Power forest fire intelligent monitoring method, device and equipment based on pass-through remote fusion and storage medium

Abstract

The application discloses a kind of power forest fire intelligent monitoring method, device and equipment based on through remote fusion, and storage medium, it is related to forest fire intelligent monitoring technical field, comprising: obtaining the position information of first monitoring node, and obtaining the temperature and humidity data and smoke concentration data corresponding to position information as environmental data;Position information and environmental data are input into least square support vector machine model for processing, and fire risk probability is obtained;When fire risk probability is greater than preset risk threshold, position information, environmental data and fire risk probability are combined to obtain fire preliminary screening information, and fire preliminary screening information is uploaded to cloud server, so that cloud server fuses satellite remote sensing data to complete fire confirmation, realize the non-blind area monitoring coverage of power forest fire along transmission line, realize the preferential transmission of key fire data, reduce communication resource waste, improve system response speed and transmission reliability of key data.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for wildfires, and in particular to an intelligent monitoring method, device, equipment, and storage medium for power wildfires based on communication, conduction, and remote sensing fusion. Background Technology

[0002] As the power grid continues to expand, high-voltage transmission lines are gradually extending into remote mountainous areas and forests prone to wildfires. These areas have complex natural environments and rugged geographical conditions, exposing transmission lines to the risk of extreme natural disasters such as wildfires for extended periods, resulting in significant safety hazards. Wildfires can not only cause transmission line tripping and insulation breakdown but also trigger widespread power outages, severely impacting the safe and stable operation of the power grid. Real-time and accurate wildfire monitoring and early warning are crucial for ensuring the safety of transmission lines and the stability of the power system.

[0003] Existing power plant wildfire monitoring technologies mainly include three types: ground-based sensor monitoring, satellite remote sensing monitoring, and drone patrols. Ground-based sensor monitoring typically uses ground-based sensor networks composed of temperature, humidity, and smoke sensors, transmitting data to the monitoring center via wireless communication. Its advantages include good real-time performance and low equipment cost, but it is limited by the coverage of the communication network, often deployed in specific areas, resulting in communication blind spots and making it difficult to achieve full coverage of remote mountainous areas and vast forest regions. Satellite remote sensing monitoring uses optical or infrared remote sensing satellites to monitor large-scale forest fires, offering advantages such as wide coverage and periodic data transmission. However, its resolution is relatively low, making it difficult to accurately identify early small fires in densely vegetated areas, and the satellite revisit cycle also limits the timely detection of fires. Drone patrol monitoring uses drones to conduct regular patrols of key areas, offering good real-time performance and the ability to acquire high-resolution images, but it is constrained by flight time, weather conditions, and patrol costs, making long-term stable operation difficult.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a method, device, equipment, and storage medium for intelligent monitoring of power forest fires based on communication, navigation, and remote sensing integration. The aim is to solve the technical problem of how to achieve early monitoring and location of power forest fires along transmission lines in remote mountainous areas without terrestrial communication coverage, with no blind spots, low latency, and high accuracy.

[0006] To achieve the above objectives, this invention provides a method for intelligent monitoring of power-related wildfires based on communication, conduction, and remote sensing fusion. The method includes the following steps:

[0007] The location information of the first monitoring node is obtained, and the temperature and humidity data and smoke concentration data corresponding to the location information are obtained as environmental data.

[0008] The location information and environmental data are input into a least squares support vector machine model for processing to obtain the fire risk probability.

[0009] When the probability of a fire risk is greater than a preset risk threshold, the location information, the environmental data, and the probability of a fire risk are combined to obtain preliminary fire screening information, which is then uploaded to a cloud server so that the cloud server can integrate satellite remote sensing data to confirm the fire.

[0010] In one embodiment, the step of inputting the location information and the environmental data into a least squares support vector machine model for processing to obtain the fire risk probability includes:

[0011] The location deviation from the historical monitoring nodes is calculated based on the location information to obtain the location deviation value;

[0012] The temperature and humidity data, the smoke concentration data, and the location deviation value are combined into an input feature vector and mapped to a high-dimensional feature space to obtain a mapping vector.

[0013] The kernel function value is calculated using the support vectors and the mapping vector, and the fire risk probability is calculated using the kernel function value and the Lagrange multiplier.

[0014] In one embodiment, the step of calculating the fire risk probability based on the kernel function value and the Lagrange multiplier includes:

[0015] The kernel function matrix is ​​weighted and summed with the Lagrange multipliers to obtain the weighted value.

[0016] The weighted value is added to the bias term to obtain the intermediate score;

[0017] The intermediate score is input into the activation function to obtain the fire risk probability.

[0018] In one embodiment, the step of combining the location information, the environmental data, and the fire risk probability to obtain preliminary fire screening information when the fire risk probability is greater than a preset risk threshold, and uploading the preliminary fire screening information to a cloud server, includes:

[0019] The location information is encoded to obtain a first data frame;

[0020] The environmental data is encoded to obtain a second data frame;

[0021] The fire risk probability is encoded to obtain the third data frame;

[0022] The first data frame, the second data frame, and the third data frame are encapsulated to obtain initial fire screening information;

[0023] The initial fire information is uploaded to a cloud server via a low-orbit satellite link.

[0024] In one embodiment, the step of uploading the initial fire screening information to a cloud server via a low-Earth orbit satellite link includes:

[0025] The initial fire screening information is divided into a preset number of data packets according to a preset packet length;

[0026] Add a checksum and a timestamp to each of the data packets, and send each data packet with the added checksum and timestamp to the cloud server in sequence;

[0027] If no confirmation frame is received from the cloud server, the corresponding data packet is retransmitted.

[0028] In one embodiment, the step of enabling the cloud server to fuse satellite remote sensing data to complete fire confirmation includes:

[0029] Send a remote sensing data request command to the cloud server;

[0030] Receive the fire confirmation result returned by the cloud server, wherein the fire confirmation result includes the location of the fire and the fire level;

[0031] Write the fire confirmation result into the local fire status flag;

[0032] The fire status marker is broadcast to other monitoring nodes in the LoRa network.

[0033] In one embodiment, the method further includes:

[0034] Receive a model update instruction sent by the cloud server, wherein the model update instruction includes the latest model parameters;

[0035] Based on the latest model parameters, the least squares support vector machine model is hot-updated to obtain the updated model.

[0036] The updated model is used to process the location information and the environmental data to obtain the updated fire risk probability.

[0037] Furthermore, to achieve the above objectives, this invention also proposes a smart power fire monitoring device based on communication, navigation, and remote sensing fusion, the device comprising:

[0038] The acquisition module is used to acquire the location information of the first monitoring node, and acquire the temperature and humidity data and smoke concentration data corresponding to the location information as environmental data;

[0039] The analysis module is used to input the location information and the environmental data into a least squares support vector machine model for processing to obtain the fire risk probability;

[0040] The upload module is used to combine the location information, the environmental data, and the fire risk probability to obtain preliminary fire screening information when the fire risk probability is greater than a preset risk threshold, and upload the preliminary fire screening information to the cloud server so that the cloud server can integrate satellite remote sensing data to complete the fire confirmation.

