Risk assessment method and system for forest fire induced by tree line discharge based on big data

Abstract

The invention discloses a big data-based tree line discharge induced forest fire risk assessment method and system, and the method comprises the steps: constructing a neural network-based tree line discharge risk assessment model which is input as a feature parameter sequence and output as a risk level; training is carried out by adopting a training sample set extracted from environmental parameter historical data before a forest fire event is induced by each time of tree line discharge, and when the risk of forest fire induced by tree line discharge in a certain target area needs to be evaluated, environmental parameter data of the target area is collected to form a to-be-evaluated characteristic parameter sequence; and inputting a tree line discharge risk assessment model to obtain a risk level of forest fire induced by tree line discharge. According to the method, historical and real-time environmental data are collected, risk assessment is performed by using the neural network model, and early warning and effective management of forest fire risks are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system safety monitoring, and more specifically, relates to a risk assessment method and system for tree-line discharge-induced wildfires based on big data. Background Art

[0002] With the rapid development of power systems, transmission lines are increasingly distributed in the natural environment. However, there are many different types of trees in the natural environment, and the interaction between different tree species and transmission lines may lead to treeline discharge, which in turn triggers wildfires. Traditional monitoring methods rely on manual inspections and simple sensors, which cannot achieve real-time monitoring and accurate prediction of wildfire risks. Therefore, it is of great practical significance to develop a method for predicting and assessing the risk of wildfires induced by treeline discharge based on big data. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a risk assessment method and system for tree line discharge-induced wildfires based on big data. By collecting historical and real-time environmental data and using a neural network model to perform risk assessment, early warning and effective management of wildfire risks can be achieved.

[0004] In order to achieve the above-mentioned purpose of the invention, the risk assessment method of tree line discharge induced wildfire based on big data of the present invention comprises the following steps:

[0005] S1: According to actual needs, several historical events of treeline discharge-induced wildfires are collected, and the characteristic parameter historical data before each treeline discharge-induced wildfire event is obtained, and several characteristic parameter sequences are extracted from them. Each characteristic parameter sequence includes characteristic parameter data of T consecutive moments, and each characteristic parameter data includes meteorological parameters, treeline space parameters, geographical environment parameters and vegetation parameters at that moment;

[0006] Each characteristic parameter sequence is recorded as Where N represents the number of characteristic parameters, and each characteristic parameter sequence X is labeled with a risk level to obtain a risk level label y, y = 1, 2, ..., K, K represents the number of preset risk levels, and the maximum value of label y indicates a higher risk of inducing wildfires;

[0007] S2: Perform data standardization on each feature parameter sequence X;

[0008] S3: Take the feature parameter sequence after data standardization as input and the corresponding risk level as the expected output to construct training samples, thus obtaining the training sample set;

[0009] S4: construct a treeline discharge risk assessment model, whose input is a characteristic parameter sequence and output is a risk level;

[0010] S5: using the training sample set obtained in step S3 to train the treeline discharge risk assessment model to obtain a trained treeline discharge risk assessment model;

[0011] S6: When it is necessary to evaluate the risk of wildfires caused by tree-line discharge in a target area, the characteristic parameter data of the target area at the most recent T consecutive moments are collected to form a characteristic parameter sequence to be evaluated. The data is then standardized using the same method as in step S2. The characteristic parameter sequence to be evaluated after data standardization is input into the tree-line discharge risk assessment model trained in step S5 to evaluate the risk level of wildfires caused by tree-line discharge.

[0012] In addition, the present invention also provides a risk assessment system for tree line discharge-induced wildfires based on big data, which is characterized by comprising a sensor network, a drone, a hyperspectrometer, and a host computer, wherein:

[0013] The sensor network is used to collect meteorological parameters, treeline spatial parameters, and geographic environmental parameters of the target area and send them to the host computer;

[0014] The drone is used to carry a hyperspectrometer, which is used to capture hyperspectral images of tree lines in the target area and upload them to the host computer;

[0015] The host computer is used to receive meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network and treeline hyperspectral images uploaded by the hyperspectral instrument, and use the pre-trained treeline discharge risk assessment model to conduct risk assessment of treeline discharge-induced wildfires; the host computer is equipped with a data processing module, a data standardization module and a pre-trained treeline discharge risk assessment model, including:

