: This invention first obtains the latitude and longitude coordinates of the wildfire ignition point, determines the predicted range of wildfire spread based on these coordinates, and divides the predicted range into several grids. This discretizes the continuous fire field into standardized spatial units, replacing the complex continuous spatial modeling of traditional physical models and effectively reducing computational dimensionality. Next, the vegetation distribution of each grid is obtained, and the initial ignition probability and ignition coefficient of the corresponding grid are determined. The ignition simulation process is repeated. Specifically, the daily meteorological parameters and real-time wind speed of the predicted range, as well as the real-time wind direction of each grid within the predicted range, are obtained to determine the relative wind direction spread probability of each grid. The relative wind direction spread probability of each grid is converted into the absolute spread probabilities in the four directions (up, down, left, and right) under the grid coordinate system, resulting in the four-directional spread probability of each grid. The grid spread probability under different wind direction conditions is considered as a conditional probability node in a Bayesian network. The ignition probability of each grid in the previous period is used as a priori, and the four-directional spread probability and ignition coefficient of each grid are used as conditions, thereby constructing a structured probability transmission network. The fire spread relationship is established to determine the current ignition probability of the corresponding grid, enabling grid-level probabilistic time-series iteration and avoiding complex solutions to the physical model. Then, the current spread time is updated, and if the current spread time is less than the preset total spread time, the current ignition probability of all grids is used as the ignition probability of the corresponding grid in the previous time period for the next ignition simulation. Otherwise, the current ignition probability of each grid in each ignition simulation is integrated into a time-series sequence of ignition probabilities for the corresponding grid. This lightweight iterative computation replaces the long-term physical simulation of large-scale fire scenes, improving computational efficiency and meeting real-time prediction requirements. Finally, based on the time-series ignition probabilities of all grids, the wildfire spread risk assessment result is determined.

[0017] In summary, this invention constructs a fire inference process based on gridded probabilistic iteration, transforming wildfire spread prediction from solving complex differential equations to lightweight Bayesian probabilistic inference. This solves the problem of low computational efficiency caused by the need to solve complex differential equations in existing physical models, and improves the timeliness of wildfire spread risk assessment. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a wildfire spread risk assessment method based on Bayesian networks provided in an embodiment of the present invention. Figure 2This is a structural schematic diagram of the wildfire spread prediction range provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the grid coordinate system for predicting the spread of wildfires according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the distribution of basic wind direction spread probability provided by an embodiment of the present invention; Figure 5 This is a schematic diagram showing the distribution of the intermediate wind direction spread probability according to an embodiment of the present invention; Figure 6 This is a schematic diagram of a von Neumann neighborhood structure in a Bayesian network provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the dynamic prediction result visualization structure of forest fire spread probability provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of a wildfire spread risk assessment device based on Bayesian networks provided in an embodiment of the present invention. Detailed Implementation

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

[0020] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated.

[0021] like Figure 1 As shown, in order to address the problem of low computational efficiency caused by the need to solve complex differential equations in existing physical models, an embodiment of the present invention provides a wildfire spread risk assessment method based on Bayesian networks. This method includes at least the following steps: Step S1: Obtain the latitude and longitude coordinates of the wildfire's epicenter; For step S1, in the early stage of wildfire development, the latitude and longitude coordinates of the wildfire point are mainly obtained by satellite remote sensing monitoring. The energy signals released by the wildfire are captured by satellite sensors, and the latitude and longitude coordinates of the fire point are accurately determined by analyzing and judging these energy signals.

[0022] Preferably, other monitoring methods such as drone video surveillance or radar monitoring can also be used to further improve the accuracy and timeliness of fire point coordinate acquisition.

[0023] Step S2: Determine the predicted range of wildfire spread based on latitude and longitude coordinates, and divide the predicted range into several grids; For step S2, as Figure 2 As shown, a system is established with the latitude and longitude coordinates of the fire point as the geometric center. kilometer × A kilometer-wide area was used as the predicted range for wildfire spread; then, the predicted range was divided into... indivual The grid consists of grids. Each grid cell is a node in a Bayesian network, participating in the subsequent calculations of the spread probability and the ignition probability.

[0024] Establish a coordinate system with the top left corner of the prediction range as the origin, the horizontal direction to the right as the X-axis, and the vertical direction downwards as the Y-axis, and label the horizontal and vertical coordinates. For example... Figure 3 As shown, the grid coordinates of the red grid are (3,3).

[0025] It should be noted that, This indicates the preset length of the predicted wildfire spread in the east-west direction. This indicates the preset length of the predicted wildfire spread along a north-south direction. and It is preset, determining the size of the prediction range. When establishing the prediction range, the influence of terrain factors such as slope is not considered; it is based solely on the location of the fire point and the preset length. Complete the definition of the prediction range.

