Neural network-based lightning stroke forest fire numerical simulation method, system and equipment and medium

Abstract

The invention belongs to the technical field of numerical simulation, and discloses a lightning stroke forest fire numerical simulation method, system and device based on a neural network and a medium, and the method comprises the steps: constructing a lightning stroke forest fire numerical simulation model based on an observation physical quantity, constructing a full-connection neural network, and enabling the input of the full-connection neural network to be coordinate points, the output is vector representation of the electric field and the magnetic field of the corresponding coordinate point; the coordinate points are space coordinates and time coordinates; sampling a group of training coordinate point data in the domain of definition; inputting the sampled coordinate data into the full-connection neural network for classification prediction, and training according to a target loss function to obtain a trained full-connection neural network; and predicting the distribution condition of the electric field and the magnetic field at any coordinate point in the domain of definition based on the trained full-connection neural network. According to the technical scheme, numerical simulation can be efficiently and accurately achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of numerical simulation, and particularly relates to a method, system, device and medium for numerically simulating lightning-caused forest fires based on a neural network. Background Art

[0002] Lightning strikes are one of the important causes of forest fires. Accurately simulating and predicting the occurrence and development of lightning-caused forest fires is of great significance for forest resource protection and fire prevention and control.

[0003] Traditional numerical simulation methods based on physical equations (such as the finite difference method, the finite element method, etc.) have some difficulties in dealing with complex problems such as lightning-caused forest fires. They have high computational complexity and low efficiency in dealing with complex boundary conditions and high-dimensional problems, and it is difficult to accurately satisfy physical constraints. The progress of deep learning technology has provided a new perspective for solving partial differential equations. The neural network-based solution method can directly approximate the solution of the equation under meshless conditions by embedding the physical equation into the loss function and using automatic differentiation to calculate the derivative. This method not only overcomes the dependence of traditional methods on mesh generation, but also shows significant advantages in high-dimensional problems, complex boundary handling, and multi-physics field coupling. The industrial community has begun to explore its application potential. For example, in the aerospace field, by using a neural network to solve complex fluid mechanics equations to simulate the airflow around an aircraft, the aerodynamic characteristics can be quickly and accurately predicted to help optimize the aircraft's shape design and improve flight performance and fuel efficiency; in the mechanical energy field, neural networks can be used to solve combustion equations and fluid mechanics equations to optimize the combustion process and intake and exhaust systems of engines, and quickly predict the performance parameters of engines, such as power and fuel consumption rate, to help optimize the engine design and improve the engine's efficiency and reliability; in electronic engineering, neural networks can be used to solve semiconductor physics equations to simulate the electrical characteristics and performance of devices, quickly predict the current-voltage characteristics of devices, etc., to help optimize the structural design and process parameters of devices and improve the performance and reliability of devices. In view of the above background, the present invention applies a neural network to solve Maxwell's equations to efficiently and accurately achieve the numerical simulation of lightning-caused forest fires. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for numerically simulating lightning-caused forest fires based on a neural network to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a method for numerically simulating lightning-caused forest fires based on a neural network, including:

[0006] Constructing a numerical simulation model of lightning-caused forest fires based on observed physical quantities, where the numerical simulation model is Maxwell's equation in the form of a first-order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition;

[0007] Construct a fully-connected neural network, where the input of the fully-connected neural network is a coordinate point, and the output is the vector representation of the electric field and magnetic field of the corresponding coordinate point; the coordinate point includes spatial coordinates and time coordinates.

[0008] Sample a set of training coordinate point data within the domain of definition.

[0009] Input the sampled coordinate data into the fully-connected neural network for classification prediction, and train according to the target loss function to obtain a trained fully-connected neural network.

[0010] Based on the trained fully-connected neural network, predict the distribution of the electric field and magnetic field at any coordinate point within the domain of definition.

[0011] Optionally, the physical quantities include the electric field and magnetic field.