[0016] The data processing module is used to receive meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network and treeline hyperspectral images uploaded by the hyperspectral instrument, and extract vegetation parameters from the treeline hyperspectral images; the data processing module forms a characteristic parameter sequence X with the characteristic parameter data of the current moment and the previous T-1 consecutive moments, each characteristic parameter data includes the meteorological parameters, treeline spatial parameters, geographical environment parameters and vegetation parameters at the moment, and then sends the characteristic parameter sequence X to the data standardization module;

[0017] The data standardization module is used to standardize the characteristic parameter sequence X and then input it into the tree line discharge risk assessment model;

[0018] The treeline discharge risk assessment model is used to evaluate the risk level of inducing wildfires based on the characteristic parameter sequence after data standardization.

[0019] The present invention provides a risk assessment method and system for tree-line discharge-induced wildfires based on big data, and constructs a tree-line discharge risk assessment model based on a neural network. The model has a characteristic parameter sequence as input and a risk level as output. The model is trained using a training sample set extracted from historical environmental parameter data before each tree-line discharge-induced wildfire event. When it is necessary to assess the risk of tree-line discharge-induced wildfires in a target area, the environmental parameter data of the target area is collected to form a characteristic parameter sequence to be assessed, and then the sequence is input into the tree-line discharge risk assessment model to obtain the risk level of tree-line discharge-induced wildfires.

[0020] The present invention collects historical and real-time environmental data and uses a neural network model to conduct risk assessment, thereby achieving early warning and effective management of wildfire risks, thereby achieving early prediction and effective management of wildfire risks, which not only improves the accuracy and timeliness of wildfire warnings, but also improves the safety management capabilities of power systems. The present invention uses automated feature extraction and risk prediction, which significantly improves the efficiency and accuracy of risk assessment, while reducing the need for manual intervention, providing a strong technical guarantee for the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of a specific implementation method of the risk assessment method of tree line discharge induced wildfire based on big data of the present invention;

[0022] Figure 2 is a structural diagram of the treeline discharge risk assessment model in the present invention;

[0023] Figure 3 is a structural diagram of a bidirectional long short-term memory network based on an attention mechanism in this embodiment;

[0024] Figure 4 It is a structural diagram of the risk assessment system of tree line discharge induced wildfire based on big data of the present invention;

[0025] Figure 5 Schematic diagram of the actual application of the risk assessment system for tree line discharge-induced wildfires based on big data in this embodiment. DETAILED DESCRIPTION

[0026] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0027] Example

[0028] Figure 1This is a flowchart of a specific implementation of the risk assessment method of tree line discharge-induced wildfires based on big data of the present invention. Figure 1 As shown, the risk assessment method of tree line discharge-induced wildfire based on big data of the present invention includes the following steps:

[0029] S101: Collect historical data:

[0030] According to actual needs, several historical events of treeline discharge-induced wildfires are collected, and the historical data of characteristic parameters before each treeline discharge-induced wildfire event are obtained. Several characteristic parameter sequences are extracted from them. Each characteristic parameter sequence includes environmental parameter data at T consecutive moments, and each parameter data includes meteorological parameters, treeline spatial parameters, geographical environment parameters and vegetation parameters at that moment.

[0031] Each characteristic parameter sequence is recorded as Where N represents the number of characteristic parameters, and each characteristic parameter sequence X is labeled with a risk level to obtain a risk level label y, y = 1, 2, ..., K, K represents the number of preset risk levels, and the larger the value of label y, the higher the risk of inducing wildfires.