[0041] Furthermore, to achieve the above objectives, the present invention also proposes a power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion. The device includes: a memory, a processor, and a power fire intelligent monitoring program based on communication, navigation, and remote sensing fusion stored in the memory and executable on the processor. The power fire intelligent monitoring program based on communication, navigation, and remote sensing fusion is configured to implement the steps of the power fire intelligent monitoring method based on communication, navigation, and remote sensing fusion as described above.

[0042] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing a power fire intelligent monitoring program based on communication, navigation, and remote sensing fusion. When the power fire intelligent monitoring program based on communication, navigation, and remote sensing fusion is executed by a processor, it implements the steps of the power fire intelligent monitoring method based on communication, navigation, and remote sensing fusion as described above.

[0043] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion as described above.

[0044] One or more technical solutions proposed in this application have at least the following technical effects:

[0045] By employing a hybrid space-ground communication network architecture for intelligent monitoring of power transmission line wildfires, this system innovatively integrates low-Earth orbit satellite internet, terrestrial LoRa self-organizing networks, and traditional communication methods, achieving seamless coverage and communication assurance across the monitoring area. Through cloud-edge collaborative hierarchical fire risk screening and edge processing algorithms, lightweight optimization algorithms ensure low-power, high-efficiency operation of edge devices and prioritize the transmission of critical fire data, improving the system's timeliness and stability. A smart fusion deep learning model based on multi-source heterogeneous data (CNN-MLP hybrid model) enables high-precision identification of early-stage fires and small fire points, improving the accuracy and reliability of traditional single-data source identification methods. A multi-mode communication node hardware design integrating BeiDou positioning, LoRa communication, and low-Earth orbit satellite communication ensures the engineering application and stable operation of the space-ground fusion communication and intelligent data processing model in complex environments. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating an embodiment of the intelligent monitoring method for power wildfires based on the fusion of communication, navigation, and remote sensing in this application.

[0049] Figure 2 This is a flowchart of cloud-edge collaborative hierarchical processing provided in Embodiment 1 of the intelligent monitoring method for power wildfires based on communication, navigation and remote sensing integration in this application;

[0050] Figure 3 This is a diagram of the CNN-MLP hybrid model structure provided in Embodiment 1 of the intelligent monitoring method for power wildfires based on communication, navigation and remote sensing fusion in this application;

[0051] Figure 4 This is a flowchart illustrating Embodiment 2 of the intelligent monitoring method for power wildfires based on communication, navigation, and remote sensing integration provided in this application;

[0052] Figure 5 This is a diagram of the space-ground collaborative three-dimensional monitoring network architecture according to an embodiment of this application;

[0053] Figure 6 This is a system hardware design block diagram of an embodiment of this application;

[0054] Figure 7 This is a schematic diagram of the module structure of the intelligent power wildfire monitoring device based on communication, navigation and remote sensing fusion according to an embodiment of this application;

[0055] Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the intelligent monitoring method for power wildfires based on communication, navigation, and remote sensing integration in the embodiments of this application.

[0056] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0057] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0058] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0059] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion. The following description uses a power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion as an example to illustrate this embodiment and the subsequent embodiments.

[0060] Based on this, the embodiments of this application provide a method for intelligent monitoring of power wildfires based on communication, navigation, and remote sensing fusion, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the intelligent monitoring method for power field fires based on the fusion of communication, navigation, and remote sensing in this application.

[0061] In this embodiment, the intelligent monitoring method for power wildfires based on communication, navigation, and remote sensing fusion includes steps S10 to S30:

[0062] Step S10: Obtain the location information of the first monitoring node, and obtain the temperature and humidity data and smoke concentration data corresponding to the location information as environmental data;

[0063] It should be noted that the first monitoring node refers to the monitoring equipment installed along the transmission line, which is used to collect data on the surrounding environment.

[0064] Location information refers to the geographical location data of the monitoring nodes obtained through the BeiDou positioning system, which is used to determine the specific location of the monitoring points.

[0065] Environmental data includes temperature and humidity data and smoke concentration data. These data reflect the environmental conditions at the monitoring points and can be used to determine whether wildfires have occurred.

[0066] In its implementation, this solution involves installing monitoring nodes integrated with BeiDou positioning, temperature and humidity, and smoke sensors in high-risk wildfire areas along the transmission lines, forming a dense network of monitoring points. These nodes are interconnected via low-power LoRa self-organizing network technology to achieve local data transmission and cover communication blind spots.

[0067] Sensor data is aggregated to a convergence node via a LoRa self-organizing network. The convergence node has a built-in ARM edge processor and a low-orbit satellite communication module to process and analyze massive amounts of data from different monitoring nodes and upload key data to the satellite network to ensure the transmission of critical data from remote areas.

[0068] It should be understood that this invention proposes a systematic technical solution, the core of which lies in constructing an intelligent monitoring system that integrates space-ground coordination, cloud-edge collaboration, data fusion, and hardware and software integration. This is specifically achieved through the following four interrelated and collaborative key components:

[0069] The construction of a three-dimensional monitoring network integrating space, ground, and air, and a cloud-edge collaborative hierarchical processing mechanism: This part is responsible for establishing a data transmission network for wildfire monitoring along power transmission lines. It involves deploying data acquisition nodes integrating BeiDou positioning and temperature, humidity, and smoke sensors in key areas, utilizing a low-power LoRa self-organizing network for local networking; setting up aggregation nodes with edge computing capabilities and low-Earth orbit satellite communication functions; and transmitting key fire information to the cloud in real time, where it is simultaneously accessed and analyzed using satellite remote sensing data sources. This network architecture fundamentally solves the problem of insufficient coverage in traditional terrestrial communication, providing a physical foundation for the integrated "communication, navigation, and remote sensing" system, achieving full coverage of the monitoring area and precise location of fire points.

[0070] Edge processing algorithm (LS-SVM optimization): The edge processing algorithm deployed at the aggregation node can perform real-time analysis of massive amounts of raw sensor data to conduct preliminary screening of wildfire risks. Only critical data identified as high-risk (including location, sensor data, and analysis results) is prioritized for uploading to the cloud via low-Earth orbit satellite links. This mechanism significantly reduces the amount of invalid data transmission on low-bandwidth satellite links, ensuring the timeliness of critical fire information transmission and alleviating the burden on the cloud through edge preprocessing. A lightweight least-squares support vector machine (LS-SVM) model is used as the edge risk analysis algorithm, and its hyperparameters are optimized using a genetic algorithm. The algorithm is designed to meet the stringent constraints of edge processors in terms of computing power, storage space, and power consumption, while maintaining high prediction accuracy and millisecond-level real-time response capability. This optimized algorithm can run effectively at resource-constrained edge devices and is a key technical support for reducing communication volume and improving response speed.