[0026] Step S3: Obtain the vegetation distribution of each grid cell and determine the initial ignition probability and ignition coefficient of the corresponding grid cell; In a preferred embodiment, the vegetation distribution of each grid cell is obtained, and the initial ignition probability and ignition coefficient of the corresponding grid cell are determined, including: Obtain the vegetation distribution for each grid cell; Based on the vegetation distribution of each grid and the preset correspondence between vegetation type and ignition coefficient, the ignition coefficient of the corresponding grid is determined. The initial fire probability of the corresponding grid is determined based on the vegetation distribution of each grid and the preset correspondence between vegetation type and initial fire probability.

[0027] For step S3, vegetation cover distribution information within each grid is obtained through remote sensing or field surveys to determine the main vegetation type of the grid. In this embodiment, the main vegetation types include combustible and non-combustible types. Combustible types include forest land such as flat pine needles, dead branches and fallen leaves, thatch and weeds, sedge and dwarf birch, pasture and grassland, red pine, Chinese pine and Yunnan pine, etc., while non-combustible types include land types such as roads, lakes and barren land where wildfires will not spread.

[0028] Based on the pre-defined correspondence between vegetation types and fire coefficients (as shown in Table 1), a corresponding fire coefficient is matched for each grid cell. Among them, the ignition coefficient These coefficients are used to characterize the intensity of the response of vegetation to the spread of fire. The values ​​are derived from the experience of a large number of wildfire burning experiments and historical cases.

[0029] Table 1. Correspondence between vegetation type and fire coefficient At the same time, the initial ignition probability of each grid is set. Specifically, the initial ignition probability of the grid where the fire point is located is set to 1, which means that the grid is already in a burning state; the initial ignition probability of non-combustible grids is set to 0, which means that they have no possibility of burning or being ignited by the spread of fire.

[0030] Step S4: Repeat the fire simulation process to obtain the fire probability time series sequence of all grids; The fire simulation process includes: obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid. The relative wind direction spread probability of each grid is converted into the absolute spread probability in the four directions of up, down, left, and right in the grid coordinate system, so as to obtain the four-way spread probability of each grid. Using the fire probability of each grid in the previous period as a priori, and the four-way spread probability and fire coefficient of each grid as conditions, the current fire probability of the corresponding grid is determined; where the fire probability of the previous period is the initial fire probability. Update the current spread time, and if the current spread time is less than the preset total spread time, use the current ignition probability of all grids as the ignition probability of the corresponding grid in the previous period when the next ignition simulation process is executed; otherwise, integrate the current ignition probability of each grid when the ignition simulation process is executed into the ignition probability time sequence of the corresponding grid. For step S4, before executing the fire simulation process in step S4, the total spread time is set. The starting time of the simulation Mark time 0 and determine the time interval between each simulation as . .

[0031] It should be noted that the total duration of the spread... The time interval can be set by comprehensively considering factors such as the predicted area of ​​spread, the initial size of the wildfire, and the type of surface vegetation, as well as the response time of power operation and maintenance personnel during disaster relief and rescue processes; For the temporal resolution of spread prediction, the time interval The smaller the slice, the more spread segments are generated, resulting in a more refined spread process. In this embodiment, the total spread time is... Set to 360 minutes.

[0032] By repeatedly executing the fire simulation process, iteratively calculating and outputting the fire probability time series of all grids, the dynamic law of wildfire spread over time can be fully depicted.

[0033] Specifically, each fire simulation process includes the following steps S401 to S404: Step S401: Obtain the daily meteorological parameters and real-time wind speed of the forecast range, as well as the real-time wind direction of each grid in the forecast range, and determine the relative wind direction spread probability of each grid. In a preferred embodiment, the daily meteorological parameters include: daily maximum temperature, average noon wind speed, and daily minimum humidity; Obtain the daily meteorological parameters and real-time wind speed for the forecast area, as well as the real-time wind direction for each grid cell within the forecast area, and determine the relative wind direction spread probability for each grid cell, including: Obtain the daily maximum temperature, midday average wind speed, daily minimum humidity, and real-time wind speed for the forecast range; The initial spread rate of the predicted area is calculated based on the daily maximum temperature, the average wind speed at noon, and the daily minimum humidity. Calculate the wind speed adjustment coefficient for the forecast range based on real-time wind speed; The downwind spread rate of each grid is calculated based on the initial spread rate and wind speed adjustment coefficient of the predicted range, as well as the combustible coefficient of each grid. The combustible coefficient is determined based on the vegetation distribution of each grid and the pre-defined correspondence between vegetation type and combustible coefficient. Obtain the real-time wind direction for each grid cell within the forecast range; For each grid cell, the downwind spread rate is taken as the mean of the normal distribution, and the square root of the downwind spread rate is taken as the standard deviation of the normal distribution to obtain the normal distribution of the downwind spread rate. The relative wind direction spread probability is determined based on the normal distribution of real-time wind direction and downwind spread rate.

