Forest fire spreading prediction method and system based on physical information neural network

Abstract

The invention relates to the technical field of forest fire spreading prediction, in particular to a forest fire spreading prediction method and system based on a physical information neural network. The method comprises the following steps: acquiring forest fire sample data, and preprocessing the forest fire sample data to obtain a sample set; constructing an artificial neural network model; defining a physical constraint error term according to mass conservation and momentum conservation, defining a data error term, and constructing a loss function according to the physical constraint error term and the data error term; training and testing the model by using the sample set and the loss function; and acquiring real-time data of the forest fire, preprocessing the real-time data of the forest fire, and inputting the data into the model to predict the spreading of the forest fire. According to the method, physical knowledge can be fused into deep learning driven based on historical fire data, and the model is guided by using a physical rule in the learning process, so that the generalization ability of the forest fire spreading prediction model is improved, and the forest fire spreading situation is more reliably and accurately predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire spread prediction, and more specifically, to a forest fire spread prediction method and system based on physical information neural network. Background Art

[0002] With the impact of climate change, the situation of forest fires in my country is severe. In this new era, in order to comprehensively strengthen the prevention and control of forest and grassland fires in my country, the national level actively promotes scientific and technological innovation to empower fire prevention and control. Against this background, remote sensing technology and machine learning are now the mainstream for predicting forest fires. However, since forest fires themselves have the characteristics of high timeliness requirements for predictions and relatively little historical data, this poses many challenges to the use of scientific and technological predictions.

[0003] Although remote sensing technology can quickly identify potential fire spots and monitor large forest areas in real time, it lacks accuracy in multi-dimensional data fusion and synchronization, and the cost of remote sensing equipment is relatively high;

[0004] Although machine learning avoids the shortcomings of multi-data fusion and synchronization, improves the accuracy of prediction to a certain extent, and reduces costs, due to the complexity of the forest environment, it is unable to perceive various interference factors such as light changes, shadows, and occlusions in detail, thus affecting the accuracy of the prediction; and it requires a large number of data samples for training. For forest fires that lack a large amount of historical data, the trained model lacks generalization capabilities.

[0005] Therefore, the use of traditional remote sensing and machine learning methods cannot accurately predict the spread of forest fires, and there is an urgent need to improve the prediction methods. Summary of the invention

[0006] The present invention aims to overcome at least one defect (shortcoming) of the above-mentioned prior art and provide a forest fire spread prediction method and system based on physical information neural network, which is used to more reliably and accurately predict the forest fire spread trend.

[0007] According to a first aspect of the present application, a forest fire spread prediction method based on a physical information neural network is provided, the method comprising:

[0008] Acquire forest fire sample data, and preprocess the forest fire sample data to obtain a sample set;

[0009] Construct artificial neural network models;

[0010] A physical constraint error term is defined according to mass conservation and momentum conservation, and a data error term is defined, and a loss function L is constructed according to the physical constraint error term and the data error term;

[0011] Using the sample set and the loss function L to train and test the artificial neural network model to obtain a trained artificial neural network model;

[0012] Real-time forest fire data is acquired, the real-time forest fire data is preprocessed to obtain preprocessed real-time forest fire data, and the preprocessed real-time forest fire data is input into the trained artificial neural network model to perform forest fire spread prediction to obtain a prediction result.

[0013] It is understandable that the present application continues to use the artificial neural network model to predict the fire line coordinates of forest fires, and cites physical information to establish a loss function to guide and verify the training of the model, which can improve the generalization ability of the model, thereby making the predicted output data more accurate and reliable.

[0014] Optionally, the forest fire sample data includes historical forest fire sample data and / or experimental sample data of simulating forest fire in a laboratory.