[0071] Multi-source heterogeneous data intelligent fusion algorithm: This part is deployed in the cloud and is responsible for the final high-precision fire situation determination. A CNN-MLP hybrid model is designed specifically for fusing edge-up sensor time-series data (one-dimensional) and satellite remote sensing image data (two-dimensional). The CNN branch (based on EfficientNet) extracts spatial features, and the MLP branch (combined with LSTM / self-attention) extracts temporal features. A dynamic weight fusion mechanism is used for collaborative analysis to output high-precision wildfire occurrence probability and fire location information. This algorithm effectively overcomes the limitations of a single data source and significantly improves the accuracy and comprehensiveness of identifying small fires and early-stage fires.

[0072] System Hardware Design: This section provides physical support for the aforementioned software algorithms and network architecture. It includes: data acquisition nodes powered by solar modules and integrating temperature, humidity, smoke sensors, BeiDou positioning, and LoRa communication modules; aggregation nodes equipped with high-performance ARM edge processors, LoRa, and low-Earth orbit satellite communication modules; and a cloud server platform based on GPU clusters and distributed storage. This hardware system design ensures the feasibility, stable operation, and long-term maintenance of the entire solution in complex field environments.

[0073] like Figure 2 As shown, Figure 2 This diagram illustrates the cloud-edge collaborative layered processing flow and showcases the model architecture of the multi-source data intelligent fusion algorithm:

[0074] After monitoring nodes collect BeiDou coordinates, temperature, humidity, and smoke concentration data in real time, they first perform missing value imputation and normalization preprocessing. Then, they use FDS simulated wildfire data to train samples, forming chromosomes by encoding kernel function parameters and regularization parameters, and using negative mean square error as the fitness. Through selection, crossover, mutation, and iterative optimization using a genetic algorithm, the optimal hyperparameters suitable for the LS-SVM model are obtained. The model is then used to infer the processed feature vectors and output the fire risk probability. When the probability is greater than 0.5, the BeiDou coordinates, sensor data, and analysis results are merged into the initial fire screening information and cached locally. This information is then immediately uploaded to the cloud via a low-orbit satellite terminal. The cloud then integrates satellite remote sensing images to perform spatial hotspot verification, ultimately accurately locating the fire point and providing the fire level. This achieves a closed-loop process for power plant wildfires, from rapid edge detection to accurate cloud confirmation.

[0075] CNN branch: EfficientNet extracts spatial features from satellite imagery (including depthwise separable convolutions and attention mechanisms). MLP branch: LSTM+self-attention analyzes sensor temporal data. Fusion layer: Dynamic weights combine features from both branches to output fire risk probability and segmentation map.

[0076] Step S20: Input the location information and environmental data into the least squares support vector machine model for processing to obtain the fire risk probability;

[0077] It should be noted that the least squares support vector machine model is a machine learning model used for data classification and regression analysis, and here it is used to calculate the probability of fire risk based on the input data.

[0078] The fire risk probability is a value output by the model, representing the likelihood of a wildfire occurring at the current monitoring point.

[0079] It should be understood that LS-SVM (Least Squares Support Vector Machine) is an ideal edge processing algorithm for implementing wildfire risk analysis on edge processors. Its advantages are mainly reflected in three aspects: First, the algorithm adopts a prediction mechanism based on linear combination of support vectors, and the computational complexity is only related to the number of support vectors. In a typical wildfire monitoring scenario, a single inference only requires 1.2ms, meeting the millisecond-level response requirements of real-time early warning. Second, the model only needs to store a small number of support vectors (usually 10-20KB), which can be permanently stored in the chip memory, greatly saving storage resources of edge devices. Finally, its convex optimization characteristics ensure that stable and reliable prediction results can still be obtained in resource-constrained environments, and the regularization parameters effectively avoid the problem of overfitting with small samples. These characteristics enable LS-SVM to adapt to the stringent resource constraints of the edge (such as a power budget of <5mW) while maintaining a prediction accuracy of over 92%, perfectly meeting the comprehensive requirements of wildfire monitoring systems for real-time performance, reliability, and energy efficiency.

[0080] Furthermore, before inputting the least squares support vector machine model, the following steps are also included:

[0081] ImageNet normalization was performed on satellite remote sensing images, and sliding window normalization was performed on temperature and humidity data and smoke concentration data to eliminate dimensional differences and stabilize the input distribution.

[0082] In practice, due to the significant differences in the numerical range and distribution of satellite imagery, sensor data, and meteorological data, standardization is often required. Satellite imagery uses ImageNet-standardized mean and standard deviation, while sensor data is normalized using a sliding window method.

[0083]

[0084] in, μRGB and σRGB are the mean and standard deviation of the three RGB channels, respectively, and I is the pixel value (0–255) of a certain pixel in any RGB channel of the original satellite image. The normalized pixel values, ranging from approximately [-3, 3], are directly fed into the CNN branch.

[0085] Sensor data normalization:

[0086]

[0087] Satellite imagery normalization uses pre-trained ImageNet statistics to make the input distribution more stable, while sensor data is normalized using a sliding window to preserve local temporal features while eliminating dimensional differences.

[0088] Furthermore, the kernel function parameters and regularization parameters of the least squares support vector machine model are obtained through a genetic algorithm, specifically including the following steps:

[0089] Encode the kernel function type and regularization parameters as chromosomes;

[0090] The chromosomes are subjected to crossover and arithmetic crossover processes to obtain offspring chromosomes;

[0091] Perform mutation operations on the offspring chromosomes;

[0092] Chromosomes are selected to be retained based on the fitness function, resulting in optimized kernel function parameters and regularization parameters.

[0093] It should be noted that genetic algorithms, as an optimization method simulating the natural evolutionary process, were used in this study for the automatic optimization of hyperparameters in the LS-SVM model. Its core principle is to continuously evolve parameter combinations through operations such as selection, crossover, and mutation to obtain the optimal solution. Considering the characteristics of LS-SVM, we designed a hybrid encoding strategy, with chromosome design incorporating kernel function type and regularization parameters.

[0094]

[0095] Where k ∈ {linear, poly, rbf}, the value is limited to one of three types: “linear”, “poly”, or “radial basis”, which is used to determine the kernel mapping method of LS-SVM.

[0096] C ∈ , 0.1 ≤ C ≤ 10, represents the penalty strength for classification error; the larger the value, the more sensitive the model is to training error, and the value range is limited to between 0.1 and 10, and the optimal value is searched by the genetic algorithm during the evolution process.

[0097] Chromosome is a chromosome encoding form that combines the kernel function type k with the regularization parameter C into a binary vector, which serves as the individual (chromosome) in the genetic algorithm and is used for crossover, mutation, and selection operations.

[0098] When the RBF core is selected, the default settings are... This setting can automatically adapt to the data scale, ensuring that features of different dimensions receive reasonable weight allocation.

[0099] The fitness function guides the evolutionary direction, and negative mean squared error is used as the evaluation criterion.

[0100]

[0101] in, The sample size of the test set. For the true value, The predicted values ​​are for the LS-SVM model. This design allows the algorithm to automatically minimize the prediction error while avoiding overfitting through evaluation on the test set.

[0102] A tournament selection strategy is employed, where three individuals are randomly selected from the population each time, and the one with the highest fitness is retained. This strategy maintains population diversity while ensuring the transmission of superior genes.