[0034] For step S401, the meteorological parameters of the forecast area for the day are collected by ground meteorological stations, including the daily maximum temperature of the forecast area, the average wind speed at noon in the forecast area, the daily minimum humidity of the forecast area, and the real-time wind speed above the forecast area.

[0035] Based on the predicted area's daily maximum temperature, midday average wind speed, and daily minimum humidity, the initial spread rate of the predicted area is calculated using the following formula. This quantifies the basic fire spread capacity under meteorological conditions, visually demonstrating that high temperatures, strong winds, and low humidity significantly increase the initial spread potential of wildfires. ; in, This indicates the daily maximum temperature for the predicted area, in units of... , This indicates the average wind speed at noon in the predicted area. This represents the daily minimum humidity for the predicted area, in units of... , This indicates the initial spread rate of the predicted range.

[0036] Based on the real-time wind speed above the predicted area, the wind speed adjustment coefficient for the predicted range is calculated using the following formula. This exponentially amplifies the impact of wind speed on the spread rate, accurately capturing the intensifying effect of wind speed changes on fire propagation: ; in, This indicates the real-time wind speed above the forecast area, in units of... , The wind speed adjustment factor represents the forecast range.

[0037] Next, based on the vegetation distribution of each grid cell and the preset correspondence between vegetation type and combustible coefficient, the combustible coefficient is determined. Specifically, based on the pre-defined correspondence between vegetation types and combustible coefficients (as shown in Table 2), a corresponding combustible coefficient is matched for each grid cell. Among them, the combustibility coefficient A configuration pattern used to characterize the flammability and chemical properties of a combustible material and whether they are conducive to combustion.

[0038] Table 2. Correspondence between vegetation type and combustible material coefficient For each grid cell, the downwind spread rate is calculated using the following formula, based on the initial spread rate and wind speed adjustment factor of the predicted range, and the combustible material coefficient of each grid cell: ; in, Indicates the rate of spread downwind, in units of .

[0039] By using the above multi-parameter coupled downwind spread rate calculation, the basic response of the initial spread rate to the overall dryness and wind conditions of the region is preserved. The wind speed adjustment coefficient is used to accurately amplify the strengthening effect of the real-time wind field on the fire. At the same time, by using the combustible coefficient, the differences in flammability of different vegetation types are directly mapped to the spread rate, thus realizing the quantification of the downwind spread rate of wildfires at the grid scale.

[0040] Subsequently, the rate of spread with the wind As the mean of the normal distribution, and the square root of the downwind spread rate. Using the standard deviation of the normal distribution, we obtain the normal distribution of the downwind spread rate: ; in, The random variable representing the rate of spread downwind follows a normal distribution. This represents the mean of a normal distribution, and its value is the downwind spread rate of the grid. ,Right now This represents the expected speed at which the fire spreads downwind. The standard deviation of the normal distribution is represented by the square root of the downwind spread rate. ,Right now It is used to reflect the random fluctuations caused by environmental disturbances, combustion instability and other factors during the actual spread of fire.

[0041] By modeling the fire using the normal distribution described above, we can not only preserve the core trend of the downwind spread rate, but also quantify the uncertainty of fire spread through the standard deviation, which is more in line with the random fluctuation characteristics of the actual spread of wildfires.

[0042] Next, the real-time wind direction corresponding to each grid in the prediction range is obtained. The real-time wind direction covers wind field information in all directions, including east, south, northeast, and northwest, so as to clarify the dominant wind direction at each grid.

[0043] Based on this, and combined with the normal distribution of the downwind spread rate, the relative wind direction spread probability is determined according to the real-time wind direction type of each grid, and the influence of wind field direction changes on the probability of fire spread between adjacent grids is quantified, thus realizing the transformation from downwind spread capability to multi-directional spread probability.

[0044] Preferably, the relative wind direction spread probability includes: basic wind direction spread probability or intermediate wind direction spread probability; the basic wind direction spread probability includes: downwind direction spread probability, first downwind left and right side spread probability and headwind direction spread probability; the intermediate wind direction spread probability includes: second downwind left and right side spread probability and headwind left and right side spread probability. Based on the normal distribution of real-time wind direction and downwind spread rate, the relative wind direction spread probability is determined, including: When the real-time wind direction is within the basic wind direction, the probability of spreading in the downwind direction, the probability of spreading to the left and right sides of the first downwind direction, and the probability of spreading in the headwind direction are calculated based on the normal distribution of the downwind spread rate and the preset grid length. The basic wind direction includes: east, south, west or north. When the real-time wind direction is in the middle wind direction, the spread probability in the left and right directions of the second downwind and the spread probability in the left and right directions of the headwind are calculated based on the normal distribution of the downwind spread rate and the preset grid length; the middle wind direction includes: northeast, southeast, southwest or northwest wind.

[0045] Specifically, with due north as the reference, a direction is divided every 45° clockwise. The basic wind direction refers to the four wind directions of east, south, west, and north, while the intermediate wind direction refers to the four wind directions of northeast, southeast, southwest, and northwest.