[0012] Optionally, the numerical simulation model is specifically:

[0013]

[0014] In the formula, E is the electric field, B is the magnetic field, ρ is the charge density, ε0 is the vacuum permittivity, J is the current density, μ0 is the magnetic permeability, is the divergence, is the curl, and t is the time coordinate.

[0015] Optionally, the training process of the fully-connected neural network specifically includes:

[0016] Based on the quasi-random sampling method, sample a set of training coordinate point data within the domain of definition, input the training coordinate point data into the fully-connected neural network, calculate the corresponding spatial partial derivative and time partial derivative of the output of the fully-connected neural network, and substitute the spatial partial derivative and time partial derivative into Maxwell's equations to calculate the residuals of each component.

[0017] Calculate the mean square error between the sampling points of the initial conditions and periodic boundary conditions and the output of the fully-connected neural network, and construct a target loss function based on the calculated residuals and mean square error.

[0018] Based on the gradient descent optimization algorithm, minimize the target loss function, and in the optimization process, iteratively adjust the network parameters to make the residuals approach zero, and complete the training of the fully-connected neural network.

[0019] Optionally, after the training of the fully-connected neural network is completed, verification is carried out, specifically including:

[0020] After the training is completed, select test coordinate points within the domain of definition.

[0021] Input the test points into the trained fully-connected neural network to obtain predicted values.

[0022] The spatiotemporal distribution of the predicted values is displayed by visualization means, and the predicted values are compared with the analytical solution or numerical solution to evaluate the accuracy and physical accuracy.

[0023] A numerical simulation system for lightning-caused forest fires based on neural networks, comprising:

[0024] A numerical model construction module for constructing a numerical simulation model of lightning-caused forest fires based on observed physical quantities, where the numerical simulation model is a Maxwell equation in the form of a first-order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition;

[0025] A prediction model training module for constructing a fully connected neural network, where the input of the fully connected neural network is a coordinate point, and the output is a vector representation of the electric field and magnetic field at the corresponding coordinate point; the spatial and temporal coordinates of the coordinate point; sampling a set of training coordinate point data within the domain of definition; inputting the sampled coordinate data into the fully connected neural network for classification prediction, and training according to the target loss function to obtain a trained fully connected neural network;

[0026] A lightning-caused forest fire numerical simulation module for predicting the distribution of the electric field and magnetic field at any coordinate point within the domain of definition according to the trained fully connected neural network.

[0027] An electronic device, comprising a memory and a processor, where the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the described numerical simulation method for lightning-caused forest fires based on neural networks.

[0028] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the described numerical simulation method for lightning-caused forest fires based on neural networks.

[0029] The technical effects of the present invention are as follows:

[0030] The present invention constructs a numerical simulation model of lightning-caused forest fires based on observed physical quantities, and the mathematical model is the Maxwell equation constructed on the domain of periodic boundary conditions; uses a physics-informed neural network to solve the partial differential equation, and uses the mean square error function to calculate the residual loss; uses automatic differentiation to replace the construction of differences in traditional numerical methods; the simulation results characterize the convergence trend of the numerical solution of the mathematical model; this embodiment can efficiently and accurately achieve numerical simulation. Description of the Drawings

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0032] The accompanying drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the accompanying drawings:

[0033] Figure 1 is the implementation flowchart in the embodiments of the present invention;

[0034] Figure 2 is the structural diagram of the neural network in the embodiments of the present invention;

[0035] Figure 3 is the change diagram of the loss function during the process of training the neural network in the embodiments of the present invention;

[0036] Figure 4 is the comparison diagram of the solution obtained by applying the neural network to solve the equation and the standard solution in the embodiments of the present invention;

[0037] Figure 5 is the schematic diagram of the mean square error between the predicted solution and the standard solution in the embodiments of the present invention. Detailed implementation manners

[0038] Now, various exemplary implementation manners of the present invention will be described in detail. This detailed description should not be considered as a limitation to the present invention, but should be understood as a more detailed description of certain aspects, characteristics, and implementation schemes of the present invention.