[0015] It is understandable that forest fires are extremely destructive natural disasters. Once they occur, if there is no human intervention, they can spread rapidly in a short period of time, and the situation is difficult to control, and they cause serious harm and losses to forest resources, ecosystems, and human life. Therefore, when a forest fire occurs in the early stage, more human intervention is needed as soon as possible to extinguish the fire as quickly as possible and reduce losses as much as possible; at the same time, prevention and control measures for forest fire prevention and disaster prevention will be strengthened to prevent open flames from entering the mountains and igniting, which can also reduce the occurrence of forest fires from another angle. Therefore, the collection of historical data on forest fires is scarce, so only using real forest fire historical data to train the model cannot achieve the training effect. In addition to collecting historical data on forest fires, this application also uses a laboratory to simulate forest fires, specifically simulating various terrains, light, vegetation and other characteristics in real forests, and simulating the occurrence of fires. This method can avoid the destructiveness of real fires, and increase the data volume of training samples to achieve the training effect of the model, improve the learning effect of the model, and make the data predicted by the model more accurate and reliable.

[0016] Optionally, the obtaining of forest fire sample data and preprocessing of the forest fire sample data to obtain a sample set specifically includes:

[0017] Obtaining fire line information in the forest fire sample data, and establishing a two-dimensional coordinate system according to the fire line information, and obtaining position coordinates (x, y) of a number of fire line points at each moment according to the position of the fire line in the two-dimensional coordinate system;

[0018] According to the position coordinates of the same firing point at different times, calculate the actual speed u of the firing point in the X-axis direction xand the actual speed u in the Y-axis direction y , and then according to the u x and the u y Calculate the actual velocity vector u of the firing point obs ;

[0019] The fire line data of each forest fire is taken as a sample, and the fire line data of multiple forest fires constitute a sample set, wherein the fire line data includes the position coordinates (x, y) of the plurality of fire line points and the corresponding time t and the corresponding actual velocity vector u obs .

[0020] It can be understood that, by obtaining each coordinate of the firing point of the firing line data and the corresponding time, the actual velocity vector u of the firing point is calculated. obs Splitting a fire line into multiple fire line points for acquisition helps to accurately predict the spread of the entire fire line; and according to the coordinate position of the same fire line point at different times, the spread speed vector of the fire line point is calculated to provide labels for the training and verification of the following model, thereby improving the efficiency of model training and learning.

[0021] Optionally, the constructing of an artificial neural network model specifically includes:

[0022] The position coordinates (x, y) of each firing point and the corresponding time t, initial atmospheric pressure p 0 as input data of the input layer; and using the output data of the input layer as input data of the hidden layer;

[0023] The hidden layer processes and extracts features from the output data of the input layer and outputs data, and uses the output data of the hidden layer as input data of the output layer;

[0024] The output layer processes the output data of the hidden layer to obtain the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads, and calculates the predicted speed vector u of the fire point and the predicted coordinates (x(t)) of the fire point at the next time T according to the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads. T ,y T ).

[0025] It is understandable that the construction of the artificial neural network model defines the parameters of the input layer, the number of hidden layers and the parameters of the output layer, providing a model framework for the prediction method, making the input and output of the model more standardized and organized.

[0026] Optionally, the physical constraint error term is defined according to the conservation of mass and the conservation of momentum, specifically including:

[0027] Constructing the boundary function F based on the physical information of mass conservation 1 :

[0028]

[0029] Wherein, ρ is the gas density released by the combustion of combustibles in the forest fire; k is the thermal decomposition rate of the combustible particles in the forest fire;

[0030] Physical information boundary function F based on momentum conservation 2 :

[0031]

[0032] Among them, τ ij is the viscous stress tensor of the gas released by the combustion of combustibles in the forest fire, and m is the sum constant of the fixed external forces of conservation of momentum;

[0033] where τ ij Follow the formula:

[0034]

[0035] Among them, S ij is the deformation rate tensor of the combustible material in the forest fire, δ ij A parameter factor of the deformation rate tensor;

[0036] The deformation rate tensor S of the combustible material in the forest fire is ij The formula to follow is:

[0037]

[0038] Among them, the deformation rate tensor parameter factor δ ij The formula to follow is:

[0039]

[0040] Wherein, i is the coordinate x of the training sample at time t, and j is the coordinate y of the training sample at time t;

[0041] Therefore, the physical constraint error term L defined based on mass conservation and momentum conservation is PED for:

[0042] L PED =λ 1 ‖F 1 ‖2 +λ 2 ‖F 2 ‖ 2

[0043] Among them, λ 1 is the error hyperparameter of the mass conservation term, λ 2 is the error hyperparameter of the momentum conservation term.