[0046] Determine whether the real-time wind direction is a basic wind direction or an intermediate wind direction, such as Figure 4 As shown, if the real-time wind direction belongs to the basic wind direction, the probability of spreading in the downwind direction, the probability of spreading in the left and right sides of the first downwind direction, and the probability of spreading in the headwind direction are calculated in sequence.

[0047] Regarding the calculation of the probability of fire spreading in the downwind direction, the probability of a fire spreading in the downwind direction can be calculated as follows: [The probability is missing from the original text, so the translation is incomplete.] Within the grid, the probability of a flame spreading along the current burning grid and reaching an adjacent grid, i.e., the probability of spreading in the downwind direction, is equal to the probability that the travel time of the flame between two grids is less than a given time interval. The travel time is equal to the grid length divided by the spread rate. Therefore, the probability of a fire spreading in the wind direction can be estimated as follows: : ; in, This indicates the time it takes for the flame to travel between two grid squares. Indicates the length of the grid. Indicates the rate of spread downwind The probability density function corresponds to the normal distribution mentioned above. The cumulative density function represents the rate of spread downwind.

[0048] The normal distribution of the downwind spread rate has been established. Based on this, to facilitate the calculation of the cumulative distribution probability, the downwind spread rate is... Standardized to a standard normal distribution variable The specific formula is as follows: ; Therefore, the probability of the wildfire spreading downwind is: ; in, The cumulative distribution function representing the standard normal distribution can be obtained by looking up the standard normal distribution table.

[0049] Regarding the calculation of the probability of spread in the left and right directions of the first tailwind, the probability of spread in the left and right directions of the first tailwind includes the probability of spread in the left direction and the probability of spread in the right direction of the first tailwind, and the probability of spread in the left direction of the first tailwind is equal to the probability of spread in the right direction of the first tailwind, both being the probability of spread in the tailwind direction. This conforms to the true law that fire spreads in the downwind direction and evenly to both sides under the basic wind direction, avoiding simulation distortion caused by excessively high or low probability on one side.

[0050] Regarding the calculation of the probability of spread in the upwind direction, the probability of spread in the upwind direction is equal to the probability of spread in the downwind direction. This reduces the likelihood of propagation in the upwind direction, reflecting the physical characteristic that wildfires are difficult to spread in the opposite direction. It retains the possibility of propagation in the weak upwind direction under extreme wind conditions or combustion disturbances, while avoiding an excessively high probability of upwind propagation that contradicts the actual fire spread pattern. This makes the distribution of the probability of multi-directional spread more consistent with the actual spread characteristics of wildfires.

[0051] Taking the above scenario of wind direction from south to north as an example, the fire spread rate... Follows a normal distribution, mean for Standard deviation for and time interval Set to 10 minutes, grid length Calculate the critical propagation rate under the condition of 100 meters: ; Substituting into the standardized formula, we get: ; By consulting the standard normal cumulative distribution function, we can obtain: ; Ultimately, the probability of the wildfire spreading northward with the wind was calculated: ; Probability of spreading in the downwind direction Based on this, the probability of spread in the left and right directions with the first tailwind and the probability of spread in the headwind direction are further calculated: ; ; in, This indicates the probability of spreading in the downwind direction. This indicates the probability of spread in the left direction with the first tailwind. This indicates the probability of spread in the first downwind right direction. This indicates the probability of spreading in the direction of the headwind.

[0052] like Figure 5 As shown, if the real-time wind direction is a neutral wind direction, the spread probability of the second tailwind to the left and right sides and the spread probability of the headwind to the left and right sides are calculated in sequence. The spread probability of the second tailwind to the left and right sides includes the spread probability of the second tailwind to the left side and the spread probability of the second tailwind to the right side. The spread probability of the headwind to the left and right sides includes the spread probability of the headwind to the left side and the spread probability of the headwind to the right side.

[0053] Regarding the calculation of the probability of spread in the left and right directions of the second tailwind, similarly, the calculation process for the probability of spread in the tailwind direction is referred to above to obtain the probability of spread in the tailwind direction under the current intermediate wind direction scenario. Since the intermediate wind direction forms a 45° angle with the basic wind direction, the probability of spread in the tailwind direction is compared with the coefficient. By performing a product operation, we can obtain the spread probability in the second downwind left direction and the second downwind right direction, respectively. The specific formula is as follows: ; The calculation of the probability of spread in the left and right directions against the headwind is based on the probability of spread in the downwind direction under the current intermediate wind direction scenario, and is compared with the coefficient. By performing a product operation, we can obtain the probabilities of spread in the left direction and the right direction against the wind, respectively. The specific formulas are as follows: ; in, This indicates the probability of spread in the left direction of the second tailwind. This indicates the probability of spread in the second downwind right direction. This indicates the probability of spread in the downwind direction under a mid-wind scenario. This indicates the probability of spreading in the direction to the left of the headwind. This indicates the probability of spreading in the direction to the right of the headwind.