[0032] In this embodiment, meteorological parameters include temperature, humidity, wind speed, and wind direction of the corresponding area; treeline spatial parameters include the distance and angle between trees and transmission lines; geographic environment parameters include soil moisture and terrain category; and vegetation parameters include vegetation spectral reflectance, vegetation index, pigment content, moisture content, and pest and disease indicators. Meteorological data, treeline spatial parameters, and geographic environment parameters can be collected by deploying sensors and monitoring equipment in the target area. Vegetation parameters are obtained by analyzing hyperspectral images collected by a hyperspectral instrument carried by an unmanned aerial vehicle, wherein spectral reflectance can be directly obtained through hyperspectral data. Spectral reflectance reflects the ability of leaves to reflect light at different wavelengths, and can be used to identify and evaluate the growth and health of trees; vegetation index calculation can be calculated using a specific mathematical model, and vegetation index includes the Normalized Difference Vegetation Index (NDVI), Chlorophyll Absorption Index (CAI), etc. These indices can reflect the health and biochemical characteristics of leaves, and can reflect the density and health of vegetation, which is of great significance for predicting trees that may come into contact with transmission lines; by analyzing the spectral absorption characteristics within a specific wavelength range, the leaf density can be estimated. The content of chlorophyll, carotenoids and other pigments in the leaves is crucial for evaluating the physiological state and growth trend of trees. The distance to the tree line can be estimated through the growth trend of trees. The water in the leaves will produce absorption characteristics in certain specific bands of the spectrum (such as 1.4um, 1.9um and 2.7um). The water content can be estimated by analyzing the spectral data of these bands, which is of reference value for predicting the growth status of trees under drought conditions and possible fire risks. Pests and diseases can affect the spectral characteristics of leaves. By analyzing abnormal changes in spectral data, leaves that may be affected by pests and diseases can be identified, which is crucial for timely detection and treatment of vegetation problems that may increase the risk of fire.

[0033] After research, it is found that the environmental parameter data set in this embodiment can fully reflect various environmental factors that may affect tree line discharge and wildfire risks. By integrating these parameters, a more accurate risk assessment can be provided for the safety management of transmission lines, so that effective prevention and emergency measures can be taken to reduce the possibility of wildfires.

[0034] S102: Data Standardization:

[0035] Data standardization is performed on each characteristic parameter sequence X to reduce the impact of deviations and outliers and improve the quality and reliability of the data. The specific method of data standardization in this embodiment is:

[0036] First, for each parameter x in the feature parameter sequence X t,n Normalization is performed, t=1,2,…,T, n=1,2,…,N, N represents the number of environmental parameters, and the normalization formula is as follows:

[0037]

[0038] in, Indicates the normalized parameter value, min n 、max n Respectively represent the minimum and maximum value of the nth parameter.

[0039] Then the parameters are decentralized to obtain the decentralized parameter values

[0040]

[0041] Among them, avg n Represents the average of all normalized values ​​of the nth parameter.

[0042] An outlier detection is performed on the decentralized parameter value of each parameter, and when an outlier is detected, the outlier is replaced by the average value of the decentralized parameter value. In this embodiment, the outlier detection adopts the IQR (interquartile range) outlier detection method.

[0043] S103: Constructing a training sample set:

[0044] The feature parameter sequence after data standardization is taken as input, and the corresponding risk level is taken as the expected output to construct the training samples, thus obtaining the training sample set.

[0045] S104: Constructing a treeline discharge risk assessment model:

[0046] In order to better assess the risk of wildfires induced by treeline discharge, the present invention constructs a treeline discharge risk assessment model based on a neural network, whose input is a feature parameter sequence and output is a risk level. Figure 2 is a structural diagram of the tree line discharge risk assessment model in the present invention. Figure 2 As shown, the tree line discharge risk assessment model in the present invention includes a convolutional neural network (CNN), a bidirectional long short-term memory network (BiLSTM-Bidirectional Long Short-Term Memory) and a risk determination module, wherein:

[0047] The convolutional neural network is used to process the feature parameter sequence, extract the features of the feature parameter data at each moment, and send the feature sequence to the bidirectional long short-term memory network. The convolutional neural network is composed of multiple layers of convolutional layers and pooling layers, and uses its convolutional layers and pooling layers to automatically learn the local features and spatial hierarchical structures in the environmental parameter data.

[0048] Convolutional neural networks are mainly used to process multidimensional spatial data and extract complex features. The risk assessment of wildfires induced by treeline discharge in the present invention relies on a variety of input data, namely meteorological parameters, treeline spatial parameters, geographical environment parameters and vegetation parameters, and convolutional neural networks are particularly suitable for extracting important features from such high-dimensional and complex input data, providing rich feature representations for subsequent time series analysis. For example, through the convolution operation, it is possible to identify a pattern with a higher discharge risk when the distance between trees and power lines is close and the wind speed is high. Through the local receptive field mechanism, relevant risk features can be extracted under conditions such as the local relationship between trees and power lines and wind speed, thereby more accurately predicting the discharge risk in certain areas. Through a multi-layer convolution structure, multi-level features of data can be extracted. For example, in the low-level convolution layer, CNN can extract local features such as temperature and humidity, and in the higher-level convolution layer, more complex spatial relationships can be extracted, such as the relative distance between trees and power lines and terrain, etc. The hierarchical feature representation improves the accuracy of risk assessment. CNN effectively integrates multiple data sources through multi-channel input (such as each channel represents different types of data), and comprehensively extracts the correlation between different data, so as to better assess the discharge risk.