[0044] It is understandable that the physical constraint error term is constructed based on physical information, in which the relevant principles of fluid mechanics are used to introduce the relevant physical characteristics of forest fires into the calculation. By using the classical model of forest fire dynamics and utilizing the physical changes of combustible particles during the spread of fire, the model is physically constrained, thereby achieving a more reliable and accurate prediction of the spread of forest fires.

[0045] Optionally, the defining of the data error term specifically includes:

[0046] The data error term L data The specific formula is as follows:

[0047] L data =‖uu ovs ‖ 2

[0048] It is understandable that comparing the training predicted velocity vector with the actual velocity vector and calculating the error term therebetween helps to measure the reliability and accuracy of the model training prediction.

[0049] Optionally, constructing a loss function L according to the physical constraint error term and the data error term specifically includes:

[0050] The loss function L is specifically:

[0051] L=λ PED L PED +λ data L data

[0052] Among them, λ PED is the physical information hyperparameter, λ data is the data error hyperparameter.

[0053] It is understandable that the weighted operation of the physical constraint error term and the data error term to construct the loss function can integrate the two dimensions of physical constraints and data error constraints, and provide a more rigorous verification mechanism for the training predictions, which is conducive to eliminating prediction data with large errors and improving the accuracy of the model.

[0054] According to a second aspect of the present application, a forest fire spread prediction system based on a physical information neural network is provided, the system comprising:

[0055] A preprocessing module, used to obtain forest fire sample data, and preprocess the forest fire sample data to obtain a sample set;

[0056] A construction module is used to construct an artificial neural network model; a physical constraint error term is defined according to mass conservation and momentum conservation, and a data error term is defined, and a loss function L is constructed according to the physical constraint error term and the data error term;

[0057] A training-testing module, used to train and test the artificial neural network model using the training samples and the loss function L to obtain a trained artificial neural network model;

[0058] The prediction module is used to obtain real-time forest fire data, preprocess the real-time forest fire data to obtain preprocessed real-time forest fire data, and input the preprocessed real-time forest fire data into the trained artificial neural network model to predict the spread of forest fire and obtain a prediction result.

[0059] According to the third aspect of the present application, an electronic device is provided, comprising a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement a forest fire spread prediction method based on a physical information neural network as described in the first aspect above.

[0060] According to a fourth aspect of the present application, there is provided a computer storage medium on which a computer program is stored. When the computer program is executed, the method for predicting the spread of forest fires based on a physical information neural network as described in the first aspect above is implemented.

[0061] Based on any of the above aspects, the embodiment of the present application provides a method and system for predicting the spread of forest fires based on a physical information neural network, which obtains forest fire data and preprocesses the forest fire data to obtain a sample set; at the same time, an artificial neural network model is constructed and a loss function L is constructed based on momentum and mass conservation; then the artificial neural network model is trained and tested based on the training samples and the loss function to obtain a trained model; finally, real-time forest fire data is obtained and input into the model for prediction to obtain a prediction result. This application focuses on the prediction of forest fire spread, innovatively integrates physical knowledge into deep learning driven by fire history data, and uses physical laws to guide the model during the learning process, thereby improving the generalization ability of the forest fire spread prediction model, especially in the common implementation scenarios such as less historical data, it can be physically constrained by the classical model of forest fire dynamics without basically affecting the deep learning method based on fire history data, so as to more reliably and accurately predict the spread of forest fires. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0063] Figure 1 A flow chart of a forest fire spread prediction method based on physical information neural network provided in this embodiment.

[0064] Figure 2 This is a flow chart of the method for obtaining a sample set provided in this embodiment.

[0065] Figure 3 A schematic diagram of the two-dimensional coordinates of the fire line provided in this embodiment.

