Cross-platform remote power transmission channel forest fire monitoring and early warning system and method

Abstract

The application provides a cross-platform remote power transmission channel forest fire monitoring and early warning system and method, relates to the technical field of power systems, and comprises an adaptive sensing network unit, a data processing unit, a visual display unit, a fire situation early warning unit and a collaborative decision-making module. Through the deployment of multi-modal sensor nodes, various environmental parameters such as temperature, humidity, wind speed, soil humidity, light intensity and air pressure can be collected in real time. Information fusion algorithms are used to weight and integrate these data, generate a comprehensive environmental risk index, and thus provide more comprehensive and accurate environmental monitoring.

Description

Technical Field

[0001] This invention relates to the field of power system technology, specifically to a cross-platform remote transmission channel wildfire monitoring and early warning system and method. Background Technology

[0002] Wildfire monitoring is a crucial foundational task for ensuring the safe and stable operation of power transmission channels. Wildfire monitoring of power transmission channels can be mainly divided into monitoring methods based on satellite remote sensing technology and monitoring methods based on distributed wireless sensor networks (WSNs). The main detection methods in near-field terminal remote real-time line monitoring and wildfire monitoring and early warning of power transmission channels include smoke detection, video surveillance, infrared temperature detection, and lidar monitoring using a network of sensor terminals deployed along the line. Satellite remote sensing-based detection methods are low-cost, highly efficient, and can detect large areas and wide ranges; however, they are limited by satellite orbital rules and cannot perform real-time detection, and are significantly affected by weather conditions. Sensor network detection methods offer better real-time performance and comprehensiveness, but are costly and difficult to implement, and are only suitable for key prevention in high-risk fire areas by deploying high-precision, long-distance detectors (such as lidar detectors).

[0003] According to the Chinese Patent Publication No. CN112802287B, a system and method for monitoring, early warning, and locating wildfires on power transmission lines includes the following steps: S1. Determine whether there is a possibility of a fire based on remote sensing data from geostationary satellites. If so, locate the fire site using a remote sensing image target positioning algorithm and execute step S2; S2. Drive at least one monitoring drone to the fire site; S3. Acquire the images transmitted by the monitoring drone and display them on a GIS monitoring platform; S4. Confirm the possibility of a fire based on the images, and if a fire is confirmed, circle the fire area based on the images.

[0004] According to the method for controlling the risk of power transmission line tripping due to wildfires, which integrates meteorological and artificial fire prevention measures, as disclosed in Chinese Patent Publication No. CN105243459A, the method includes the following steps: First, historical wildfire disaster monitoring data near the power transmission line corridor is obtained through a power transmission line wildfire monitoring and early warning system. Based on the obtained data, a power transmission line wildfire tripping probability model PR considering precipitation factors, a power transmission line wildfire tripping probability model PT considering vegetation conditions, a power transmission line wildfire tripping probability model PF considering wildfire distance and fire prevention measures, and a power transmission line wildfire tripping probability model PV considering breakdown type are established. Finally, a comprehensive probability model for power transmission line wildfire tripping is obtained, P = PR·PT·PF·PV. Corresponding control measures are taken based on the comprehensive probability of power transmission line wildfire tripping obtained from real-time monitoring to cope with power transmission line wildfire tripping accidents.

[0005] The aforementioned patent documents and prior art have the following technical problems when used:

[0006] 1. Traditional fire monitoring systems often rely on a single parameter such as temperature or smoke concentration, resulting in a narrow data collection coverage and low sensitivity in fire monitoring.

[0007] 2. The early warning mechanism of traditional fire monitoring systems is mostly based on threshold judgment. An alarm will only be triggered when a certain parameter exceeds a predetermined threshold, which is prone to missed alarms and false alarms. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a cross-platform remote power transmission channel wildfire monitoring and early warning system, which solves the problems of low sensitivity and the tendency for missed or false alarms in traditional fire monitoring systems.

[0009] Another objective of this invention is to provide a cross-platform remote power transmission channel wildfire monitoring and early warning method.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A cross-platform remote power transmission channel wildfire monitoring and early warning system includes:

[0012] An adaptive sensor network unit is used to collect environmental parameters and weight and integrate the collected environmental parameters to obtain integrated data.

[0013] A data processing unit is used to verify, store, and classify the received integrated data;

[0014] The fire early warning unit is used to integrate and train stored historical fire data with received integrated data to form a training dataset and output the generated potential fire prediction data.

[0015] The visualization unit is used to display the integrated data and potential fire prediction data, and generate a fire risk assessment interface;

[0016] The collaborative decision-making unit is used to collaboratively calculate the integrated data, generate dynamically adjusted monitoring strategies and early warning thresholds, and feed the calculation results back to the adaptive sensor network unit.

[0017] Cross-platform remote power transmission channel wildfire monitoring and early warning methods include:

[0018] Collect environmental parameters, and then weight and integrate the collected environmental parameters to obtain integrated data;

[0019] The integrated data is then verified, stored, and categorized.