[0054] Step S402: Convert the relative wind direction spread probability of each grid into the absolute spread probability in the four directions of up, down, left, and right in the grid coordinate system to obtain the four-way spread probability of each grid. In a preferred embodiment, the four-way propagation probability includes: downward propagation probability, upward propagation probability, leftward propagation probability, and rightward propagation probability; For step S402, the relative wind direction spread probability of each grid is converted into the absolute spread probability in the four directions of up, down, left, and right in the grid coordinate system. Essentially, it maps the relative direction description based on the wind, headwind, or crosswind to the absolute orientation of up, down, left, and right in the grid, while introducing the time dimension to mark the timeliness of the probability.

[0055] Combination Figure 6The von Neumann neighborhood structure in the Bayesian network shown matches the relative spread probabilities of downwind, headwind, and crosswind directions to the grid axes one by one, based on the correspondence between real-time wind direction (basic wind direction or intermediate wind direction) and grid coordinate axes. ,Down ,Left ,right The time points are then clearly marked using subscript formatting in four absolute directions. , refer to At any given moment, each grid cell is ultimately obtained. The absolute spread probability at any given time to the four adjacent cells (up, down, left, and right), i.e., the downward spread probability. Upward spread probability Leftward spread probability and rightward spread probability .

[0056] Step S403: Using the fire probability of each grid cell in the previous period as a priori, and using the four-way spread probability and fire coefficient of each grid cell as conditions, determine the current fire probability of the corresponding grid cell; wherein, the fire probability of the previous period is the initial fire probability. In a preferred embodiment, the current ignition probability of a corresponding grid cell is determined by using the ignition probability of each grid cell in the previous time period as a priori, and the four-way spread probability and ignition coefficient of each grid cell as conditions, including: The probability of downward spread, upward spread, leftward spread, and rightward spread of each grid cell is summed and then weighted and corrected in conjunction with the ignition coefficient of the corresponding grid cell to obtain the neighborhood spread contribution probability of the corresponding grid cell. The neighborhood spread contribution probability of each grid cell is integrated with the fire probability of the corresponding grid cell in the previous period to obtain the current fire probability of the corresponding grid cell.

[0057] For step S403, for each grid cell ,exist Probability of fire at any time It is made of grid The probability of four-way spread at any given time, and the grid at... Probability of fire at any time This is determined collectively. Under the same fire intensity, the probability of ignition varies among different grid cells due to factors such as vegetation moisture content, vegetation porosity, and combustible fuel load. Therefore, for each grid cell... exist Probability of fire at any time As a priori, for the grid Downward propagation probability Upward spread probability Leftward spread probability and rightward spread probability Summing is performed to obtain the total diffusion potential from the surrounding neighborhood, and then combined with the grid. ignition coefficient Weighted adjustments are made to reflect the sensitivity of different vegetation types to fire spread, thus obtaining the grid. The probability of neighborhood spread.

[0058] Then, the grid Neighborhood spread contribution probability and grid exist The ignition probabilities at different times are summed to obtain the grid. exist The probability of fire at any given moment is calculated using the following formula: ; in, Represents grid exist The probability of fire at any given moment. Represents grid exist The probability of fire at any given moment. Represents grid exist Downward propagation probability at time, Represents a grid exist The probability of upward spread at any given time, Represents grid exist The probability of leftward spread at time 1. Represents a grid exist The probability of rightward spread at time step.

[0059] It should be noted that during the first fire simulation, there is no fire probability data for earlier periods. In this case, the fire probability for the previous period is the initial fire probability.

[0060] Step S404: Update the current spread time, and if the current spread time is less than the preset total spread time, use the current ignition probability of all grids as the ignition probability of the corresponding grid in the previous period when the next ignition simulation process is executed; otherwise, integrate the current ignition probability of each grid when the ignition simulation process is executed into the ignition probability time sequence of the corresponding grid. In a preferred embodiment, the fire probability time series includes: several wildfire ignition probabilities ordered in chronological order; For step S404, the current propagation time is updated using the following formula, advancing the extrapolation time forward by one time interval. : ; in, Indicates the current spread moment after the update. This indicates the period of spread before the update.

[0061] Then, determine the current spread time after the update. Whether it spreads for a total duration If the current spread time after the update Less than the total duration of spread If the predicted end time has not been reached, then the current fire probability of all grid cells calculated in this round will be used as the fire probability of the previous period in the next round of fire simulation, realizing the transfer and iteration of the time sequence state, and repeating steps S401 to S404; if the updated current spread time The total duration of the spread has been reached or exceeded. If the iteration stops, the current ignition probability of each grid obtained in each round of simulation is taken as the wildfire ignition probability, and sorted in chronological order to integrate into the ignition probability time series of the corresponding grid, thus obtaining complete spatiotemporal evolution data of wildfire spread, providing traceable data basis for subsequent analysis of wildfire spread trend and assessment of risk areas.