[0066] Figure 4 This is a flow chart of the method for building an artificial neural network model provided in this embodiment.

[0067] Figure 5 This is a flow chart of the method for training an artificial neural network model provided in this embodiment.

[0068] Figure 6 This embodiment provides a module diagram of a forest fire spread prediction system based on a physical information neural network.

[0069] Figure 7 A device structure diagram of an electronic device provided in this embodiment. DETAILED DESCRIPTION

[0070] The drawings of this application are only used for illustrative purposes and should not be construed as limiting the present application. In order to better illustrate the following embodiments, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; it is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0071] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0072] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0073] Forest fires are extremely destructive natural disasters. Once they occur, they can spread rapidly in a short period of time without human intervention. The situation is difficult to control and causes serious damage and losses to forest resources, ecosystems and human life. Therefore, predicting forest fires is of great significance and value to forest safety. However, the mainstream use of remote sensing technology and machine learning to predict forest fires lacks accuracy and reliability, and the scarcity of training samples makes the machine learning model lack generalization ability, so it is necessary to improve the traditional prediction method.

[0074] This embodiment provides a technical solution that can solve the above-mentioned problem. The specific implementation methods of this application are described in detail below in conjunction with the accompanying drawings.

[0075] For example, a method for predicting forest fire spread based on a physical information neural network is provided in an embodiment of the present application. Figure 1 As shown, a flowchart of the method is provided, and the method comprises the following steps:

[0076] S110, obtaining forest fire sample data, and preprocessing the forest fire sample data to obtain a sample set;

[0077] Specifically, the forest fire data includes historical sample data of forest fires and / or experimental sample data of forest fires simulated in a laboratory.

[0078] It is understandable that forest fires are extremely destructive natural disasters. Once they occur, if there is no human intervention, they can spread rapidly in a short period of time, and the situation is difficult to control, and it causes serious harm and losses to forest resources, ecosystems, and human life. Therefore, when a forest fire occurs in the early stages, more human intervention is needed as soon as possible to extinguish the fire as quickly as possible and minimize losses; at the same time, prevention and control measures for forest fire prevention and disaster prevention will be strengthened to prevent open flames from entering the mountains and igniting them, which can also reduce the occurrence of forest fires from another perspective. Therefore, the collection of historical data on forest fires is scarce, so training models using only real historical data on forest fires cannot achieve training results. In addition to collecting historical data on forest fires, this application also uses laboratories to simulate forest fires, specifically simulating various terrains, light, vegetation, and other characteristics in real forests, and simulating the occurrence of fires.

[0079] Specifically, the step S110 of obtaining forest fire sample data and preprocessing the forest fire sample data to obtain a sample set can be subdivided into the following steps: Figure 2 As shown, including:

[0080] S111, obtaining fire line information in the forest fire sample data, and establishing a two-dimensional coordinate system according to the fire line information, and obtaining position coordinates (x, y) of a plurality of fire line points at each moment according to the position of the fire line in the two-dimensional coordinate system;

[0081] S112, calculating the actual speed u of the firing point in the X-axis direction according to the position coordinates of the firing point at different times x and the actual speed u in the Y-axis direction y , and then according to the u x and the u y Calculate the actual velocity vector u of the firing point obs ;

[0082] S113, taking the fire line data of each forest fire as a sample, and the fire line data of multiple forest fires constitute a sample set, wherein the fire line data includes the position coordinates (x, y) of the plurality of fire line points and the corresponding time t and the corresponding actual velocity vector u obs .

[0083] It is understandable that the fire spread trajectory can be determined by performing image processing on the actual forest fire spread video and / or the forest fire simulated video in the laboratory. Figure 3 As shown, the fire line of forest fire data is obtained, and a two-dimensional spatial coordinate system is constructed through the position of the flame spreading front (i.e., the farthest point of the fire line) and the position of the root of the fire line. That is, the X-axis is established by the extension line of the straight line where the root of the fire line is located, and the Y-axis is established by the extension line of the vertical straight line between the position of the flame spreading front (i.e., the farthest point of the fire line) and the root of the fire line. The coordinates of the fire point at a certain position of the fire line at a certain time t are obtained according to the two-dimensional coordinates and can be expressed as (x, y).