[0020] The stored historical fire data and the received integrated data are integrated and trained to form a training dataset, and the generated potential fire prediction data is output.

[0021] Display the integrated data and potential fire prediction data to generate a fire risk assessment interface;

[0022] The integrated data is collaboratively calculated to generate dynamically adjusted monitoring strategies and early warning thresholds, and the calculation results are fed back for information sharing.

[0023] The present invention has the following beneficial effects:

[0024] 1. This invention, by deploying multimodal sensor nodes, can collect various environmental parameters in real time, including temperature, humidity, wind speed, soil moisture, light intensity, and air pressure. An information fusion algorithm is used to weight and integrate these data to generate a comprehensive environmental risk index, thereby providing more comprehensive and accurate environmental monitoring.

[0025] 2. This invention employs a self-organizing network protocol based on a bio-inspired algorithm. The adaptive sensor network can dynamically adjust its topology according to environmental changes and node health status. This mechanism enables the system to self-repair and reconfigure network connections even when sensor nodes fail, ensuring system continuity and stability.

[0026] 3. The fire early warning unit of this invention uses a generative adversarial network (GAN) to perform in-depth analysis of historical and real-time data, enabling it to accurately predict potential fire scenarios and generate timely early warning signals. This approach improves the accuracy of fire prediction and the timeliness of early warning. Attached Figure Description

[0027] Figure 1 This is a schematic block diagram illustrating the workflow of the method of the present invention;

[0028] Figure 2 This is a block diagram of the system modules of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Example 1

[0031] As shown in the figure, the cross-platform remote power transmission channel wildfire monitoring and early warning system includes the following components:

[0032] The adaptive sensor network unit comprises sensors, each constituting a sensor node. These sensor nodes also include multimodal sensors for environmental parameter acquisition, including sensors for temperature, humidity, wind speed, soil moisture, light intensity, and vegetation type. The multimodal sensors use an information fusion algorithm to weight and integrate these parameters. Each sensor node collects environmental parameters, including wind speed, humidity, temperature, and air pressure. The sensor nodes of the adaptive sensor network unit employ a self-organizing network protocol based on a bio-inspired algorithm to achieve dynamic topology reconfiguration. This protocol automatically adjusts the connections between nodes based on environmental changes and node health status, optimizing data transmission paths and improving the network's anti-interference capabilities. The adaptive sensor network unit also includes a self-healing module with built-in machine learning algorithms for automatic network topology reconfiguration when sensor nodes fail.

[0033] Multimodal sensors are used to collect environmental parameters, including temperature (T), humidity (H), wind speed (W), soil moisture (SW), light intensity (LI), and vegetation type (VT).

[0034] Information fusion algorithm: Weighted integration formula:

[0035]

[0036] E: Integrated environmental parameter index after fusion, w i The weight of the i-th parameter reflects its importance in fire risk assessment and is usually determined through expert evaluation or historical data analysis; p i The i-th environmental parameter value collected; this formula uses weight normalization to ensure that the comprehensive index is between 0 and 1, which facilitates subsequent decision-making;

[0037] Self-organizing network protocols employ dynamic topology reconfiguration: using a biologically inspired self-organizing network protocol, connections are automatically adjusted based on environmental changes and node health status, with an improved transition probability formula.

[0038]

[0039] P ij : The transition probability from node i to node j; τ ij Path strength (pheromone), representing the amount of information transmitted in the past; η ij Heuristic factors (such as the reciprocal of distance) reflect the ease of connection. The health status of node j assesses the node's battery level and operational status; a higher value indicates a healthier node. The health status of node j assesses the node's power and operating status. A higher value indicates a healthier node. α, β, and γ are adjustment parameters with values ​​between -1 and 1, reflecting the degree of influence of each factor on the transition probability. These parameters need to be optimized through experiments.

[0040] The self-healing module employs machine learning algorithms: specifically, it uses ensemble learning (such as random forest) to achieve self-healing when nodes fail. The node reconfiguration formula is as follows:

[0041]

[0042] R j : The target node after reconfiguration, used to replace the failed node. K: The set of candidate nodes, all available nodes. Confidence k The reliability score of node k is output by a machine learning model, based on historical data and real-time performance evaluation.

[0043] The data processing unit is built into each sensor node. It processes the data collected by the sensor node and verifies the data through blockchain. The data processing unit has a built-in multi-level ensemble learning framework that combines random forest, support vector machine and deep learning model to extract features and classify the data collected by the sensor node. It is used to identify fire risks and generate early warning strategies based on dynamic environmental changes.

[0044] Data processing and verification utilize blockchain data verification:

[0045] The data processing unit processes and verifies the collected data locally. Data hash formula:

[0046] Hash d =H(d)=H(T||D||ID)

[0047] Hash d H: Hash value of the data record. H: Hash function, ensuring data immutability. T: Timestamp, recording the data generation time. D: Data collected by the sensor. ID: Unique identifier for the sensor node.

[0048] Feature extraction and classification employ an ensemble learning framework:

[0049] Combining random forest, support vector machine, and deep learning models for feature extraction and classification; random forest feature selection formula.