[0049] The bidirectional long short-term memory network is used to process the feature sequence, obtain the risk features corresponding to the input feature parameter sequence and input them into the risk determination module. The bidirectional long short-term memory network is an extension of the LSTM (Long Short-Term Memory, Long Short-Term Memory Network LSTM). It processes data in two directions: forward and reverse, and then concatenates and outputs the forward features and reverse features. The bidirectional long short-term memory network can capture the time dependency and long-term memory in the data, providing more comprehensive time dynamic information for risk assessment.

[0050] In the present invention, an attention mechanism is integrated into the bidirectional long short-term memory network to enhance the model's ability to capture key information. By learning the importance of different features in the data, the model can pay more attention to the most critical information for risk prediction, thereby improving the accuracy and efficiency of the prediction. Figure 3 is a structural diagram of a bidirectional long short-term memory network based on the attention mechanism in this embodiment. Figure 3 As shown, the bidirectional long short-term memory network based on the attention mechanism in this embodiment includes an input embedding module, a bidirectional long short-term memory network and an attention layer, wherein:

[0051] The input embedding module is used to embed the feature parameter sequence and send the embedded vector to the forward long short-term memory layer and the reverse long short-term memory layer. Through the input module, discrete input can be converted into a continuous vector representation.

[0052] The bidirectional long short-term memory network is used to perform forward and backward processing on the embedding vector, concatenate the representation of each time step in the forward context representation vector and the reverse context representation vector, obtain the bidirectional context representation vector and send it to the attention module.

[0053] The attention module is used to process the bidirectional context representation vector using the attention mechanism, extract the weight of each time step, and perform weighted average on the bidirectional context representation vector to obtain the risk feature. The attention mechanism is a commonly used neural network mechanism, and its specific process will not be repeated here.

[0054] The risk determination module is used to process the risk characteristics and obtain the risk level at the last moment in the characteristic parameter data. In this embodiment, the risk determination module adopts a fully connected layer.

[0055] S105: Training Treeline Discharge Risk Assessment Model:

[0056] The training sample set obtained in step S103 is used to train the treeline discharge risk assessment model to obtain a trained treeline discharge risk assessment model.

[0057] In this embodiment, the calculation formula of the loss function L used in the treeline discharge risk assessment model training is as follows:

[0058]

[0059] Among them, y k represents the true probability that the input feature parameter sequence determined by the risk level label y belongs to the kth risk level, It represents the probability that the input feature parameter sequence obtained by the treeline discharge risk assessment model belongs to the kth risk level.

[0060] In order to improve the training effect of the treeline discharge risk assessment model, the Newton-Raphson Optimization (NRBO) algorithm is used in this embodiment to update the model parameters. The NRBO algorithm is mainly used to minimize the objective function through an iterative optimization method, thereby optimizing the parameters of the model. For the present invention, NRBO can help quickly find the optimal model parameter combination so that the model can accurately predict the risk of treeline discharge under different meteorological and environmental conditions. There is often a complex nonlinear relationship between the risk of treeline discharge and various factors (such as temperature, humidity, wind speed, tree characteristics, etc.), and NRBO has obvious advantages in dealing with these nonlinear relationships. Using Taylor series expansion, the nonlinear characteristics of the data can be considered when updating the parameters, thereby effectively dealing with nonlinear relationships. In addition, the NRBO algorithm can efficiently find the model parameters that minimize the loss function by using the gradient information of the first-order and second-order derivatives of the loss function, and can better deal with complex nonlinear models than the traditional gradient descent method, avoiding falling into the local optimal solution, thereby improving the performance of the treeline discharge risk assessment model finally trained and improving the assessment accuracy.