[0062] Preferably, after obtaining the ignition probability time series of all grids, the method further includes: For each time period, based on the probability of wildfire ignition in each grid and the correspondence between the preset ignition probability range and the color code, each grid is filled with color to obtain the wildfire spread prediction map for the corresponding time period. Based on wildfire spread prediction maps for all time periods, generate dynamic visualization results of wildfire spread.

[0063] Specifically, in order to intuitively present the spatiotemporal evolution of wildfire spread, the time series of ignition probabilities for all grids is visualized. Specifically, based on the pre-set correspondence between ignition probability intervals and color card codes (as shown in Table 3), the ignition probability of each grid is matched to the corresponding probability interval, and then the grid is filled with the color card code of the corresponding interval, thereby generating a wildfire spread prediction map for that period. Subsequently, the wildfire spread prediction maps for all periods are concatenated in chronological order to finally generate a dynamic visualization result of wildfire spread.

[0064] Table 3. Correspondence between fire probability ranges and color code charts In this embodiment, as Figure 7 The diagram shows a visualization of the dynamic prediction results for the probability of forest fire spread. Figure 7 (a) to Figure 7(d) shows the spatial distribution of wildfire ignition probability in the gridded area at 30 min, 60 min, 90 min and 180 min respectively. Each grid cell is presented with the corresponding ignition probability in a visual way through color. The low probability range corresponds to the green series, the medium probability range corresponds to the yellow series, and the high probability range corresponds to the red series, clearly distinguishing the degree of ignition risk of different grids.

[0065] At the same time, combined Figure 7 (a) to Figure 7 (d) It can be seen that as the time step progresses, the range and location of the high probability of fire can be observed intuitively, depicting the dynamic process of the fire spreading from the initial combustion point to the surrounding grid and the high probability of spread gradually expanding, thus fully presenting the spatiotemporal spread pattern of the wildfire within the predicted time.

[0066] Step S5: Determine the wildfire spread risk assessment results based on the fire probability time series of all grids.

[0067] In a preferred embodiment, the wildfire spread risk assessment results include: the predicted area is in a state of potential fire threat or the predicted area is in a state of no significant fire threat; Based on the time series of ignition probabilities for all grid cells, the results of the wildfire spread risk assessment are determined, including: If the probability of a wildfire ignition exceeds a preset ignition probability threshold, the predicted area is determined to be in a state of potential fire threat; otherwise, the predicted area is determined to be in a state of no significant fire threat.

[0068] For step S5, after obtaining the ignition probability time series of all grids through the ignition deduction process in step S4, the ignition probability time series of all grids are traversed. If the ignition probability of any grid exceeds a preset threshold, in this embodiment, the ignition probability threshold is set to 0.8, that is, the ignition probability falls within the interval [0.8, 0.9) or... Within the interval, the predicted range is determined to be in a state of potential fire threat; if the ignition probability of all grids does not exceed the preset threshold, the predicted range is determined to be in a state of no significant fire threat.

[0069] Preferably, the predicted fire probability falls within the range of [0.8, 0.9) or... The regions within the interval are overlaid with the distribution information of important infrastructure such as power transmission line corridors, substations, major transportation routes, residential areas, and ecologically sensitive areas in the actual geographic information system. The output includes the estimated arrival time of wildfires and early warning information. For example, it is predicted that a certain substation will be in the second-level wildfire warning spread zone after "hh:mm", and will be upgraded to the first-level wildfire warning spread zone after "hh:mm".

[0070] This invention improves the computational efficiency and spatial accuracy of wildfire spread prediction compared to standalone statistical or physical models by iteratively calculating the grid spread probability and ignition probability based on Bayesian networks.

[0071] In terms of computational efficiency, the inherent probabilistic reasoning characteristics of Bayesian networks, combined with the gridded von Neumann neighborhood structure, decompose global propagation prediction into probability transfer and update of local neighborhoods, avoiding the overhead of solving complex partial differential equations in physical models, and significantly reducing the computational complexity of long-term series prediction. In terms of spatial accuracy, this invention quantifies key influencing factors such as wind field and vegetation type into four-way spread probability and ignition coefficient. Through probabilistic inference of Bayesian network, it achieves collaborative modeling of multiple factors, which not only retains the ability of physical model to depict the fire spread law, but also avoids the statistical model's neglect of spatial heterogeneity. It can accurately capture the spread differences and local risk mutations between different grids, making the prediction results more consistent with the spatiotemporal diffusion characteristics of real wildfires.