[0050]

[0051] F: Overall feature importance score; M: Number of trees in the forest; N m: Number of samples in the m-th tree; G: Loss function (Gini index) for each tree, used to evaluate classification performance; f m (x i The m-th tree is paired with the input feature x. i The prediction, y i : The output value of the loss function.

[0052] The fire early warning unit has a built-in generative adversarial network (GAN). The GAN is used to predict potential fire scenario data and improve the generalization ability of the fire prediction model. The GAN of the fire early warning unit is trained with heterogeneous data, including historical fire records, real-time monitoring data and meteorological data, to generate multi-dimensional potential fire predictions.

[0053] Generative Adversarial Networks (GANs) for Fire Prediction Models:

[0054]

[0055] G(z): Generator, which generates potential fire scenario data based on random noise z; The generated prediction data; D(y) and D(y) are the discrimination results of generated data and real data, respectively, used to optimize the model.

[0056] The visualization unit has a built-in augmented reality module that displays monitoring data and generates a fire risk assessment interface. The visualization unit also includes an augmented reality module and a virtual reality module. The augmented reality module overlays real-time monitoring data onto the monitoring interface, while the virtual reality module provides a risk simulation environment. The visualization unit includes risk area distribution, historical fire trend analysis, and future early warning models.

[0057] Augmented Reality Output:

[0058] AR display =Overlay(E,R)

[0059] AR display Augmented reality display content; E: Environmental parameter fusion results; R: Real-time monitoring data;

[0060] Virtual Reality Simulation:

[0061] VR simulation =Simulate(R,H)

[0062] VR simualtion Fire risk simulation in virtual reality; R: real-time data; H: historical fire data used for simulation scenarios.

[0063] Collaborative Decision-Making Module: This module incorporates a swarm intelligence algorithm to enable information sharing and collaborative decision-making among sensor nodes. It dynamically adjusts monitoring strategies and early warning thresholds. The module utilizes a swarm intelligence algorithm for distributed intelligent decision-making by the sensor nodes. Sensor nodes evaluate shared information using adaptive algorithms and dynamically adjust monitoring strategies and early warning thresholds. The collaborative decision-making module employs a swarm intelligence algorithm.

[0064] Definition: Swarm intelligence algorithms simulate the behavior of groups in nature and are used for distributed intelligent decision-making;

[0065] Information sharing mechanism: Information update formula:

[0066] I ij (t+1)=I ij (t)+β·(S h (t)-I ij (t))

[0067] I ij (t): The degree of trust that node i has in the information of node j at time t; S j (t): The actual state information of node j (such as environmental parameters); β: The learning rate, which controls the speed of information updates; This formula is used to dynamically adjust the trust level of information from other nodes based on the received information.

[0068] Adaptive algorithms dynamically adjust monitoring strategies and early warning thresholds:

[0069] Threshold adjustment formula:

[0070] T new =T old +α·(E current -E target )

[0071] T new New monitoring threshold; T old Old monitoring thresholds; E current Current environmental risk assessment value (provided by sensor nodes); E target : Target risk assessment value, usually set by the system; α: Adjustment factor, controlling the range of change in the threshold;

[0072] Distributed decision model decision output formula:

[0073]

[0074] D i : The final decision output of node i (such as issuing an early warning or adjusting the monitoring strategy); N i S is the set of neighboring nodes of node i. i: The self-state of node i (such as health status, data validity); f is a comprehensive function that makes decisions based on information from neighboring nodes and its own state;

[0075] Sensor nodes form a distributed intelligent network through a collaborative decision-making module, enabling information sharing and strategy adjustment under different environmental conditions.

[0076] Data flow:

[0077] Each sensor node collects data and generates information.

[0078] Nodes share state and trust levels through an information update mechanism.

[0079] The monitoring strategy and early warning thresholds are dynamically adjusted based on current environmental parameters.

[0080] The adaptive sensor network unit comprises multimodal sensor nodes, a microprocessor (MCU), a wireless communication module, a power supply module, and a self-healing module. The multimodal sensor nodes integrate temperature, humidity, wind speed, soil moisture, light intensity, and vegetation type sensors, enabling comprehensive collection of environmental parameters. These sensors are connected to the microprocessor, which performs data acquisition and preliminary processing, and establishes connections with other nodes via wireless communication modules (such as ZigBee, LoRa, or Wi-Fi). The power supply module includes a rechargeable battery and a solar power source, ensuring long-term stable operation of the nodes. The self-healing module incorporates health monitoring sensors to monitor the node status in real time and automatically triggers network topology reconfiguration when node failure is detected using machine learning algorithms, ensuring system reliability and integrity.