[0061] S106: Treeline discharge risk assessment for target areas:

[0062] When it is necessary to evaluate the risk of wildfires caused by tree-line discharge in a target area, the environmental parameter data of the target area at the most recent T consecutive moments are collected to form a characteristic parameter sequence to be evaluated, and then the data is standardized using the same method in step S102. The characteristic parameter sequence to be evaluated after data standardization is input into the tree-line discharge risk assessment model trained in step S105 to evaluate the risk level of wildfires caused by tree-line discharge.

[0063] In this embodiment, the meteorological parameters in the environmental parameter data are collected by meteorological sensors, the tree line space parameters are collected by laser rangefinders, the soil moisture in the geographical environmental parameters is collected by humidity sensors arranged in the soil, and the terrain category is obtained by querying the coordinates of the target area from a pre-set terrain database.

[0064] In order to better implement the risk assessment method of tree-line discharge-induced wildfires based on big data of the present invention, the present invention also designs a risk assessment system of tree-line discharge-induced wildfires based on big data. Figure 4 This is a structural diagram of the risk assessment system for tree line discharge-induced wildfires based on big data of the present invention. Figure 4 As shown, the risk assessment system of tree line discharge-induced wildfire based on big data of the present invention includes a sensor network 1, a drone 2, a hyperspectrometer 3, and a host computer 4, wherein:

[0065] The sensor network 1 is used to collect meteorological parameters, tree line space parameters and geographical environment parameters of the target area and send them to the host computer 4. In this embodiment, the sensor network 1 includes a meteorological sensor 11, a laser rangefinder 12 and a soil moisture sensor 13, wherein the meteorological sensor 11 is used to collect meteorological parameters, including temperature, humidity, wind speed and wind direction; the laser rangefinder 12 is installed on the conductor of the transmission line to measure the distance and angle between the tree and the transmission line, and the soil moisture sensor 13 is arranged in the soil to collect soil moisture.

[0066] In order to ensure the reliability of the environmental sensor network, in this embodiment, a conductor CT energy acquisition module is configured for the sensor network 1 to obtain power from the power transmission line and supply power to the environmental sensor network.

[0067] In this embodiment, the laser ranging module 12 is based on phase information detection technology, has high precision, low power consumption and strong anti-interference ability, can operate stably on high-voltage wires, and includes a modulated laser transmitting system, a receiving system and a signal processing system. The transmitting system is composed of a semiconductor laser diode, a drive modulation circuit and a soft start circuit. The receiving system is composed of an avalanche photodiode, a temperature compensation circuit and an amplification filter circuit. The signal processing part converts the signal frequency through a mixing circuit and determines the phase difference using automatic digital ranging technology, and is finally processed by a single-chip microcomputer. Electronic technologies such as modulation signal stability, photoelectric detection signal signal-to-noise ratio, mixing circuit noise and phase measurement accuracy processing are used to ensure data acquisition accuracy in a high-voltage and strong magnetic field environment.

[0068] In addition, in order to facilitate the staff to view the on-site conditions, this embodiment also configures a visual monitoring module in the sensor network 1, which is turned on or off by the control signal of the host computer 4, and is used to collect and store on-site images of the transmission line and upload them to the host computer 4.

[0069] Drone 2 is used to carry hyperspectrometer 3, which is used to capture hyperspectral images of the tree line in the target area and upload them to the host computer. Hyperspectral imaging technology can capture the reflectance characteristics of plant leaves in different spectral bands, thereby obtaining vegetation parameters in the target area. These data are crucial for assessing the potential contact risk between trees and power lines, as they can help predict the growth trends of trees and possible tree line contact points. The use of drones provides a fast, efficient and safe means of monitoring, which can continuously cover a wide area without invading the natural environment.

[0070] The host computer 4 is used to receive the meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network 1 and the treeline hyperspectral images uploaded by the hyperspectrometer 3, and use the pre-trained treeline discharge risk assessment model to conduct risk assessment of treeline discharge-induced wildfires. The host computer is equipped with a data processing module 41, a data standardization module 42 and a pre-trained treeline discharge risk assessment model 43, where:

[0071] The data processing module 41 is used to receive the meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network 1 and the treeline hyperspectral image uploaded by the hyperspectrometer 3, and extract vegetation parameters from the treeline hyperspectral image. In this embodiment, the vegetation parameters include vegetation spectral reflectance, vegetation index, pigment content, moisture content and pest index. The data processing module 41 forms a characteristic parameter sequence X with the characteristic parameter data of the current moment and the previous T-1 consecutive moments, each characteristic parameter data includes the meteorological parameters, treeline spatial parameters, geographical environment parameters and vegetation parameters at the moment, and then sends the characteristic parameter sequence X to the data standardization module 42.