[0084] For example, suppose that at t 1 The coordinates of a point on the fire line at the moment are (x 1 ,y 1 ), and the fire line point obtained is at t 2 The coordinates of the moment are (x 2 ,y 2 ). According to the values ​​and directions of the two coordinates, the actual speed u of the firing point in the X-axis direction in the current time period can be calculated respectively. x and the actual speed u in the Y-axis direction y .

[0085] Understandably, in t 1 to 2 In the time period, the time change dt is:

[0086] dt=t 2 -t 1

[0087] The displacement of the firing point in the X-axis direction in the time variation of dt is:

[0088] dx=x 2 -x 1

[0089] The displacement of the firing point in the Y-axis direction in the time variation of dt is:

[0090] dy=y 2 -y 1

[0091] Therefore, the actual speed u in the X-axis direction is x for:

[0092]

[0093] Therefore, the actual speed u in the Y-axis direction is y for:

[0094]

[0095] Finally, according to the u x and the u y Calculate the actual velocity vector u of the firing point obs .

[0096] It is understandable that the actual velocity vector u obs Contains the magnitude and direction of velocity, so according to u obs It can accurately reflect the spreading direction and speed of the fire point.

[0097] S120, constructing an artificial neural network model;

[0098] Specifically, the artificial neural network model includes an input layer, a hidden layer, and an output layer;

[0099] It can be understood that the input layer of the artificial neural network model is responsible for receiving and preprocessing data; preferably, in this example, the amount of input layer data is as follows: Figure 2 As shown, in the fire lines obtained from the forest fire data, 1000 fire line points are randomly sampled as random data sample points to complete the input to the input layer;

[0100] The hidden layer is responsible for feature extraction and processing of data; the hidden layer can be set to one or more layers, and each additional layer increases the complexity and expression ability of the network, enabling it to learn more complex features. Preferably, this example is set to 3 layers;

[0101] The output layer is responsible for generating the final prediction or classification result.

[0102] Specifically, Figure 4 Building an artificial neural network model can be broken down into the following steps:

[0103] S121, the position coordinates (x, y) of each firing point and the corresponding time t, initial atmospheric pressure p 0 as input data of the input layer; and using the output data of the input layer as input data of the hidden layer;

[0104] It can be understood that the initial atmospheric pressure p 0 Derived from previous measurements in the laboratory and / or obtained from historical data from previous studies.

[0105] S122, the hidden layer processes and extracts features from the output data of the input layer and outputs data, and uses the output data of the hidden layer as input data of the output layer;

[0106] S123, the output layer processes the output data of the hidden layer to obtain the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads, and calculates the predicted speed vector u of the fire point and the predicted coordinates (x(t)) of the fire point at the next time T according to the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads. T ,y T ).

[0107] S130, defining a physical constraint error term according to mass conservation and momentum conservation, and defining a data error term, and constructing a loss function L according to the physical constraint error term and the data error term;

[0108] Specifically, the physical constraint error term is defined according to the conservation of mass and momentum, and the boundary function F needs to be constructed based on the physical information of the conservation of mass. 1 :

[0109]

[0110] Wherein, ρ is the gas density released by the combustion of combustibles in the forest fire; k is the thermal decomposition rate of the combustible particles in the forest fire;

[0111] It can be understood that the ρ and k are obtained by pre-measurement in the laboratory and / or based on historical data from previous studies.

[0112] Physical information boundary function F based on momentum conservation 2 :

[0113]

[0114] Among them, τ ij is the viscous stress tensor of the gas released by the combustion of combustibles in the forest fire, and m is the sum constant of the fixed external forces of conservation of momentum;

[0115] It is understandable that the m is obtained by measuring in the laboratory in advance and / or based on historical data from previous studies.

[0116] where τ ij The formula to follow is:

[0117]

[0118] Among them, S ij is the deformation rate tensor of the combustible material in the forest fire, δ ijA parameter factor of the deformation rate tensor;

[0119] It is understandable that the δ ij Derived from previous measurements in the laboratory and / or obtained from historical data from previous studies.