[0081] Multimodal sensor nodes collect environmental parameters (such as temperature, humidity, and wind speed) and transmit the collected data to a microprocessor for processing. The microprocessor uses an information fusion algorithm to weightedly integrate data from multiple sensors, generating a comprehensive environmental risk index. All sensor nodes automatically adjust their connections using a bio-inspired self-organizing network protocol, achieving dynamic topology reconfiguration. This mechanism ensures optimized data transmission paths and improves the network's anti-interference capabilities. A self-healing module continuously monitors the health status of the sensor nodes; if a node failure is detected, it immediately reconfigures the network topology using machine learning algorithms to ensure system stability.

[0082] Work steps:

[0083] Initialization: The sensor node is powered on and starts up, performs a self-test and establishes an initial network topology to ensure that all nodes can communicate normally.

[0084] Data acquisition: Multimodal sensor nodes continuously collect environmental parameters, such as temperature, humidity, wind speed, soil moisture, light intensity, and vegetation type.

[0085] Data fusion: After receiving sensor data, the microprocessor uses information fusion algorithms to weight and integrate the data to generate a comprehensive environmental risk index.

[0086] Data transmission: The processed data is transmitted to the data processing unit or other adjacent nodes via a wireless communication module, supporting real-time monitoring and collaborative work.

[0087] Self-organizing networks: Adaptive sensor networks, based on bio-inspired algorithms, achieve dynamic topology reconstruction, automatically adjust network connections, and ensure efficient data transmission and interference resistance.

[0088] Fault monitoring and repair: The self-repair module monitors the node status in real time. When a node failure is detected, it triggers network reconfiguration through machine learning algorithms to ensure stable system operation.

[0089] The data processing unit consists of a computing platform, storage devices, a network interface, and a security module. The computing platform provides computing power and supports various data processing and analysis algorithms. The storage devices store sensor data, model parameters, and historical records. The network interface enables high-speed communication with the sensor network, fire early warning unit, and visualization unit. The security module incorporates blockchain technology to ensure the authenticity and integrity of data transmission.

[0090] The data processing unit acquires real-time data from sensor nodes via a network interface from the adaptive sensing network unit and verifies the data using blockchain technology. The verified data is stored in a database and analyzed using a multi-layered ensemble learning framework (including random forests, support vector machines, and deep learning models). These models are capable of feature extraction and classification of sensor data, identifying fire risks, and generating dynamic early warning strategies.

[0091] Work steps:

[0092] Data reception: Receives data transmitted from the sensor network via the network interface.

[0093] Data verification: Blockchain technology is used to verify the integrity and authenticity of data, ensuring secure transmission.

[0094] Data storage: Validated data is stored in a database as the basis for analysis and prediction.

[0095] Feature extraction and classification: Through an ensemble learning framework, features are extracted from the data and classified to identify potential fire risks.

[0096] Early warning strategy generation: Based on the analysis results, the monitoring strategy is dynamically adjusted and an early warning strategy is generated and sent to the collaborative decision-making module and the visualization display unit.

[0097] The fire early warning unit includes computing devices, storage devices, and a network interface. The computing devices run a Generative Adversarial Network (GAN) model for training and prediction on data. The storage devices store model parameters, training data, and historical records. The network interface enables communication with the data processing unit and the collaborative decision-making module. The fire early warning unit uses a GAN to predict potential fire scenarios through deep learning training on heterogeneous data (including historical fire records, real-time monitoring data, and meteorological data). The GAN consists of a generator and a discriminator. The generator generates data samples of potential fire scenarios, and the discriminator evaluates the authenticity of the generated data. By continuously iterating and optimizing the model, the accuracy of fire prediction is improved. Finally, the early warning results are transmitted to the collaborative decision-making module and the visualization unit.

[0098] Work steps:

[0099] Data preparation: Collect and integrate historical fire records, real-time monitoring data and meteorological data to form a training dataset.

[0100] Model training: Deep learning training is performed using a GAN model. The generator generates potential fire scenario samples, and the discriminator evaluates the generation results and provides feedback to optimize the model.

[0101] Scenario prediction: Predict potential fire scenarios using a trained model and output risk assessment results.

[0102] Early warning output: The prediction results are transmitted to the collaborative decision-making module and the visualization display unit to generate real-time early warning information.

[0103] Model updates: Regularly retrain the model based on new data to ensure the timeliness and accuracy of prediction results.

[0104] The visualization unit consists of a monitor, computer or workstation, AR glasses or VR headset, and interactive devices. The monitor and workstation run the visualization software, supporting real-time data rendering and display. AR glasses or VR headsets are used for augmented and virtual reality displays, providing a more intuitive risk simulation environment. Interactive devices (such as a mouse, keyboard, touchscreen, or gesture recognition device) support real-time user control and operation.

[0105] The visualization unit uses augmented reality (AR) and virtual reality (VR) technologies to present monitoring data and analysis results to users in an intuitive way. The augmented reality module overlays real-time monitoring data onto the real environment, allowing users to directly see fire risk indicators on the user interface. The virtual reality module provides a risk simulation environment, allowing users to experience different fire scenarios through interactive devices, aiding in decision-making.

[0106] Work steps:

[0107] Data acquisition: Receive real-time monitoring data and analysis results from the data processing unit and the fire early warning unit.