[0103] Crossover operation: Kernel function type uses swap crossover, directly swapping the parent kernel function selection; regularization parameters use arithmetic crossover.

[0104]

[0105] in, The result is a uniformly random number in the range [0,1]. Mutation operation: The kernel function mutates randomly with a 50% probability, and the regularization parameter is randomly generated again in the range [0.1,10] with a mutation probability set to 20% to ensure sufficient exploration capability.

[0106] An advanced (μ+λ) evolutionary strategy is employed: u = 10 parent population size, λ = 20 offspring population size, with a termination condition of a maximum of 50 generations. This strategy preserves the best individuals in each generation while enhancing search capability through a larger offspring population size. Experiments show that this configuration converges to a satisfactory solution within a relatively small number of generations.

[0107] Step S30: When the probability of fire risk is greater than the preset risk threshold, the location information, environmental data and the probability of fire risk are combined to obtain the initial fire screening information, and the initial fire screening information is uploaded to the cloud server so that the cloud server can integrate satellite remote sensing data to complete the fire confirmation.

[0108] It should be noted that the preset risk threshold is a pre-set risk value. When the calculated probability of fire exceeds this value, it is considered that there may be a risk of wildfire.

[0109] The initial fire screening information is a combination of location information, environmental data, and fire risk probability information, used to make a preliminary judgment on whether a wildfire exists.

[0110] A cloud server is a server used to receive and process data uploaded from various monitoring points, and it has powerful computing and storage capabilities.

[0111] In practice, the cloud platform first receives abnormal sensor data uploaded from the edge, and simultaneously uses infrared and visible light remote sensing images from satellites to scan fire hotspots over a wide area. This data is then combined and analyzed with ground sensor data to improve the accuracy of identification and address the problem of insufficient satellite identification of small fires.

[0112] By deeply integrating BeiDou navigation, low-orbit satellite communication, and satellite remote sensing technologies, a "three-in-one" collaborative monitoring closed loop is constructed. First, all data acquisition nodes are equipped with BeiDou positioning sensors, reporting their own positions with meter-level accuracy in real time, forming a dynamic electronic fence and enabling manageable and controllable equipment. When the ARM edge processor built into the aggregation node analyzes the data transmitted from the monitoring nodes and triggers a fire warning, it immediately transmits the fire point coordinates, combined with data from temperature, humidity, and smoke sensors, directly back via low-orbit satellite communication protocol. Satellite remote sensing data is simultaneously integrated, performing hotspot scanning verification of the warning area. AI algorithms intelligently fuse remote sensing data and ground sensor data to analyze the probability of wildfire occurrence, forming a closed-loop process from "precise positioning by ground sensors to edge processor analysis to real-time transmission via low-orbit satellite communication to cross-verification of remote sensing images." This fusion mechanism enables rapid confirmation of fire location, reduces false alarm rates, and significantly improves the response speed and reliability of the monitoring system.

[0113] Furthermore, the step of the cloud server fusing satellite remote sensing data to confirm the fire situation also includes:

[0114] The satellite remote sensing images and the initial fire screening information are input into the CNN-MLP hybrid model. The CNN branch uses depthwise separable convolution and spatial attention to extract spatial features, while the MLP branch uses LSTM and self-attention to extract temporal features. The spatial and temporal features are then fused through dynamic weights to output the fire risk probability and fire point segmentation map.

[0115] The CNN-MLP hybrid model is jointly optimized using a weighted multi-task loss function, and the optimized model parameters are then sent to the least squares support vector machine model to update its parameters.

[0116] like Figure 3 As shown, Figure 3This diagram illustrates the architecture of the CNN-MLP hybrid model, showcasing the hardware components of the monitoring and aggregation nodes: Monitoring Node: BeiDou module, LoRa communication, sensors (temperature, humidity / smoke), solar power. Aggregation Node: ARM edge processor, satellite communication module, multi-stage voltage regulator circuit. Cloud Server: GPU cluster and distributed storage. Starting from two parallel links—satellite remote sensing imagery and ground sensor time-series data—the satellite imagery on the left undergoes deep separable convolution and spatial attention mechanisms via a CNN branch to quickly locate thermal anomaly areas and suppress background noise, generating spatial features. The ground sensors on the right feed one-dimensional time-series data such as temperature, humidity, and smoke concentration into an LSTM and self-attention layer to capture abrupt changes and long-term dependencies, outputting time-series features. The two features are then adaptively weighted and merged through a dynamic weight fusion mechanism, preserving both spatial positioning accuracy and temporal evolution information. Finally, the CNN-MLP hybrid model synchronously outputs fire hazard probability and fire point segmentation maps, achieving integrated wildfire identification through large-scale satellite remote sensing scanning and fine-grained ground node perception.

[0117] In the implementation, a lightweight EfficientNet is used as the backbone network, combined with depthwise separable convolutions to reduce computation, and a spatial attention mechanism is used to enhance the response of fire regions, extracting spatial features of fire regions from satellite images. The depthwise separable convolution method is as follows:

[0118]

[0119] The spatial attention mechanism works as follows:

[0120]

[0121] =

[0122] Where I is the input feature map. DepthwiseConv (channel-wise convolution): Performs spatial convolution on each channel of the input separately to extract the spatial information of that channel. PointwiseConv (pointwise convolution): Linearly combines the outputs of the channel-wise convolutions along the channel dimension using a 1×1 convolution, fusing information and adjusting the number of channels. ⊕ indicates concatenating or combining the results of both. Depthwise separable convolution decomposes standard convolution into channel-wise and pointwise convolutions, reducing computation to 1 / 8 to 1 / 9. The spatial attention mechanism generates an attention map through global average pooling and 1×1 convolutions, highlighting important regions.

[0123] LSTM is used to capture long-term dependencies, and a self-attention mechanism is combined to enhance the feature representation of key time points, which is used to model the temporal dependencies of sensor data.

[0124] LSTM unit:

[0125]

[0126] Self-attention:

[0127]

[0128] LSTM controls the flow of information through input gates and forget gates to solve the problem of long-term dependency. The self-attention mechanism uses query-key-value calculation to dynamically allocate weights at different time points.

[0129] By using learnable attention weights, the contribution ratios of CNN and MLP branches can be dynamically adjusted, and spatial and temporal features can be adaptively integrated.

[0130] Modal importance weighting, feature fusion:

[0131]

[0132]

[0133] The weight α is generated by a fully connected layer + Sigmoid, and its range is [0,1]. It integrates spatial and temporal information by fusing features. The larger the α is, the more important the spatial features are.

[0134] The main task is to predict the fire risk probability, while the auxiliary task is to generate a fire point segmentation map to improve the quality of spatial features and output the fire risk probability simultaneously.

[0135] Fire risk probability regression:

[0136]

[0137] The fire risk probability is mapped to the [0,1] interval using Sigmoid. The segmentation task uses transposed convolution upsampling, and the supervised CNN branch learns more accurate spatial features.