[0039] It should be understood that the terms described in the present invention are only for describing specific implementation manners and are not used to limit the present invention. Additionally, for the numerical ranges in the present invention, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Each intermediate value within any stated value or stated range, as well as each smaller range between any other stated value or intermediate value within the stated range, is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded from the range.

[0040] Without departing from the scope or spirit of the present invention, various improvements and changes can be made to the specific implementation manners of the specification of the present invention, which are obvious to those skilled in the art. Other implementation manners obtained from the specification of the present invention are obvious to those skilled in the art. The specification and embodiments of this application are only exemplary.

[0041] As used herein, terms such as "comprising", "including", "having", "containing", etc. are all open-ended terms, meaning including but not limited to.

[0042] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will detail this application with reference to the accompanying drawings and in conjunction with the embodiments.

[0043] As Figure 1 shown, in this embodiment, a numerical simulation method of lightning-struck forest fires based on a neural network is provided, including: constructing a numerical simulation model of lightning-struck forest fires based on observed physical quantities, where the numerical simulation model is a Maxwell equation in the form of a first-order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition; constructing a fully-connected neural network, where the input of the fully-connected neural network is a coordinate point and the output is a vector representation of the electric field and magnetic field at the corresponding coordinate point; the spatial coordinates and time coordinates of the coordinate point; sampling a set of training coordinate point data within the domain of definition; inputting the sampled coordinate data into the fully-connected neural network for classification prediction and training according to the target loss function to obtain a trained fully-connected neural network; predicting the distribution of the electric field and magnetic field at any coordinate point within the domain of definition based on the trained fully-connected neural network.

[0044] This embodiment provides a numerical simulation method of lightning-struck forest fires based on an ARC-PINN physics-informed neural network, which relates to the field of numerical simulation; the method includes: constructing a numerical simulation model of lightning-struck forest fires based on observed physical quantities; the mathematical model is a Maxwell equation constructed on the domain of definition with periodic boundary conditions; using a physics-informed neural network to solve the partial differential equation and calculating the residual loss using the mean square error function; using automatic differentiation to replace the construction of differences in traditional numerical methods; the simulation results characterize the convergence trend of the numerical solution of the mathematical model; this embodiment can efficiently and accurately achieve numerical simulation.

[0045] The purpose of this embodiment is to provide a numerical simulation method of lightning-struck forest fires based on an ARC-PINN physics-informed neural network, which can efficiently achieve numerical simulation.

[0046] To achieve the above purpose, this embodiment provides the following solutions:

[0047] Define the problem and the physical model: The model is a Maxwell equation for periodic boundary conditions constructed based on observed physical quantities related to lightning-struck forest fires.

[0048] The Maxwell equation specifically includes:

[0049]

[0050] where \(E\) is the electric field, \(B\) is the magnetic field, \(\rho\) is the charge density, \(J\) is the current density, \(\varepsilon_0\) is the vacuum permittivity, and \(\mu_0\) is the magnetic permeability. The Maxwell equations involve vector fields \((E, B)\) and use vector operations (divergence curl ), which makes them more complex than scalar partial differential equations. The electric field \(E\) is the function to be solved for in the equations.

[0051] The initial conditions are \(E\) x (x,y,0)=\(\sin(\pi x)\), \(E\) y (x,y,0)=0, \(B\) z (x,y,0),

[0052] The periodic boundary condition is \(u(0,y,t)=u(1,y,t)\).

[0053] Sampling collocation points: The method of sampling collocation points is to generate random points within the domain of definition, which is specifically implemented through the torch.rand function in PyTorch. This method belongs to uniform random sampling.

[0054] The number of PDE residual points \((xyt\_pde)\) is \(N\) p =10000, and the domain of definition is \((\Omega = [0,1]\times[0,1]\times[0,0.5])\) (corresponding to \(x\), \(y\), \(t\)). The specific process is to use the torch.rand function to generate a tensor of \(10000\times3\), where each element is uniformly randomly distributed in the interval \([0,1)\), and then multiply the tensor by torch.tensor([1.0,1.0,0.5]) to scale the time dimension to \([0,0.5]\) while keeping \(x\), \(y\) in \([0,1]\). Finally, sampling points uniformly distributed throughout the three-dimensional space-time are obtained.