[0108] Graphics rendering: Using augmented reality and virtual reality technologies to render data into graphics.

[0109] Interface Display: Display the fire risk assessment interface on a high-resolution display or AR / VR device, including real-time monitoring data, historical trends, and future early warning models.

[0110] User interaction: Users can adjust monitoring parameters in real time through interactive devices, or select different data views to view and further analyze the risk situation.

[0111] Feedback mechanism: User operations are fed back to the data processing unit, and the system optimizes the monitoring strategy and improves the subsequent monitoring and early warning process based on the feedback.

[0112] The collaborative decision-making module consists of distributed computing nodes, a communication module, and a storage module. The distributed computing nodes typically share hardware with the sensor nodes and possess both computing and communication capabilities. The communication module supports information sharing and synchronization between nodes. The storage module is used to store shared information, decision-making strategies, and historical data.

[0113] The collaborative decision-making module utilizes swarm intelligence algorithms, employing Particle Swarm Optimization (PSO) or ant colony optimization to achieve collaborative decision-making among nodes. Each sensor node exchanges environmental data and status information through an information-sharing mechanism, forming a global understanding. Based on this shared information, the collaborative decision-making module can dynamically adjust monitoring strategies and early warning thresholds, enabling the system to adapt to environmental changes and optimize monitoring efficiency. After receiving the prediction results from the fire early warning unit, the system adjusts its strategies based on node information and model analysis results, and feeds this information back to the sensor network, forming a closed loop.

[0114] Work steps:

[0115] Information collection: Each sensor node shares the collected data and status information with the collaborative decision-making module.

[0116] Swarm intelligence computing: Utilizing swarm intelligence algorithms to analyze shared information, perform collaborative computing, and generate optimal monitoring strategies and early warning thresholds.

[0117] Strategy Update: Update the calculation results to each node and adjust its monitoring behavior and early warning mechanism.

[0118] Real-time feedback: After a node executes a new strategy, it will feed back the execution results to the collaborative decision-making module to optimize the overall monitoring effect.

[0119] Continuous optimization: Through continuous information sharing and decision feedback, the monitoring strategy and early warning system are continuously adjusted to improve the system's adaptability and stability.

[0120] Overall system workflow:

[0121] Deployment phase: Adaptive sensor network units are deployed in key areas along the power transmission channel to initially establish network connections and topology.

[0122] Data acquisition and transmission: The adaptive sensor network unit acquires environmental data in real time and transmits it to the data processing unit through the wireless communication module.

[0123] Data processing and analysis: The data processing unit receives, verifies, and analyzes data to identify potential fire risks and generate early warning strategies.

[0124] Fire prediction and early warning: The fire early warning unit uses a GAN model to train on heterogeneous data, generate potential fire scenarios and provide risk assessment.

[0125] Collaborative Decision-Making and Adjustment: The collaborative decision-making module adjusts monitoring strategies and early warning thresholds in real time through information sharing and intelligent algorithms to optimize overall system performance.

[0126] Visualization and User Interaction: The visualization unit displays environmental monitoring data and fire prediction results in real time, providing operators with information for decision-making and control.

[0127] Feedback and Optimization: The system continuously optimizes itself through user feedback, node feedback, and data updates to improve monitoring and early warning effects, forming a complete closed loop. Specific Implementation Example 2

[0129] As shown in the figure, the cross-platform remote power transmission channel wildfire monitoring and early warning method has the following working steps:

[0130] Sp1: Deploy adaptive sensor network units in key areas of the power transmission channel. The adaptive sensor network unit consists of multiple sensor nodes, each of which is used to collect environmental parameters, including wind speed, humidity, temperature and air pressure.

[0131] Adaptive sensor network units are deployed in key areas along the power transmission corridor to achieve comprehensive environmental monitoring coverage. Each sensor node incorporates multimodal sensors, including resistance temperature detectors (RTDs) for temperature measurement, capacitive humidity sensors for humidity measurement, ultrasonic anemometers for wind speed measurement, and piezoresistive sensors for air pressure measurement. These sensor nodes are equipped with wireless communication modules (such as LoRa, ZigBee, or Wi-Fi) to form an adaptive network with other nodes, enabling dynamic data transmission. Controlled by a microprocessor (MCU), the sensor nodes collect environmental parameters and then use an information fusion algorithm to weightedly integrate various sensor data to generate a comprehensive environmental risk index. The network's self-organizing characteristics are based on a bio-inspired algorithm, dynamically adjusting the connections between nodes according to environmental changes and node health status to ensure network stability and efficient data transmission. Environmental parameter acquisition utilizes multimodal sensors, which weightedly integrate various parameters through an information fusion algorithm.

[0132] Weighted integration formula:

[0133]

[0134] E: Integrated environmental parameter index after fusion, w i The weight of the i-th parameter reflects its importance in fire risk assessment and is usually determined through expert evaluation or historical data analysis; p i The i-th environmental parameter value collected; this formula uses weight normalization to ensure that the comprehensive index is between 0 and 1, which facilitates subsequent decision-making;

[0135] The sensor nodes employ a self-organizing network protocol based on bio-inspired algorithms to achieve dynamic topology reconstruction. The improved formula for connection relationship transition probability is automatically adjusted based on environmental changes and node health status.