[0138] In one feasible implementation, step S30 includes steps A11 to A15:

[0139] A11: Encode the location information to obtain the first data frame;

[0140] It should be noted that the encoding process involves converting data into a format suitable for transmission.

[0141] A data frame is an encapsulated data unit that contains a specific type of information.

[0142] A12: Encode the environmental data to obtain the second data frame;

[0143] A13: Encode the fire risk probability to obtain the third data frame;

[0144] A14: Encapsulate the first data frame, the second data frame, and the third data frame to obtain the initial fire screening information;

[0145] It should be noted that the encapsulation process combines multiple data frames into a complete data packet.

[0146] A15: Upload initial fire information to the cloud server via a low-orbit satellite link.

[0147] It should be noted that communication links using low-Earth orbit satellites for data transmission are suitable for remote areas.

[0148] Further, step A15 includes:

[0149] The initial fire information is divided into a preset number of data packets according to the preset packet length;

[0150] Add a checksum and timestamp to each data packet, and send each data packet with the added checksum and timestamp to the cloud server in sequence;

[0151] If no confirmation frame is received from the cloud server, the corresponding data packet is retransmitted.

[0152] The steps to enable cloud servers to integrate satellite remote sensing data for fire confirmation include:

[0153] Send a remote sensing data request command to the cloud server;

[0154] Receive the fire confirmation result returned by the cloud server, which includes the location of the fire and the fire level;

[0155] Write the fire confirmation result to the local fire status flag;

[0156] The fire status is broadcast to other monitoring nodes in the LoRa network.

[0157] In the implementation, the initial fire information is divided into multiple data packets, each with a checksum and timestamp added before being sent sequentially. If no acknowledgment frame is received, the corresponding data packet is retransmitted. This is a mechanism to ensure reliable data transmission.

[0158] It should be noted that the remote sensing data request command is a request sent to the cloud server to obtain satellite remote sensing data.

[0159] The fire confirmation result is the result obtained by the cloud server based on a comprehensive judgment of satellite remote sensing data and uploaded preliminary fire information, including the location of the fire and the fire level.

[0160] Local fire status flags store fire status confirmation results locally for recording and tracking fire status.

[0161] LoRa networking is a low-power wide-area network technology used for data transmission between monitoring nodes.

[0162] Furthermore, this solution also includes steps A21 to A23:

[0163] A21: Receive model update instructions from the cloud server, which include the latest model parameters;

[0164] It should be noted that the model update command is a command generated by the cloud server based on new data or optimized model parameters, and is used to update the model at the edge.

[0165] The latest model parameters are those that have been retrained or optimized in the cloud, and these parameters can better reflect the current monitoring environment and fire characteristics.

[0166] In practice, the edge device listens for model update commands sent from the cloud in real time through a communication link (such as WebSocket) established with the cloud server. When a command is received, the edge device parses the command content and extracts the latest model parameters.

[0167] A22: The least squares support vector machine model is hot-updated based on the latest model parameters to obtain the updated model;

[0168] It should be noted that hot update is the process of dynamically updating model parameters without shutting down the system or affecting normal operation. This ensures the real-time performance and continuity of the monitoring system.

[0169] In practical implementation, at the edge, the parameters of the least squares support vector machine model are replaced or adjusted based on the latest received model parameters. This process is usually achieved through model deserialization or parameter injection. Specifically, the new model parameters are directly assigned to the corresponding parameter variables of the model without re-initializing the entire model.

[0170] A23: The updated fire risk probability is obtained by processing the location information and environmental data through the updated model.

[0171] It should be noted that the updated model, namely the least squares support vector machine model after hot update, has a more accurate fire prediction capability.

[0172] The updated fire risk probability, which is the fire risk probability recalculated using the updated model, can more accurately reflect the current fire risk status at the monitoring point.

[0173] In practice, new location information and environmental data are input into the updated model. The model then calculates the updated fire risk probability based on the new parameters. This process is similar to the initial calculation of the fire risk probability, but it uses the latest model parameters, thereby improving the accuracy and reliability of the prediction.

[0174] Understandably, the optimized LS-SVM model described above is deployed on an ARM processor to process and analyze sensor data from monitoring nodes, perform initial screening of wildfire risks, and classify wildfires as occurring when the risk probability is greater than 0.5. The values ​​of the node's BeiDou positioning, temperature and humidity, and smoke sensors are then uploaded to the cloud via a low-orbit satellite terminal.

[0175] It should be understood that this invention constructs a space-ground collaborative intelligent monitoring system for power and wildfires based on communication, navigation, and remote sensing integration. It adopts a hybrid communication network that integrates low-orbit satellite internet and ground-based LoRa self-organizing network to ensure the monitoring needs of power and wildfires along transmission lines in remote areas. By introducing BeiDou navigation, ground sensor acquisition nodes can be accurately located. Ground sensor aggregation nodes are equipped with edge processors to analyze the data sent from the acquisition nodes. Key data and preliminary analysis results are transmitted to the cloud via low-orbit satellite terminals. The cloud uses a multi-source heterogeneous data intelligent fusion algorithm to analyze the two-dimensional spatial imagery of satellite remote sensing and the one-dimensional time-series data of ground sensors to achieve high-precision early fire identification.

[0176] This invention effectively overcomes the problems of limited communication coverage, wasted communication resources, insufficient data fusion, and high false alarm rates in traditional power grid wildfire monitoring systems. Through the described system and method, full-coverage monitoring of power grid wildfires along transmission corridors in remote areas can be achieved, eliminating monitoring blind spots; the accuracy of early fire detection can be significantly improved, and the speed of early warning response can be accelerated; and the stable and reliable transmission of critical power grid wildfire monitoring data can be ensured. Therefore, the power grid's ability to prevent and control wildfire risks is significantly enhanced, and the safety and stability of power grid operation are strengthened, demonstrating significant engineering application value.

[0177] This embodiment provides a method for intelligent monitoring of power plant wildfires based on communication, navigation, and remote sensing integration. It constructs a hybrid communication network integrating low-orbit satellite internet and terrestrial self-organizing networks to address the insufficient coverage of traditional communication. By introducing BeiDou navigation, it solves the problem of inaccurate fire location, achieving blind-spot-free wildfire monitoring along power transmission lines in remote areas and ensuring real-time transmission of fire data and accurate fire location. It introduces a cloud-edge collaborative hierarchical processing mechanism and edge algorithms: through collaborative computing between cloud and edge devices, intelligent processing algorithms deployed at the edge perform data preprocessing and preliminary analysis, solving the aforementioned problems of communication resource waste, data transmission congestion, and data processing efficiency at ground nodes, ensuring real-time transmission of critical fire information. Finally, it develops a multi-source heterogeneous data intelligent fusion algorithm: by fusing heterogeneous data from ground-based sensor monitoring and satellite remote sensing in the cloud, it improves the accuracy and comprehensiveness of fire identification and early warning, addressing the problem of insufficient accuracy from a single data source.

[0178] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 Step S20 includes steps S201 to S203:

[0179] Step S201: Calculate the position deviation from the historical monitoring nodes based on the location information to obtain the position deviation value;

[0180] It should be noted that location information refers to the geographical location data of the monitoring nodes obtained through the BeiDou positioning system.