[0055] Constructing the neural network: As Figure 2 shown, the ARC_PINN neural network constructs a neural network that is a multi-layer perceptron (MLP) containing 1 input layer, 3 hidden layers, and 1 output layer.

[0056] Among them, the input layer is a linear layer with an input dimension of 3. The corresponding inputs are the spatial coordinates \(x\), \(y\) and the time coordinate \(t\), that is, the spatial and time dimensions in a two-dimensional problem. The output dimension is 50, which maps the input 3D features to a 50D feature space, providing more expressive power for subsequent non-linear transformations.

[0057] The hidden layer consists of 3 layers. Each hidden layer is composed of a fully connected layer and a Tanh activation function layer. Each hidden layer has 50 neurons. Each fully connected layer recombines the 50-dimensional features of the previous layer to form a new feature representation. This linear mapping allows the network to explore different linear combinations in the input space. The non-linear activation function layer introduces the non-linear activation function Tanh, enabling the network to capture non-linear patterns (such as the fluctuating sine shape) in the solutions of Maxwell's equations. The three hidden layers increase the depth of the network. Deep networks have stronger expressive power than single-layer networks and can better fit the high-order features of partial differential equations. Here, the deep neural network gradually constructs the complex E from the spatio-temporal coordinates (x, y, t). x , E y , B z distribution.

[0058] The output layer has 3 neurons, which map the high-dimensional features (50-dimensional) extracted by the hidden layer back to the electromagnetic field components (3-dimensional) in physical space, representing E x , E y , B z , which directly corresponds to the variables in Maxwell's equations and is used to calculate the physical residual and satisfy the initial and boundary conditions.

[0059] Construct the loss function: When using the ARC-PINN deep neural network to solve Maxwell's equations, the construction of the loss function is the core part, which incorporates physical constraints, initial conditions, and boundary conditions into the training process of the neural network. The training objective of the neural network is to minimize the loss function. This loss function consists of multiple parts:

[0060] Construct the total loss function in the form of L = L PDE + λ1L IC + λ2L BC , where L PDE is the mean square sum of the equation residuals, and L IC and L BC are the mean square errors of the initial and boundary conditions, and λ1 and λ2 are weight factors.

[0061] PDE residual loss: This part of the loss calculates the difference between the predicted solution of the model and the actual physical equation, enabling the physical quantities output by the neural network to satisfy the dynamic relationship of Maxwell's equations. The specific PDE residual loss is as follows,

[0062]

[0063] where N p = 10000 is the number of sampling points.

[0064] Initial condition residual loss: Ensure that the network output at (t = 0) conforms to the physical initial state. The initial condition is E x (x, y, 0) = sin(πx), E y (x, y, 0) = 0, B z (x, y, 0). The specific initial condition residual loss is as follows,

[0065]

[0066] where N ic = 1000 is the number of sampling points.

[0067] Boundary condition loss: This part of the loss calculates the error of the boundary conditions. The network needs to satisfy the predefined boundary conditions, that is, at the given boundaries (x = 0) and (x = 1), the outputs of the neural network should be consistent, reflecting the periodic physical constraints. The periodic boundary condition is u(0, y, t) = u(1, y, t). The specific boundary condition loss is as follows,

[0068]

[0069] where N bc = 1000 is the number of sampling points.

[0070] The total loss function is Loss = Loss PDE + Loss ic + Loss bc .

[0071] Automatic differentiation for calculating derivatives: Automatic differentiation technology is the core technology for calculating residuals in solving Maxwell's equations using the ARC-PINN deep neural network. It is an efficient and accurate derivative calculation method, widely used in deep learning and scientific computing, especially for embedding physical constraints in physics-informed neural networks.