[0072] like Figure 8 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided; An embodiment of the present invention provides a wildfire spread risk assessment device based on Bayesian networks, including: a fire point coordinate acquisition module, a prediction range determination module, a vegetation parameter configuration module, a fire simulation module, and a spread risk assessment module; The fire point coordinate acquisition module is used to obtain the latitude and longitude coordinates of the wildfire fire point; The prediction range determination module is used to determine the predicted range of wildfire spread based on latitude and longitude coordinates, and divide the predicted range into several grids; The vegetation parameter configuration module is used to obtain the vegetation distribution of each grid and determine the initial ignition probability and ignition coefficient of the corresponding grid. The fire simulation module is used to repeatedly execute the fire simulation process to obtain the fire probability time series sequence of all grids. The fire simulation process includes: obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid; converting the relative wind direction spread probability of each grid into the absolute spread probabilities in the four directions (up, down, left, and right) in the grid coordinate system, to obtain the four-directional spread probability of each grid; using the fire probability of each grid in the previous period as a priori, and using the four-directional spread probability and fire coefficient of each grid as conditions, determining the current fire probability of the corresponding grid; wherein, the fire probability of the previous period is the initial fire probability; updating the current spread time, and if the current spread time is less than the preset total spread time, using the current fire probability of all grids as the fire probability of the corresponding grid in the previous period when the fire simulation process is executed next time; otherwise, integrating the current fire probability of each grid in each execution of the fire simulation process into the fire probability time series sequence of the corresponding grid. The spread risk assessment module is used to determine the wildfire spread risk assessment results based on the time series of ignition probabilities of all grids.

[0073] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the Bayesian network-based wildfire spread risk assessment method provided by any of the above-described method embodiments of the present invention.

[0074] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0075] Based on the above embodiments of the wildfire spread risk assessment method based on Bayesian networks, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the wildfire spread risk assessment method based on Bayesian networks of any embodiment of the present invention.

[0076] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0077] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0078] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device via various interfaces and lines.

[0079] Based on the above-described method embodiments, another embodiment is provided: another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the wildfire spread risk assessment method based on Bayesian networks described in any of the above-described method embodiments of the present invention.

[0080] The modules / units integrated into the Bayesian network-based wildfire spread risk assessment device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0081] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for assessing wildfire spread risk based on Bayesian networks, characterized in that, include: Obtain the latitude and longitude coordinates of the wildfire's ignition point; The predicted range of wildfire spread is determined based on latitude and longitude coordinates, and the predicted range is divided into several grids; Obtain the vegetation distribution of each grid cell and determine the initial ignition probability and ignition coefficient of the corresponding grid cell; Repeat the fire simulation process to obtain the fire probability time series of all grids; Based on the time series of ignition probabilities of all grids, the results of the wildfire spread risk assessment are determined; The fire simulation process includes: obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid. The relative wind direction spread probability of each grid is converted into the absolute spread probability in the four directions of up, down, left, and right in the grid coordinate system, so as to obtain the four-way spread probability of each grid. Using the fire probability of each grid in the previous period as a priori, and the four-way spread probability and fire coefficient of each grid as conditions, the current fire probability of the corresponding grid is determined; where the fire probability of the previous period is the initial fire probability. Update the current spread time, and if the current spread time is less than the preset total spread time, use the current ignition probability of all grids as the ignition probability of the corresponding grid in the previous period when the next ignition simulation process is executed. Otherwise, integrate the current ignition probability of each grid when the ignition simulation process is executed into the ignition probability time sequence of the corresponding grid.

2. The wildfire spread risk assessment method based on Bayesian networks according to claim 1, characterized in that, The fire probability time series includes: several wildfire fire probabilities ordered in chronological order; After obtaining the ignition probability time series of all grids, the following is also included: For each time period, based on the probability of wildfire ignition in each grid and the correspondence between the preset ignition probability range and the color code, each grid is filled with color to obtain the wildfire spread prediction map for the corresponding time period. Based on wildfire spread prediction maps for all time periods, generate dynamic visualization results of wildfire spread.

3. The wildfire spread risk assessment method based on Bayesian networks according to claim 1, characterized in that, Obtain the vegetation distribution for each grid cell, and determine the initial ignition probability and ignition coefficient for the corresponding grid cell, including: Obtain the vegetation distribution for each grid cell; Based on the vegetation distribution of each grid and the preset correspondence between vegetation type and ignition coefficient, the ignition coefficient of the corresponding grid is determined. The initial fire probability of the corresponding grid is determined based on the vegetation distribution of each grid and the preset correspondence between vegetation type and initial fire probability.

4. The wildfire spread risk assessment method based on Bayesian networks according to claim 3, characterized in that, The meteorological parameters for the day include: the highest daily temperature, the average wind speed at noon, and the lowest daily humidity; The process of obtaining the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid within the prediction range, and determining the relative wind direction spread probability of each grid includes: Obtain the daily maximum temperature, midday average wind speed, daily minimum humidity, and real-time wind speed for the forecast range; The initial spread rate of the predicted area is calculated based on the daily maximum temperature, the average wind speed at noon, and the daily minimum humidity. Calculate the wind speed adjustment coefficient for the forecast range based on real-time wind speed; The downwind spread rate of each grid is calculated based on the initial spread rate and wind speed adjustment coefficient of the predicted range, as well as the combustible coefficient of each grid; wherein the combustible coefficient is determined based on the vegetation distribution of each grid and the preset correspondence between vegetation type and combustible coefficient. Obtain the real-time wind direction for each grid cell within the forecast range; For each grid cell, the downwind spread rate is taken as the mean of the normal distribution, and the square root of the downwind spread rate is taken as the standard deviation of the normal distribution to obtain the normal distribution of the downwind spread rate. The relative wind direction spread probability is determined based on the normal distribution of real-time wind direction and downwind spread rate.