[0181] The location information of historical monitoring nodes refers to the geographical location data of past monitoring nodes stored in the system, which is used to compare the location of the current node.

[0182] The location deviation value is the difference in location between the current monitoring node and the historical monitoring nodes, and is used to assess the movement of the node or changes in the environment.

[0183] By using a coordinate calculation algorithm, the latitude and longitude of the current node are compared with those of historical nodes to calculate the positional deviation value.

[0184] It should be understood that when constructing a wildfire risk prediction model, we first need to transform the monitoring problem into a mathematical optimization problem. LS-SVM, through the principle of structural risk minimization, controls model complexity while ensuring model fitting accuracy, making it particularly suitable for small-sample scenarios involving edge devices. Its core idea is to map sensor data (a 3-dimensional feature vector x_i of temperature, humidity, and smoke concentration) to a high-dimensional feature space through a kernel function, constructing the optimal separating hyperplane in this space. Unlike traditional SVM, LS-SVM uses a least-squares loss function, transforming inequality constraints into equality constraints, making the problem a solution to a system of linear equations, greatly reducing computational complexity.

[0185] Specifically, a wildfire simulation data training sample set {(x_i,y_i)} is used, where y_i∈{-1,1} represents the probability of wildfire occurrence, and a convex optimization problem is solved:

[0186]

[0187]

[0188] Where W is the weight vector in the feature space, with the same dimension as the mapped feature space, b is the bias term, and ei is the slack variable for the i-th sample. Kernel functions are used to map data to a high-dimensional space. The regularization parameter controls the balance between model complexity and training error. The solution to this optimization problem will yield the optimal parameters for the classification hyperplane. .

[0189] in Kernel functions are used to map data to a high-dimensional space. The regularization parameter controls the balance between model complexity and training error. The solution to this optimization problem will yield the optimal parameters for the classification hyperplane. .

[0190] Step S202: Combine the temperature and humidity data, smoke concentration data, and location deviation values ​​into an input feature vector and map it to a high-dimensional feature space to obtain a mapping vector;

[0191] It should be noted that the input feature vector is a vector that combines temperature and humidity data, smoke concentration data, and location deviation values, and serves as the input to the model.

[0192] A high-dimensional feature space is a space that transforms the input feature vector into a higher dimension through a mapping function, thereby enhancing the expressive power of the model.

[0193] A mapping vector is a vector that represents the input features in a high-dimensional space.

[0194] Step S203: The kernel function value is calculated using support vectors and mapping vectors, and the fire risk probability is calculated based on the kernel function value and Lagrange multipliers.

[0195] It should be noted that support vectors are sample points that are determined during the training process and have an impact on the classification boundary.

[0196] The kernel function value is the similarity between the support vector and the mapping vector calculated using the kernel function.

[0197] Lagrange multipliers are parameters introduced in optimization problems to balance the complexity of the model and the training error.

[0198] The fire risk probability is the value output by the model that represents the likelihood of a fire occurring.

[0199] In the specific implementation, the optimal parameter combination is obtained. Then, the system will deploy a specially optimized lightweight LS-SVM model, whose prediction function is expressed as:

[0200]

[0201] in, The Lagrange multipliers obtained through solving the optimization problem reflect the importance weights of each support vector. The dynamic kernel function is optimized using a genetic algorithm. The global bias obtained by solving the system of linear equations ensures the correct positioning of the prediction results in the feature space.

[0202] In one feasible implementation, step S203 includes steps A31 to A33:

[0203] A31: The kernel function matrix is ​​weighted and summed with the Lagrange multipliers to obtain the weighted value;

[0204] It should be noted that the kernel function matrix contains the kernel function values ​​of all support vectors and input vectors.

[0205] It should be understood that weighted summation is to multiply each element in the kernel function matrix by the corresponding Lagrange multiplier and then sum them.

[0206] The weighted value is the result of weighted summation.

[0207] A32: Add the weighted value to the bias term to obtain the median score;

[0208] It should be noted that the bias term, or bias parameter in the model, is used to adjust the baseline of the model output.

[0209] The median score is the result of adding the weighted value and the bias term.

[0210] A33: Input the intermediate score into the activation function to obtain the fire risk probability.

[0211] It should be noted that an activation function is a function used to convert intermediate scores into probability values, such as the Sigmoid function.

[0212] The fire risk probability is the final output value representing the likelihood of a fire occurring, ranging from 0 to 1.

[0213] This embodiment provides an intelligent power line wildfire monitoring method based on communication, navigation, and remote sensing fusion. By calculating position deviations in real time, fusing multi-dimensional environmental data and mapping it to a high-dimensional space, and combining it with an optimized support vector machine model, it achieves accurate assessment of wildfire risks along transmission lines. Position deviation calculation enables the system to dynamically sense minute displacements or environmental abrupt changes in monitoring nodes, while high-dimensional mapping enhances the ability to capture complex environmental features. The model utilizes parameters pre-optimized by a genetic algorithm to ensure rapid and accurate output of fire risk probabilities on edge devices. The cloud-based collaborative dynamic model update mechanism further improves the system's adaptability and reliability, thereby constructing an efficient and intelligent power line wildfire monitoring system.

[0214] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the intelligent monitoring method for power and wildfires based on the integration of communication, navigation and remote sensing in this application. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0215] like Figure 5 As shown, Figure 5 This diagram illustrates the architecture of a space-ground integrated three-dimensional monitoring network, showcasing the overall architecture of a space-ground integrated power and wildfire monitoring system based on communication, navigation, and remote sensing fusion. The system constructs a space-ground integrated three-dimensional monitoring network, achieving blind-spot-free intelligent monitoring of power and wildfires along transmission corridors through deep fusion of communication, navigation, and remote sensing. It also demonstrates the workflow and edge processing algorithm of the cloud-edge collaborative layered processing mechanism: through the organic synergy of real-time initial screening at the edge layer and in-depth analysis in the cloud, the contradiction between response latency and computing resources in power and wildfire monitoring is resolved.

[0216] The system's power supply design utilizes solar panels with batteries to provide 12V, protected by fuses to prevent overload. The core component of the power supply section is the TPS54526 voltage regulator, which regulates the 12V voltage to 3.8V to supply the circuit modules of the low-Earth orbit satellite communication module. In addition, a TPS564201 voltage regulator provides 5V, a TPS7333QDRG4 voltage regulator provides 3.3V, and an AMS1117-1.8 voltage regulator provides 1.8V.

[0217] The data acquisition node terminal design includes a control module (low cost, low power consumption), a communication module (LoRa), a temperature and humidity sensor, a smoke sensor, and a BeiDou positioning sensor. The data acquisition node equipment supports multiple data transmission modes, including normal transmission mode, fixed-point transmission mode, and WOR mode, and possesses excellent low-power characteristics. The LoRa module's single-hop transmission distance can reach a maximum of 1 kilometer, and the communication range can be further extended through relay mode.