[0136]

[0137] P ij : The transition probability from node i to node j; τ ij Path strength represents the amount of information transmitted in the past; η ij Heuristic factors reflect the ease of connection. The health status of node j assesses the node's battery level and operational status; a higher value indicates a healthier node. The health status of node j assesses the node's power and operating status. A higher value indicates a healthier node. α, β, and γ are adjustment parameters with values ​​between -1 and 1, reflecting the degree of influence of each factor on the transition probability. They need to be optimized through experiments.

[0138] The adaptive sensor network unit uses ensemble learning to achieve self-repair when a node fails. The node reconfiguration formula is as follows:

[0139]

[0140] R j : The target node after reconfiguration, used to replace the failed node. K: The set of candidate nodes, all available nodes. Confidence k The reliability score of node k is output by a machine learning model, based on historical data and real-time performance evaluation.

[0141] Sp2: The sensor node collects environmental parameter data in real time and transmits the data to the built-in data processing unit for local processing.

[0142] Sensor nodes acquire environmental parameters in real time through periodic or event-driven acquisition modes. When rapid changes are detected (such as a sudden increase in wind speed or a sharp drop in humidity), the system can increase the sampling frequency to ensure that environmental changes are captured. The acquired data is first pre-processed within the node to reduce noise and errors. The processed data is then transmitted via a wireless communication module to the node's built-in data processing unit, which uses a microprocessor to quickly process large amounts of data and perform preliminary data analysis. The data processing unit can automatically select the optimal data transmission path according to a protocol, ensuring data flow can continue even under heavy network load.

[0143] SP3: The data processing unit uses blockchain technology to verify the collected data, ensuring its authenticity and integrity.

[0144] To ensure the authenticity and immutability of sensor data, the data processing unit incorporates blockchain technology. Each piece of collected data is generated with a unique hash value using a hash function and stored on the blockchain along with a timestamp and node ID, forming a chain of data records. The distributed ledger technology of blockchain allows different nodes to verify the data, ensuring that data from all sensor nodes is traceable and cannot be forged. This data verification mechanism enhances the security and reliability of the entire monitoring system, preventing data anomalies caused by external data tampering or sensor node malfunctions.

[0145] The data processing unit processes and verifies the collected data locally. The data hash formula is as follows:

[0146] Hash d =H(d)=H(T||D||ID)

[0147] Hash dH: Hash value of the data record; H: Hash function, ensuring data immutability; T: Timestamp, recording the data generation time; D: Data collected by the sensor; ID: Unique identifier of the sensor node;

[0148] Combining random forest, support vector machine, and deep learning models for feature extraction and classification; random forest feature selection formula.

[0149]

[0150] F: Overall feature importance score; M: Number of trees in the forest; N m : Number of samples in the m-th tree; G: Loss function (Gini index) for each tree, used to evaluate classification performance; f m (x i The m-th tree is paired with the input feature x. i The prediction, y i : The output value of the loss function.

[0151] SP4: The fire early warning unit uses generative adversarial networks to analyze historical and real-time data, predict potential fire scenarios, and generate early warning signals.

[0152] The fire early warning unit incorporates a Generative Adversarial Network (GAN) for in-depth analysis of historical and real-time data collected by sensor nodes. The generator in the GAN model generates data samples of potential fire scenarios, while the discriminator evaluates the reliability of the generated data. Through repeated training on generation and discrimination, the model learns about changes in environmental characteristics before a fire occurs and predicts potential fire scenarios. This model can combine real-time and meteorological data to generate early warning signals for future fire risks and send them in real-time to the monitoring center and collaborative decision-making module, ensuring timely response.

[0153] Generative Adversarial Networks (GANs) for Fire Prediction Models:

[0154]

[0155] G(z): Generator, which generates potential fire scenario data based on random noise z; The generated prediction data; D(y) and D(y) are the discrimination results of generated data and real data, respectively, used to optimize the model.

[0156] SP5: The data processed by the data processing unit and the predicted data from the fire early warning unit are transmitted to the visualization display unit. The augmented reality module displays the monitoring data in real time and generates a fire risk assessment interface.

[0157] Data, verified via blockchain and processed locally, is transmitted to a visualization unit. This unit displays real-time monitoring data, including current environmental parameters and fire risk indices, through a high-resolution monitor or AR glasses. The augmented reality module overlays this data onto the real-world environment, allowing operators to directly see changes in environmental conditions on the monitoring interface. The system also generates a comprehensive fire risk assessment interface, including visual elements such as charts and heat maps, helping operators quickly understand the risks behind the data. Furthermore, this module supports interactive operations with users, allowing them to adjust monitoring parameters, view different historical data trends, and analyze environmental conditions from multiple perspectives.