5. The wildfire spread risk assessment method based on Bayesian networks according to claim 4, characterized in that, The relative wind direction spread probability includes: basic wind direction spread probability or intermediate wind direction spread probability; the basic wind direction spread probability includes: downwind direction spread probability, first downwind left and right side spread probability and headwind direction spread probability; the intermediate wind direction spread probability includes: second downwind left and right side spread probability and headwind left and right side spread probability. The determination of the relative wind direction spread probability based on the normal distribution of real-time wind direction and downwind spread rate includes: When the real-time wind direction is within the basic wind direction, the spread probability in the downwind direction, the spread probability in the left and right sides of the first downwind direction, and the spread probability in the headwind direction are calculated based on the normal distribution of the downwind spread rate and the preset grid length; the basic wind direction includes: east wind, south wind, west wind, or north wind. When the real-time wind direction is in the middle wind direction, the spread probability in the left and right directions of the second downwind and the spread probability in the left and right directions of the headwind are calculated based on the normal distribution of the downwind spread rate and the preset grid length; the middle wind direction includes: northeast wind, southeast wind, southwest wind or northwest wind.

6. The wildfire spread risk assessment method based on Bayesian networks according to claim 1, characterized in that, The four-way propagation probabilities include: downward propagation probability, upward propagation probability, leftward propagation probability, and rightward propagation probability; The process of determining the current fire probability of a corresponding grid cell by using the fire probability of the previous time period of each grid cell as a priori, and using the four-way spread probability and fire coefficient of each grid cell as conditions, includes: The probability of downward spread, upward spread, leftward spread, and rightward spread of each grid cell is summed and then weighted and corrected in conjunction with the ignition coefficient of the corresponding grid cell to obtain the neighborhood spread contribution probability of the corresponding grid cell. The neighborhood spread contribution probability of each grid cell is integrated with the fire probability of the corresponding grid cell in the previous period to obtain the current fire probability of the corresponding grid cell.

7. The wildfire spread risk assessment method based on Bayesian networks according to claim 2, characterized in that, The results of the wildfire spread risk assessment include: the predicted area is in a state of potential fire threat or the predicted area is in a state of no significant fire threat; Based on the time series of ignition probabilities for all grid cells, the results of the wildfire spread risk assessment are determined, including: If the probability of a wildfire ignition exceeds a preset ignition probability threshold, the predicted area is determined to be in a state of potential fire threat; otherwise, the predicted area is determined to be in a state of no significant fire threat.

8. A wildfire spread risk assessment device based on Bayesian networks, characterized in that, include: The module includes a fire point coordinate acquisition module, a prediction range determination module, a vegetation parameter configuration module, an ignition simulation module, and a spread risk assessment module. The fire point coordinate acquisition module is used to acquire the latitude and longitude coordinates of the wildfire fire point; The prediction range determination module is used to determine the predicted range of wildfire spread based on latitude and longitude coordinates, and divide the predicted range into several grids. The vegetation parameter configuration module is used to obtain the vegetation distribution of each grid and determine the initial ignition probability and ignition coefficient of the corresponding grid. The fire simulation module is used to repeatedly execute the fire simulation process to obtain the fire probability time series sequence of all grids. The fire simulation process includes: acquiring the daily meteorological parameters and real-time wind speed of the prediction range, as well as the real-time wind direction of each grid in the prediction range, and determining the relative wind direction spread probability of each grid; converting the relative wind direction spread probability of each grid into the absolute spread probabilities in the four directions (up, down, left, and right) under the grid coordinate system, obtaining the four-directional spread probability of each grid; using the fire probability of each grid in the previous period as a priori, and using the four-directional spread probability and fire coefficient of each grid as conditions, determining the current fire probability of the corresponding grid; wherein, initially, the fire probability of the previous period is the initial fire probability; updating the current spread time, and if the current spread time is less than the preset total spread time, using the current fire probability of all grids as the fire probability of the corresponding grid in the previous period for the next execution of the fire simulation process; otherwise, integrating the current fire probability of each grid in each execution of the fire simulation process into the fire probability time series sequence of the corresponding grid. The spread risk assessment module is used to determine the wildfire spread risk assessment results based on the ignition probability time series of all grids.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the Bayesian network-based wildfire spread risk assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the wildfire spread risk assessment method based on Bayesian networks as described in any one of claims 1-7.

Images