[0120] The deformation rate tensor S of the combustible material in the forest fire is ij The formula to follow is:

[0121]

[0122] Among them, the deformation rate tensor parameter factor δ ij The formula to follow is:

[0123]

[0124] Wherein, i is the coordinate x of the training sample at time t, and j is the coordinate y of the training sample at time t;

[0125] Therefore, the physical constraint error term L defined based on mass conservation and momentum conservation is PED for:

[0126] L PED =λ 1 ‖F 1 ‖ 2 +λ 2 ‖F 2 ‖ 2

[0127] Among them, λ 1 is the error hyperparameter of the mass conservation term, λ 2 is the error hyperparameter of the momentum conservation term; preferably, in this example, λ 1 Take 0.5, λ 2 Take 0.5;

[0128] Specifically, the definition of data error terms specifically includes:

[0129] The data error term L data The specific formula is as follows:

[0130] L data =‖uu obs ‖ 2

[0131] Specifically, constructing a loss function L according to the physical constraint error term and the data error term specifically includes:

[0132] The loss function L is specifically:

[0133] L=λ PEDL PED +λ data L data

[0134] Among them, λ PED is the physical information hyperparameter, λ data is a data error hyperparameter; preferably, in this example, λ PED Take 0.5, λ data Take 0.5.

[0135] S140, training and testing the artificial neural network model using the sample set and the loss function L to obtain a trained artificial neural network model;

[0136] Specifically, step S140 of training the artificial neural network model can be divided into the following steps: Figure 5 As shown:

[0137] S141, input the sample set into the artificial neural network model for training, the model obtains the parameter prediction values ​​u(x, t), u(y, t) and p of the output layer, and performs various automatic differentiation operations on the parameter prediction values ​​of the output layer;

[0138] S142. Define a reliable loss item threshold;

[0139] Preferably, in this example, the reliable loss term threshold is set to 10 -6 ;

[0140] S143. Obtain the numerical value required by the loss function according to the predicted value, and calculate the actual loss term according to the loss function, and compare the actual loss term with the reliable loss term threshold for verification. If the actual loss term is less than the reliable loss term, the predicted value is accurate and reliable prediction data.

[0141] It is understandable that in this example, if the actual loss term is less than 10 -6 , the output is valid data, otherwise, the data is discarded and the training process for the next set of data is carried out again.

[0142] S150, obtaining real-time forest fire data, preprocessing the real-time forest fire data to obtain preprocessed real-time forest fire data, and inputting the preprocessed real-time forest fire data into the trained artificial neural network model to perform forest fire spread prediction to obtain a prediction result.

[0143] like Figure 6 As shown, the embodiment of the present application also provides a forest fire spread prediction system based on physical information neural network. Optionally, the system may include:

[0144] Preprocessing module 311, construction module 312, training-testing module 313, prediction module 314. Among them:

[0145] The preprocessing module 311 is used to obtain forest fire sample data and preprocess the forest fire sample data to obtain a sample set;

[0146] In this embodiment, the preprocessing module 311 can be used to perform Figure 1 As shown in step S110 , for a detailed description of the preprocessing module 311 , reference may be made to the description of step S110 .

[0147] The construction module 312 is used to construct an artificial neural network model; define a physical constraint error term according to mass conservation and momentum conservation, and define a data error term, and construct a loss function L according to the physical constraint error term and the data error term;

[0148] In this embodiment, the construction module 312 can be used to execute Figure 1 As shown in step S120 and step S130, for the specific description of the building module 312, reference may be made to the description of the step S120 and step S130.

[0149] The training-testing module 313 is used to train and test the artificial neural network model using the training samples and the loss function L to obtain a trained artificial neural network model;

[0150] In this embodiment, the training-testing module 313 can be used to perform Figure 1 As shown in step S140, for the detailed description of the training-testing module 313, reference may be made to the description of step S140.