[0072] Implemented through the torch.autograd.grad function in PyTorch, using the reverse mode, to calculate the partial derivatives of the neural network outputs E x , E y , E y with respect to the inputs x, y, t. The calculation process is based on the chain rule and computational graph, ensuring accuracy and efficiency, and directly supporting the construction and optimization of Maxwell's equation residuals.

[0073] Training the Neural Network: In the process of using the ARC-PINN neural network to solve Maxwell's equations, the process of training the neural network is an iterative process based on gradient optimization. The aim is to adjust the network parameters by minimizing the loss function so that the output satisfies the physical equations, initial conditions, and boundary conditions. The training objective is to make the loss function Loss = Loss PDE +Loss ic +Loss bc as close to zero as possible. The loss function consists of three parts: PDE residual, initial condition residual, and boundary condition residual. During the training process, the Adam optimizer is used to optimize the network parameters. The Adam (Adaptive Moment Estimation) optimizer is a commonly used optimization algorithm that can adaptively adjust the learning rate of each parameter, thus accelerating convergence. The specific steps include:

[0074] Initializing the network and the optimizer, using the Adam optimizer; setting the learning rate lr = 0.001 to control the parameter update step size; generating sampling points using the random sampling method; training loop (with 100,000 iterations); updating the parameters; saving the network.

[0075] As Figure 3 shown, as the number of training iterations increases, the value of the loss function shows a downward trend.

[0076] Solving the Equation and Validation:

[0077] After the network training is completed, predictions can be made for new input data (i.e., given new spatial and temporal coordinates, predicting the corresponding electric field). By comparing the prediction results of the neural network with the true solution, the performance of the model can be evaluated.

[0078] The data is plotted using the Matplotlib library to achieve the visual numerical simulation of the physical field distribution. As Figure 4 shown, the solution obtained using the ARC-PINN deep neural network is close to the standard solution with a small error. As Figure 5 shown, by calculating the mean square error between the predicted solution and the standard solution, the accuracy is 91.49%. The numerical simulation experiment shows that the numerical solution of this algorithm has stable convergence and high precision.

[0079] The present invention provides a high-precision numerical simulation algorithm constructed based on observing physical quantities related to lightning-struck forest fires; for Maxwell's equations with periodic boundary conditions, using automatic differentiation technology to replace traditional numerical algorithms, and using a physics-informed neural network to numerically solve the equations, calculating the residual loss through the mean square error function; the numerical simulation experiment shows that the numerical solution of this algorithm has stable convergence and high precision.

[0080] Construct an ARC-PINN neural network to analyze the mathematical model. Satisfy the physical constraints through loss function embedding, use the automatic differentiation of the neural network to calculate partial derivatives for evaluating the equation residuals, and adopt the gradient descent optimization algorithm to make the residuals tend to zero to obtain the analysis result. After training, select test points within Ω, input them into the network to obtain predicted values, visualize the spatio-temporal distribution of the solution, and compare it with the analytical solution or numerical solution to evaluate the accuracy and physical accuracy.

[0081] Implementable, this embodiment also provides a lightning-induced forest fire numerical simulation system based on a neural network, including:

[0082] A numerical model construction module for constructing a numerical simulation model of lightning-induced forest fire based on observed physical quantities. The numerical simulation model is a Maxwell equation in the form of a first-order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition.

[0083] A prediction model training module for constructing a fully connected neural network. The input of the fully connected neural network is the coordinate points, and the output is the vector representation of the electric field and magnetic field at the corresponding coordinate points; the spatial coordinates and time coordinates of the coordinate points; sample a set of training coordinate point data within the domain of definition; input the sampled coordinate data into the fully connected neural network for classification prediction, and train it according to the target loss function to obtain a trained fully connected neural network.

[0084] A lightning-induced forest fire numerical simulation module for predicting the distribution of the electric field and magnetic field at any coordinate point within the domain of definition according to the trained fully connected neural network.

[0085] Implementable, this embodiment also provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned lightning-induced forest fire numerical simulation method based on a neural network.