[0072] The data standardization module 42 is used to perform data standardization on the characteristic parameter sequence X and then input the data into the treeline discharge risk assessment model 43 .

[0073] The treeline discharge risk assessment model 43 is used to evaluate the risk level of inducing wildfires based on the characteristic parameter sequence after data standardization.

[0074] In this embodiment, the host computer is implemented by an STM32 single-chip microcomputer and a central processing unit, wherein the STM32 single-chip microcomputer is equipped with a data processing module and a data standardization module, and the central processing unit is equipped with a pre-trained tree line discharge risk assessment model. The STM32 single-chip microcomputer and the sensor network communicate using a 5G communication network to ensure reliable data transmission.

[0075] In this embodiment, the host computer also includes an early warning system for receiving the risk level assessment results output by the central processing unit and automatically issuing an early warning signal when a high-risk situation is detected. The early warning signal includes the location, time and risk level of a possible wildfire.

[0076] Figure 5 : is a schematic diagram of the actual application of the risk assessment system for tree line discharge-induced wildfires based on big data in this embodiment. Figure 5 As shown, in the risk assessment system for wildfires induced by treeline discharge based on big data in this embodiment, the sensor network 1 collects meteorological parameters, treeline spatial parameters and geographical environment parameters of the target area, the drone 2 is equipped with a hyperspectral instrument 3 to capture hyperspectral images of the treelines in the target area, and the host computer 4 completes the risk assessment of wildfires induced by treeline discharge based on the above data.

[0077] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. A risk assessment method for treeline discharge-induced wildfires based on big data, characterized in that: The following steps are involved: S1: According to actual needs, several historical events of treeline discharge-induced wildfires are collected, and the characteristic parameter historical data before each treeline discharge-induced wildfire event is obtained, and several characteristic parameter sequences are extracted from them. Each characteristic parameter sequence includes characteristic parameter data of T consecutive moments, and each characteristic parameter data includes meteorological parameters, treeline space parameters, geographical environment parameters and vegetation parameters at that moment; Each characteristic parameter sequence is recorded as Where N represents the number of characteristic parameters, and each characteristic parameter sequence X is labeled with a risk level to obtain a risk level label y, y = 1, 2, ..., K, K represents the number of preset risk levels, and the maximum value of label y indicates a higher risk of inducing wildfires; S2: Perform data standardization on each feature parameter sequence X; S3: Take the feature parameter sequence after data standardization as input and the corresponding risk level as the expected output to construct training samples, thus obtaining the training sample set; S4: construct a treeline discharge risk assessment model, whose input is a characteristic parameter sequence and output is a risk level; S5: using the training sample set obtained in step S3 to train the treeline discharge risk assessment model to obtain a trained treeline discharge risk assessment model; S6: When it is necessary to evaluate the risk of wildfires caused by tree-line discharge in a target area, the characteristic parameter data of the target area at the most recent T consecutive moments are collected to form a characteristic parameter sequence to be evaluated. The data is then standardized using the same method as in step S2. The characteristic parameter sequence to be evaluated after data standardization is input into the tree-line discharge risk assessment model trained in step S5 to evaluate the risk level of wildfires caused by tree-line discharge.

2. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 1 is characterized in that: The meteorological parameters in step S1 include the temperature, humidity, wind speed and wind direction of the corresponding area; the tree line space parameters include the distance and angle between the trees and the transmission lines; the geographical environment parameters include soil moisture and terrain category; the vegetation parameters include vegetation spectral reflectance, vegetation index, pigment content, moisture content and pest and disease indicators.

3. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 1 is characterized in that: The specific method of data standardization in step S2 is: First, for each parameter x in the feature parameter sequence X t,n Normalization is performed, t=1,2,…,T, n=1,2,…,N, N represents the number of environmental parameters, and the normalization formula is as follows: in, Indicates the normalized parameter value, min n 、max n Respectively represent the minimum and maximum values ​​of the nth parameter; Then the parameters are decentralized to obtain the decentralized parameters Among them, avg n Represents the average value of all normalized values ​​of the nth parameter; An outlier detection is performed on the decentralized parameter value of each parameter. When an outlier is detected, the average value of the decentralized parameter value is used to replace the outlier.

4. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 1 is characterized in that: The treeline discharge risk assessment model in step S4 includes a convolutional neural network, a bidirectional long short-term memory network and a risk determination module, wherein: The convolutional neural network is used to process the feature parameter sequence, extract the features of the feature parameter data at each moment, and obtain the feature sequence and send it to the bidirectional long short-term memory network; The bidirectional long short-term memory network is used to process the feature sequence, obtain the risk features corresponding to the input feature parameter sequence and input them into the risk determination module; The risk determination module is used to process the risk characteristics and obtain the risk level at the last moment in the characteristic parameter sequence.

5. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 4 is characterized in that: The bidirectional long short-term memory network adopts a bidirectional long short-term memory network based on an attention mechanism, including an input module, a bidirectional long short-term memory network and an attention layer, wherein: The input module is used to embed the feature parameter sequence and send the obtained embedding vector to the forward long short-term memory layer and the reverse long short-term memory layer; The bidirectional long short-term memory network is used to process the embedding vector forward and backward, concatenate the representation of each time step in the forward context representation vector and the reverse context representation vector, obtain the bidirectional context representation vector and send it to the attention module; The attention module is used to process the bidirectional context representation vector using the attention mechanism, extract the weight of each time step, and perform weighted averaging on the bidirectional context representation vector to obtain the risk feature.

6. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 1 is characterized in that: The calculation formula of the loss function L during the training of the treeline discharge risk assessment model is as follows: Among them, y k represents the true probability that the input feature parameter sequence determined by the risk level label y belongs to the kth risk level, It represents the probability that the input feature parameter sequence obtained by the treeline discharge risk assessment model belongs to the kth risk level.

7. The risk assessment method for treeline discharge-induced wildfires based on big data according to claim 1 is characterized in that: The Newton-Raphson optimization algorithm is used to adjust the model parameters during the training of the treeline discharge risk assessment model.

8. A risk assessment system for treeline discharge-induced wildfires based on big data, characterized in that: It includes sensor network, drone, hyperspectrometer and host computer, including: The sensor network is used to collect meteorological parameters, treeline spatial parameters, and geographic environmental parameters of the target area and send them to the host computer; The drone is used to carry a hyperspectrometer, which is used to capture hyperspectral images of tree lines in the target area and upload them to the host computer; The host computer is used to receive meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network and treeline hyperspectral images uploaded by the hyperspectral instrument, and use the pre-trained treeline discharge risk assessment model to conduct risk assessment of treeline discharge-induced wildfires; the host computer is equipped with a data processing module, a data standardization module and a pre-trained treeline discharge risk assessment model, including: The data processing module is used to receive meteorological parameters, treeline spatial parameters and geographical environment parameters uploaded by the sensor network and treeline hyperspectral images uploaded by the hyperspectral instrument, and extract vegetation parameters from the treeline hyperspectral images; the data processing module forms a characteristic parameter sequence X with the characteristic parameter data of the current moment and the previous T-1 consecutive moments, each characteristic parameter data includes the meteorological parameters, treeline spatial parameters, geographical environment parameters and vegetation parameters at the moment, and then sends the characteristic parameter sequence X to the data standardization module; The data standardization module is used to standardize the characteristic parameter sequence X and then input it into the tree line discharge risk assessment model; The treeline discharge risk assessment model is used to evaluate the risk level of inducing wildfires based on the characteristic parameter sequence after data standardization.

9. The risk assessment system for treeline discharge-induced wildfires based on big data according to claim 8, characterized in that: The meteorological parameters include the temperature, humidity, wind speed and wind direction of the corresponding area; the tree line spatial parameters include the distance and angle between trees and transmission lines; the geographical environment parameters include soil moisture and terrain category; the vegetation parameters include vegetation spectral reflectance, vegetation index, pigment content, moisture content and pest and disease indicators.

10. The risk assessment system for treeline discharge-induced wildfires based on big data according to claim 8, characterized in that: The host computer also includes an early warning system for receiving the risk level assessment result output by the central processing unit and automatically issuing an early warning signal when a high-risk situation is detected. The early warning signal includes the location, time and risk level of a possible wildfire.

Images