[0151] The prediction module 314 is used to obtain forest fire real-time data, preprocess the forest fire real-time data to obtain preprocessed forest fire real-time data, and input the preprocessed forest fire real-time data into the trained artificial neural network model to perform forest fire spread prediction to obtain a prediction result;

[0152] In this embodiment, the prediction module 314 can be used to perform Figure 1 As shown in step S150 , for a detailed description of the prediction module 314 , reference may be made to the description of step S150 .

[0153] Based on the same inventive concept, this embodiment also provides an electronic device, Figure 7The structural diagram of the electronic device of this embodiment is shown, including a memory 21 and a processor 22, wherein the memory 21 stores computer-readable instructions, and the processor 22 executes the computer-readable instructions to implement the forest fire spread prediction method based on physical information neural network of this embodiment.

[0154] Preferably, the electronic device further includes a bus 23 and a communication interface 24 , and the processor 22 , the communication interface 24 and the memory 21 are connected via the bus 23 .

[0155] The memory 21 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 24 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used. The bus 23 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus 23 may be divided into an address bus, a data bus, a control bus, etc. (not fully drawn in the figure).

[0156] The processor 22 may be an integrated circuit chip with signal processing capability. In the specific implementation process, each step in the above method embodiment may be completed by the hardware integrated logic circuit or software instructions in the processor 22. The above processor 22 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, which may implement or execute the methods, steps, and logic diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor 22 may also be any conventional processor 22, etc. The steps of the method disclosed in conjunction with the embodiments of the present invention may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 21, and the processor 22 reads the information in the memory 21 and completes the steps of the method of the above embodiment in combination with its hardware.

[0157] An embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by the processor 22, the computer-executable instructions prompt the processor 22 to implement the above-mentioned optical cable buried depth detection method. The specific implementation can be found in the embodiment, which will not be repeated here.

[0158] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0159] It can be understood that the above-mentioned system embodiment and the above-mentioned method embodiment can correspond to each other, and similar descriptions of the system embodiment can refer to the method embodiment. To avoid repetition, it will not be repeated here. A forest fire spread prediction system based on a physical information neural network provided in an embodiment of the present application can execute a forest fire spread prediction method based on a physical information neural network provided in any embodiment of the present application, and has functional modules and beneficial effects corresponding to the execution method. The functional modules of the forest fire spread prediction system based on a physical information neural network can be implemented in the form of hardware, can be implemented by instructions in the form of software, and can also be implemented by a combination of hardware and software modules.

[0160] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solution of the present invention, and are not intended to limit the specific implementation methods of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the claims of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A forest fire spread prediction method based on physical information neural network, characterized in that: The method comprises: Acquire forest fire sample data, and preprocess the forest fire sample data to obtain a sample set; Construct artificial neural network models; A physical constraint error term is defined according to mass conservation and momentum conservation, and a data error term is defined, and a loss function L is constructed according to the physical constraint error term and the data error term; Using the sample set and the loss function L to train and test the artificial neural network model to obtain a trained artificial neural network model; Real-time forest fire data is obtained, the real-time forest fire data is preprocessed to obtain preprocessed real-time forest fire data, and the preprocessed real-time forest fire data is input into the trained artificial neural network model to predict the spread of forest fire and obtain a prediction result.

2. The method for predicting forest fire spread based on physical information neural network according to claim 1, characterized in that: The forest fire sample data includes forest fire historical sample data and / or experimental sample data of simulating forest fire in a laboratory.

3. The method for predicting forest fire spread based on physical information neural network according to claim 1, characterized in that: The obtaining of forest fire sample data and preprocessing of the forest fire sample data to obtain a sample set specifically includes: Obtaining fire line information in the forest fire sample data, and establishing a two-dimensional coordinate system according to the fire line information, and obtaining position coordinates (x, y) of a number of fire line points at each moment according to the position of the fire line in the two-dimensional coordinate system; According to the position coordinates of the same firing point at different times, calculate the actual speed u of the firing point in the X-axis direction x and the actual speed u in the Y-axis direction y , and then according to the u x and the u y Calculate the actual velocity vector u of the firing point obs ; The fire line data of each forest fire is taken as a sample, and the fire line data of multiple forest fires constitute a sample set, wherein the fire line data includes the position coordinates (x, y) of the plurality of fire line points and the corresponding time t and the corresponding actual velocity vector u obs .