[0218] The aggregation node terminal design includes a control module (low cost, low power consumption), a communication module (LoRa), and an edge processor. The edge processor is equipped with a quad-core ARM Cortex-A76 processor with a main frequency of up to 2.4GHz, up to 8GB of LPDDR4X memory, and supports PCIe 3.0 interface and USB 3.0, providing faster data transfer capabilities and expansion performance. It also has an improved graphics processing unit, supporting 4K video decoding and dual display output, and can run lightweight intelligent algorithms.

[0219] The low-Earth orbit (LEO) satellite terminal design includes a control module and a wireless communication module that integrates LEO satellite mobile communication network and GNSS positioning functions. It provides satellite mobile communication services and GNSS positioning capabilities. The transmission rate is 0.6kbps-2.1kbps, with a transmit power consumption of 650mA@Avg and a receive power consumption of 350mA@Avg.

[0220] The deployment plan includes compute nodes equipped with NVIDIA A100 GPU servers, each server featuring four NVIDIA A100 graphics cards with 80GB of dedicated memory and 256GB of DDR4 ECC RAM. For storage, a Ceph distributed storage system is used, providing 1PB of available storage space.

[0221] Figure 6The system hardware design block diagram shows that the 12V output from the solar panel is fed into a multi-stage voltage regulator circuit after being protected by a fuse. The voltage is successively stepped down to 5V, 3.8V, 3.3V, and 1.8V to power the STM32G0 ultra-low power controller, MQ-2 smoke sensor, high-precision temperature and humidity sensor, and Beidou positioning module. The raw data collected by the sensor group is aggregated to the ARM edge processor via the LoRa module. The processor runs an LS-SVM model to perform an initial screening of wildfire risks. If the risk probability is >0.5, the Beidou coordinates, sensor data, and analysis results are packaged and uploaded to the low-orbit satellite via the low-orbit satellite terminal. Then, the ground gateway station forwards the data to the cloud. The cloud deploys a CNN-MLP hybrid model based on an NVIDIA A100×4 GPU cluster, integrates satellite remote sensing imagery to refine the fire risk probability, and outputs a fire point segmentation map. All data is finally written to Ceph distributed storage, realizing a closed-loop end-to-end process from edge perception to accurate cloud confirmation.

[0222] This application also provides a smart power fire monitoring device based on communication, navigation, and remote sensing fusion. Please refer to [reference needed]. Figure 7 The intelligent monitoring device for power field fires based on the integration of communication, navigation, and remote sensing includes:

[0223] The acquisition module 10 is used to acquire the location information of the first monitoring node, and to acquire the temperature and humidity data and smoke concentration data corresponding to the location information as environmental data;

[0224] Analysis module 20 is used to input location information and environmental data into the least squares support vector machine model for processing to obtain the fire risk probability;

[0225] The upload module 30 is used to combine location information, environmental data and fire risk probability to obtain preliminary fire screening information when the fire risk probability is greater than a preset risk threshold, and upload the preliminary fire screening information to the cloud server so that the cloud server can integrate satellite remote sensing data to complete the fire confirmation.

[0226] The intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion provided in this application, employing the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion in the above embodiments, can solve the technical problem of how to achieve blind-spot-free, low-latency, and high-precision early monitoring and location of power fires along transmission lines in remote mountainous areas without terrestrial communication coverage. Compared with the prior art, the beneficial effects of the intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion provided in this application are the same as those of the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion provided in the above embodiments, and other technical features in the intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0227] In one embodiment, the analysis module 20 is further configured to calculate the position deviation from historical monitoring nodes based on the position information, and obtain the position deviation value;

[0228] The temperature and humidity data, smoke concentration data, and location deviation values ​​are combined into an input feature vector and mapped onto a high-dimensional feature space to obtain a mapping vector.

[0229] The kernel function value is calculated using support vectors and mapping vectors, and the fire risk probability is calculated based on the kernel function value and Lagrange multipliers.

[0230] In one embodiment, the analysis module 20 is further configured to perform a weighted summation of the kernel function matrix and the Lagrange multipliers to obtain a weighted value;

[0231] The weighted value is added to the bias term to obtain the median score;

[0232] The intermediate score is input into the activation function to obtain the fire risk probability.

[0233] In one embodiment, the upload module 30 is further configured to encode the location information to obtain a first data frame;

[0234] The environmental data is encoded to obtain the second data frame;

[0235] The probability of fire risk is encoded to obtain the third data frame;

[0236] The first data frame, the second data frame and the third data frame are encapsulated to obtain the initial fire screening information;

[0237] Initial fire information is uploaded to a cloud server via a low-orbit satellite link.

[0238] In one embodiment, the upload module 30 is further configured to divide the initial fire screening information into a preset number of data packets according to a preset packet length;

[0239] Add a checksum and timestamp to each data packet, and send each data packet with the added checksum and timestamp to the cloud server in sequence;

[0240] If no confirmation frame is received from the cloud server, the corresponding data packet is retransmitted.

[0241] In one embodiment, the upload module 30 is further configured to send a remote sensing data request instruction to the cloud server;

[0242] Receive the fire confirmation result returned by the cloud server, which includes the location of the fire and the fire level;

[0243] Write the fire confirmation result to the local fire status flag;

[0244] The fire status is broadcast to other monitoring nodes in the LoRa network.

[0245] In one embodiment, the upload module 30 is further configured to receive a model update instruction sent by a cloud server, wherein the model update instruction includes the latest model parameters;

[0246] The least squares support vector machine model is hot-updated based on the latest model parameters to obtain the updated model.

[0247] The updated model is used to process location information and environmental data to obtain the updated fire risk probability.

[0248] This application provides a power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion. The power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion includes: at least one processor; and a memory communicatively connected to at least one processor; wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable at least one processor to execute the power fire intelligent monitoring method based on communication, navigation, and remote sensing fusion in the above embodiment 1.

[0249] The following is for reference. Figure 8 This document illustrates a structural schematic diagram of a power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion, suitable for implementing embodiments of this application. The power fire intelligent monitoring device based on communication, navigation, and remote sensing fusion in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The intelligent power fire monitoring device based on communication, navigation and remote sensing fusion shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0250] like Figure 8As shown, the intelligent power and wildfire monitoring device based on communication, navigation, and remote sensing fusion can include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in ROM (Read Only Memory) 1002 or the program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the intelligent power and wildfire monitoring device based on communication, navigation, and remote sensing fusion. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the intelligent power and wildfire monitoring equipment based on communication, navigation, and remote sensing fusion to exchange data with other devices wirelessly or via wired means. Although the figure shows an intelligent power and wildfire monitoring equipment based on communication, navigation, and remote sensing fusion with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0251] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0252] The intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion provided in this application, employing the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion in the above embodiments, can solve the technical problem of how to achieve blind-spot-free, low-latency, and high-precision early monitoring and location of power fires along transmission lines in remote mountainous areas without terrestrial communication coverage. Compared with the prior art, the beneficial effects of the intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion provided in this application are the same as those of the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion provided in the above embodiments, and other technical features in this intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0253] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0254] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0255] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the intelligent monitoring method for power wildfires based on communication, navigation and remote sensing fusion in the above embodiments.