[0086] Implementable, this embodiment also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned lightning-induced forest fire numerical simulation method based on a neural network.

[0087] As mentioned above, only the specific implementation manners of this application that are better are described. However, the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in this application should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A numerical simulation method for lightning-strike forest fires based on a neural network, characterized in that, including: Construct a numerical simulation model of lightning - induced forest fires based on observed physical quantities. The numerical simulation model is the Maxwell's equation in the form of a first - order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition; Construct a fully - connected neural network. The input of the fully - connected neural network is the coordinate point, and the output is the vector representation of the electric field and magnetic field corresponding to the coordinate point; the spatial and temporal coordinates of the coordinate point; Sample a set of training coordinate point data within the domain of definition; Input the sampled coordinate data into the fully - connected neural network for classification prediction, and train it according to the target loss function to obtain a trained fully - connected neural network; Predict the distribution of the electric field and magnetic field at any coordinate point within the domain of definition based on the trained fully - connected neural network.

2. The method for numerical simulation of lightning-caused forest fires based on neural network according to claim 1, characterized in that The physical quantities include the electric field and magnetic field.

3. A numerical simulation method of lightning-induced forest fires based on a neural network according to claim 1, characterized in that, The numerical simulation model is specifically: where E is the electric field, B is the magnetic field, ρ is the charge density, ε0 is the permittivity of free space, J is the current density, μ0 is the permeability, is the divergence, × is the curl, and t is the time coordinate.

4. A numerical simulation method of lightning-induced forest fires based on a neural network according to claim 1, characterized in that, The training process of the fully - connected neural network specifically includes: Based on the quasi - random sampling method, sample a set of training coordinate point data within the domain of definition. Input the training coordinate point data into the fully - connected neural network, calculate the spatial partial derivative and temporal partial derivative corresponding to the output of the fully - connected neural network, substitute the spatial partial derivative and temporal partial derivative into the Maxwell's equation, and calculate the residuals of each component; Calculate the mean square error between the sampled points of the initial conditions and periodic boundary conditions and the output of the fully - connected neural network, and construct a target loss function based on the calculated residuals and mean square error; Based on the gradient - descent optimization algorithm, minimize the target loss function. The optimization process iteratively adjusts the network parameters to make the residuals approach zero and complete the training of the fully - connected neural network.

5. A numerical simulation method for lightning-induced forest fires based on a neural network according to claim 1, characterized in that After the training of the fully - connected neural network is completed, verification is carried out, specifically including: After training is completed, select test coordinate points within the domain of definition; Input the test points into the trained fully - connected neural network to obtain predicted values; Display the spatio - temporal distribution of the predicted values through visualization means, compare the predicted values with the analytical solution or numerical solution, and evaluate the accuracy and physical accuracy.

6. A lightning strike fire numerical simulation system based on a neural network, characterized in that, including: A numerical model construction module for constructing a numerical simulation model of lightning - induced forest fires based on observed physical quantities. The numerical simulation model is the Maxwell's equation in the form of a first - order partial differential equation constructed based on periodic boundary conditions, initial conditions, and physical quantities within the domain of definition; A prediction model training module for constructing a fully - connected neural network. The input of the fully - connected neural network is the coordinate point, and the output is the vector representation of the electric field and magnetic field corresponding to the coordinate point; the spatial and temporal coordinates of the coordinate point; sample a set of training coordinate point data within the domain of definition; input the sampled coordinate data into the fully - connected neural network for classification prediction, and train it according to the target loss function to obtain a trained fully - connected neural network; A lightning - induced forest fire numerical simulation module for predicting the distribution of the electric field and magnetic field at any coordinate point within the domain of definition according to the trained fully - connected neural network.

7. An electronic device, characterized in that, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a method for numerical simulation of lightning - induced forest fires based on a neural network according to any one of claims 1 - 5.

8. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by a processor, it implements a neural network-based numerical simulation method for lightning-caused forest fires as described in any one of claims 1-5.

Images