4. The method for predicting forest fire spread based on physical information neural network according to claim 1, characterized in that: The construction of the artificial neural network model specifically includes: The artificial neural network model includes an input layer, a hidden layer, and an output layer; The position coordinates (x, y) of each firing point and the corresponding time t and the initial atmospheric pressure p0 are used as input data of the input layer; and the output data of the input layer is used as input data of the hidden layer; The hidden layer processes and extracts features from the output data of the input layer and outputs data, and uses the output data of the hidden layer as input data of the output layer; The output layer processes the output data of the hidden layer to obtain the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads, and calculates the predicted speed vector u of the fire point and the predicted coordinates (x(t)) of the fire point at the next time T according to the predicted speed u(x, t) of the fire point in the X-axis direction at time t, the predicted speed u(y, t) of the fire point in the Y-axis direction at time t, and the predicted dynamic pressure p of the fire point when it spreads. T ,y T ).

5. The method for predicting forest fire spread based on physical information neural network according to claim 4, characterized in that: The physical constraint error term is defined according to mass conservation and momentum conservation, specifically including: Construct the boundary function F1 based on the physical information of mass conservation: Wherein, ρ is the gas density released by the combustion of combustibles in the forest fire; k is the thermal decomposition rate of the combustible particles in the forest fire; Physical information boundary function F2 based on momentum conservation: Among them, τ ij is the viscous stress tensor of the gas released by the combustion of combustibles in the forest fire, and m is the sum constant of the fixed external forces of conservation of momentum; where τ ij The formula to follow is: Among them, S ij is the deformation rate tensor of the combustible material in the forest fire, δ ij A parameter factor of the deformation rate tensor; The deformation rate tensor S of the combustible material in the forest fire is ij The formula to follow is: Among them, the deformation rate tensor parameter factor δ ij The formula to follow is: Wherein, i is the coordinate x of the training sample at time t, and j is the coordinate y of the training sample at time t; Therefore, the physical constraint error term L defined based on mass conservation and momentum conservation is PED for: L PED =λ1‖F1‖ 2 +λ2‖F2‖ 2 Among them, λ1 is the error hyperparameter of the mass conservation term, and λ2 is the error hyperparameter of the momentum conservation term.

6. The method for predicting forest fire spread based on physical information neural network according to claim 1, characterized in that: The defined data error term specifically includes: The data error term L data The specific formula is as follows: L data =‖uu obs ‖ 2 7. A method for predicting forest fire spread based on physical information neural network according to any one of claims 5 and 6, characterized in that: Constructing a loss function L according to the physical constraint error term and the data error term specifically includes: The loss function L is specifically: L=λ PED L PED +λ data L data Among them, λ PED is the physical information hyperparameter, λ data is the data error hyperparameter.

8. A forest fire spread prediction system based on physical information neural network, characterized in that: The system comprises: A preprocessing module, used to obtain forest fire sample data, and preprocess the forest fire sample data to obtain a sample set; A construction module is used to construct an artificial neural network model; a physical constraint error term is defined according to mass conservation and momentum conservation, and a data error term is defined, and a loss function L is constructed according to the physical constraint error term and the data error term; A training-testing module, used to train and test the artificial neural network model using the training samples and the loss function L to obtain a trained artificial neural network model; The prediction module is used to obtain real-time forest fire data, preprocess the real-time forest fire data to obtain preprocessed real-time forest fire data, and input the preprocessed real-time forest fire data into the trained artificial neural network model to predict the spread of forest fire and obtain a prediction result.

9. An electronic device, comprising a memory and a processor, characterized in that: The memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement a forest fire spread prediction method based on a physical information neural network as described in any one of claims 1 to 7.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for predicting the spread of forest fires based on a physical information neural network as described in any one of claims 1 to 7 is implemented.

Images