[0256] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0257] The aforementioned computer-readable storage medium may be included in a power fire intelligent monitoring device based on communication, navigation, and remote sensing integration; or it may exist independently and not be assembled into a power fire intelligent monitoring device based on communication, navigation, and remote sensing integration.

[0258] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion, the intelligent power fire monitoring device based on communication, navigation, and remote sensing fusion will: acquire the location information of the first monitoring node, and acquire the corresponding temperature and humidity data and smoke concentration data as environmental data; input the location information and environmental data into a least squares support vector machine model for processing to obtain the fire risk probability; when the fire risk probability is greater than a preset risk threshold, combine the location information, environmental data, and fire risk probability to obtain preliminary fire screening information, and upload the preliminary fire screening information to a cloud server so that the cloud server can fuse satellite remote sensing data to complete the fire confirmation.

[0259] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0260] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0261] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0262] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion. This solves the technical problem of achieving blind-spot-free, low-latency, and high-precision early monitoring and location of power fires along transmission lines in remote mountainous areas without terrestrial communication coverage. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the intelligent power fire monitoring method based on communication, navigation, and remote sensing fusion provided in the above embodiments, and will not be repeated here.

[0263] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described intelligent monitoring method for power and wildfires based on communication, navigation, and remote sensing fusion.

[0264] The computer program product provided in this application can solve the technical problem of how to achieve early monitoring and location of power-related wildfires along transmission lines in remote mountainous areas without terrestrial communication coverage, with no blind spots, low latency, and high accuracy. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the intelligent power-related wildfire monitoring method based on communication, navigation, and remote sensing fusion provided in the above embodiments, and will not be repeated here.

[0265] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A power forest fire intelligent monitoring method based on through-conduction remote fusion, characterized in that, The method comprises: obtaining position information of a first monitoring node, and obtaining temperature and humidity data and smoke concentration data corresponding to the position information as environmental data; inputting the position information and the environmental data into a least squares support vector machine model for processing to obtain a fire risk probability; when the fire risk probability is greater than a preset risk threshold, combining the position information, the environmental data and the fire risk probability to obtain fire preliminary screening information, and uploading the fire preliminary screening information to a cloud server to enable the cloud server to fuse satellite remote sensing data to complete fire confirmation; the step of inputting the position information and the environmental data into a least squares support vector machine model for processing to obtain a fire risk probability comprises: calculating a position deviation from a historical monitoring node according to the position information to obtain a position deviation value; combining the temperature and humidity data, the smoke concentration data and the position deviation value into an input feature vector to map to a high-dimensional feature space to obtain a mapping vector; calculating a kernel function value through a support vector and the mapping vector, and calculating a fire risk probability according to the kernel function value and a Lagrange multiplier; the step of the cloud server fusing satellite remote sensing data to complete fire confirmation further comprises: inputting satellite remote sensing images and the fire preliminary screening information into a CNN-MLP hybrid model, wherein a CNN branch uses a depth separable convolution and a spatial attention to extract spatial features, an MLP branch uses an LSTM and a self-attention to extract temporal features, and the spatial features and the temporal features are fused through dynamic weights to output a fire risk probability and a fire point segmentation map.

2. The method of claim 1, wherein, the step of calculating a fire risk probability according to the kernel function value and a Lagrange multiplier comprises: performing weighted summation processing on a kernel function matrix and a Lagrange multiplier to obtain a weighted value, the kernel function matrix containing a matrix of all support vectors and input vector kernel function values; adding the weighted value and a bias term to obtain an intermediate score; inputting the intermediate score into an activation function to obtain a fire risk probability.

3. The method of claim 1, wherein, the step of combining the position information, the environmental data and the fire risk probability to obtain fire preliminary screening information when the fire risk probability is greater than a preset risk threshold, and uploading the fire preliminary screening information to a cloud server comprises: encoding the position information to obtain a first data frame; encoding the environmental data to obtain a second data frame; encoding the fire risk probability to obtain a third data frame; packaging the first data frame, the second data frame and the third data frame to obtain fire preliminary screening information; uploading the fire preliminary screening information to a cloud server through a low-orbit satellite link.

4. The method of claim 3, wherein, the step of uploading the fire preliminary screening information to a cloud server through a low-orbit satellite link comprises: dividing the fire preliminary screening information into a preset number of data packets according to a preset packet length; adding a check code and a timestamp to each data packet, and sequentially sending each data packet with the added check code and timestamp to the cloud server; when no confirmation frame returned by the cloud server is received, retransmitting the corresponding data packet.

5. The method of claim 1, wherein, The step of fusing satellite remote sensing data to complete fire confirmation by the cloud server comprises: sending a remote sensing data request instruction to the cloud server; receiving a fire confirmation result returned by the cloud server, wherein the fire confirmation result contains a fire point position and a fire level; writing the fire confirmation result into a local fire state marker; broadcasting the fire state marker to other monitoring nodes in the LoRa network.

6. The method of claim 1, wherein, The method further comprises: receiving a model update instruction issued by the cloud server, wherein the model update instruction includes the latest model parameters; performing a hot update on the least squares support vector machine model based on the latest model parameters to obtain an updated model; processing the position information and the environmental data through the updated model to obtain an updated fire risk probability.

7. A power forest fire intelligent monitoring device based on the remote fusion of the pass-through, characterized in that, The device comprises: an acquisition module configured to acquire position information of a first monitoring node, and acquire temperature and humidity data and smoke concentration data corresponding to the position information as environmental data; an analysis module configured to input the position information and the environmental data into a least squares support vector machine model for processing to obtain a fire risk probability, and further configured to calculate a position deviation from historical monitoring nodes based on the position information to obtain a position deviation value, combine the temperature and humidity data, the smoke concentration data, and the position deviation value into an input feature vector to map to a high-dimensional feature space to obtain a mapping vector, calculate a kernel function value through a support vector and the mapping vector, and calculate a fire risk probability based on the kernel function value and a Lagrange multiplier; an upload module configured to, when the fire risk probability is greater than a preset risk threshold, combine the position information, the environmental data, and the fire risk probability to obtain fire preliminary screening information, and upload the fire preliminary screening information to a cloud server to enable the cloud server to fuse satellite remote sensing data to complete fire confirmation, and further configured to input satellite remote sensing images and the fire preliminary screening information into a CNN-MLP hybrid model, wherein a CNN branch uses a depth separable convolution and a spatial attention to extract spatial features, an MLP branch uses an LSTM and a self-attention to extract temporal features, and the spatial features and the temporal features are fused through dynamic weights to output a fire risk probability and a fire point segmentation map.

8. A power forest fire intelligent monitoring device based on the remote fusion of the pass-through, characterized in that, The device comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the power forest fire intelligent monitoring method based on the fusion of remote sensing data and satellite remote sensing data according to any one of claims 1 to 6.

9. A storage medium, characterized by The storage medium is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the power forest fire intelligent monitoring method based on the fusion of remote sensing data and satellite remote sensing data according to any one of claims 1 to 6.

Images