[0158] Augmented Reality Output:

[0159] AR display =Overlay(E,R)

[0160] AR display Augmented reality display content; E: Environmental parameter fusion results; R: Real-time monitoring data;

[0161] Virtual Reality Simulation:

[0162] VR simulation =Simulate(R,H)

[0163] VR simulation Fire risk simulation in virtual reality; R: real-time data; H: historical fire data used for simulation scenarios.

[0164] SP6: The collaborative decision-making module analyzes the data from each sensor node through swarm intelligence algorithms, shares information, and dynamically adjusts monitoring strategies and early warning thresholds to improve response efficiency.

[0165] The collaborative decision-making module employs swarm intelligence technologies such as Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) to achieve information sharing and collaborative analysis among multiple nodes. Each node can adjust its monitoring strategy based on data transmitted from other nodes, thus forming a dynamic adaptive network. The system analyzes the health status of sensor nodes and real-time environmental data to adjust the communication frequency, sampling rate, and sensitivity of monitoring parameters between nodes, ensuring the timeliness and accuracy of fire risk monitoring. Furthermore, the collaborative decision-making module adjusts the system's warning thresholds based on real-time sensor data and historical data, ensuring that alarms are issued rapidly when environmental changes exceed set thresholds.

[0166] Information sharing mechanism, information update formula:

[0167] I ij (t+1)=I ij (t)+β·(S j(t)-I ij (t))

[0168] I ij (t): The degree of trust that node i has in the information of node j at time t; S j (t): The actual state information of node j (such as environmental parameters); β: The learning rate, which controls the speed of information updates; This formula is used to dynamically adjust the trust level of information from other nodes based on the received information.

[0169] Dynamically adjust monitoring strategies and early warning thresholds; threshold adjustment formula:

[0170] T new =T old +α·(E current -E target )

[0171] T new New monitoring threshold; T old Old monitoring thresholds; E current Current environmental risk assessment value (provided by sensor nodes); E target : Target risk assessment value, usually set by the system; α: Adjustment factor, controlling the range of change in the threshold;

[0172] Distributed decision model decision output formula:

[0173]

[0174] D I : The final decision output of node i (such as issuing an early warning or adjusting the monitoring strategy); N i S is the set of neighboring nodes of node i. i : The self-state of node i (such as health status, data validity); f is a comprehensive function that makes decisions based on information from neighboring nodes and its own state.

[0175] SP7: Based on the real-time data and early warning information provided by the visualization unit, the monitoring center operator makes emergency response decisions and implements corresponding measures.

[0176] When the fire early warning unit or collaborative decision-making module detects a potential fire risk, the warning signal and related data are transmitted to the monitoring center. Operators can view real-time environmental parameters, fire risk assessment interfaces, and predicted fire scenario data through a visualization display unit. Based on this information, operators will make emergency decisions, such as whether to dispatch on-site inspection personnel or directly activate fire-fighting equipment for early fire suppression. Operators can also adjust the system's monitoring strategy through the interactive interface, increasing the monitoring frequency of high-risk areas or temporarily setting new warning thresholds to cope with sudden environmental changes. The system will collect operator instructions in real time and feed them back to each module, forming a closed-loop monitoring and early warning mechanism to ensure seamless collaboration between monitoring and emergency response.

[0177] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0178] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cross-platform remote power transmission channel wildfire monitoring and early warning system, characterized in that, include: An adaptive sensor network unit is used to collect environmental parameters and weight and integrate the collected environmental parameters to obtain integrated data. The adaptive sensor network unit includes sensors, each sensor constituting a sensor node. Each sensor node is used to collect environmental parameters, including wind speed, humidity, temperature, and air pressure. A data processing unit is used to verify, store, and classify the received integrated data; the data processing unit is built into each sensor node. The fire early warning unit is used to integrate and train stored historical fire data and received integrated data to form a training dataset and output the generated potential fire prediction data; the fire early warning unit has a built-in generative adversarial network; A visualization unit is used to display the integrated data and potential fire prediction data, and generate a fire risk assessment interface; the visualization unit includes an augmented reality module and a virtual reality module. The collaborative decision-making unit is used to collaboratively calculate the integrated data, generate dynamically adjusted monitoring strategies and early warning thresholds, and feed the calculation results back to the adaptive sensor network unit. The collaborative decision-making module incorporates a swarm intelligence algorithm.

2. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The sensor nodes employ a self-organizing network protocol based on a bio-inspired algorithm for dynamic topology reconfiguration. This protocol automatically adjusts the connection relationships between nodes based on environmental changes and node health status to optimize data transmission paths.

3. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 2, characterized in that: The sensor node also includes a multimodal sensor, which collects environmental parameters. The multimodal sensor includes sensors for temperature, humidity, wind speed, soil moisture, light intensity, and vegetation type. The multimodal sensor uses an information fusion algorithm to weight and integrate the various environmental parameters.

4. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The data processing unit includes a computing platform, a storage device, a network interface, and a security module. The computing platform is used for data processing and computation; the storage device is used to store sensor data, model parameters, and historical records; and the network interface is used for communication with the adaptive sensor network unit, the fire early warning unit, the visualization display unit, and the collaborative decision-making unit.

5. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The augmented reality module overlays real-time monitoring data onto the monitoring interface; the virtual reality module is used to provide a risk simulation environment.

6. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The generative adversarial network is trained using heterogeneous data, including historical fire records, real-time monitoring data, and meteorological data, to generate multi-dimensional potential fire predictions.

7. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The collaborative decision-making module includes distributed computing nodes, a communication module, and a storage module. The distributed computing nodes are used for collaborative computing; the communication module is used to share synchronization information with sensor nodes; and the storage module is used to store shared information, decision-making strategies, and historical data.

8. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The adaptive sensor network unit also includes a self-repair module, which incorporates a machine learning algorithm to automatically reconfigure the network topology when a sensor node fails.

9. The cross-platform remote power transmission channel wildfire monitoring and early warning system according to claim 1, characterized in that: The visualization unit also includes risk area distribution, historical fire trend analysis, and future early warning models.

10. A monitoring and early warning method for a cross-platform remote power transmission channel wildfire monitoring and early warning system, characterized in that, include: Collect environmental parameters, and then weight and integrate the collected environmental parameters to obtain integrated data; The integrated data is then verified, stored, and categorized. The stored historical fire data and the received integrated data are integrated and trained to form a training dataset, and the generated potential fire prediction data is output. Display the integrated data and potential fire prediction data to generate a fire risk assessment interface; The integrated data is collaboratively calculated to generate dynamically adjusted monitoring strategies and early warning thresholds, and the calculation results are fed back for information sharing.

11. The cross-platform remote power transmission channel wildfire monitoring and early warning method according to claim 10, characterized in that: The environmental parameters are acquired using a multimodal sensor. The multimodal sensor then uses an information fusion algorithm to weight and integrate the various parameters. The weighted integration formula is as follows: ； E represents the integrated environmental parameter index after fusion; For the first The weights of each parameter reflect their importance in fire risk assessment and are usually determined through expert evaluation or historical data analysis. For the first time collected The formula uses weight normalization to ensure that the comprehensive index is between 0 and 1, which facilitates subsequent decision-making. The sensor nodes employ a self-organizing network protocol based on bio-inspired algorithms to achieve dynamic topology reconstruction. The improved formula for connection relationship transition probability is automatically adjusted based on environmental changes and node health status. ； Let be the transition probability from node i to node j; Path strength represents the amount of information transmitted in the past; As a heuristic factor, it reflects the ease of connection; To assess the health status of node j, evaluate the node's battery level and operating status; a higher value indicates a healthier node. To assess the health status of node k, we evaluate the node's battery level and operational status; a higher value indicates a healthier node. To adjust the parameters, the values ​​are between -1 and 1, reflecting the degree of influence of each factor on the transition probability, and need to be optimized through experiments; The adaptive sensor network unit uses ensemble learning to achieve self-repair when a node fails. The node reconfiguration formula is as follows: ； K represents the target node after reconfiguration, used to replace the failed node; K is the set of candidate nodes, all available nodes. The reliability score for node k is output by a machine learning model, based on historical data and real-time performance evaluation.

12. The cross-platform remote power transmission channel wildfire monitoring and early warning method according to claim 10, characterized in that: The collected data is processed and verified locally, using the following data hash formula: ； H is the hash value of the data record; H is the hash function, ensuring that the data cannot be tampered with. T is the timestamp, recording the time the data was generated; D represents the data collected by the sensor; ID represents the unique identifier of the sensor node. Combining random forest, support vector machine, and deep learning models for feature extraction and classification; the feature selection formula for random forest is: ； F represents the overall feature importance score; M represents the number of trees in the forest; Let G be the number of samples in the m-th tree; G is the loss function (Gini index) for each tree, used to evaluate classification performance. For the m-th tree pair of input features The prediction; This is the output value of the loss function.

13. The cross-platform remote power transmission channel wildfire monitoring and early warning method according to claim 10, characterized in that: In the step of displaying the integrated data and potential fire prediction data Augmented Reality Output: ； To display augmented reality content; E represents the result of environmental parameter fusion; R represents real-time monitoring data; Virtual Reality Simulation: ； This is for fire risk simulation in virtual reality; R represents real-time data; H represents historical fire data used to simulate scenarios.

14. The cross-platform remote power transmission channel wildfire monitoring and early warning method according to claim 10, characterized in that: In the step of integrating and training the stored historical fire data with the received integrated data, an adversarial network and a fire prediction model are generated. ； The generator generates potential fire scenario data based on random noise z; For the generated prediction data; and The results of the generated data and the real data are used to optimize the model.

15. The cross-platform remote power transmission channel wildfire monitoring and early warning method according to claim 10, characterized in that: In the information sharing step described above Information sharing mechanism, information update formula: ； Let be the value of node i's trust in node j's information at time t; This refers to the actual state information of node j (such as environmental parameters). The learning rate controls the speed at which information is updated; this formula is used to dynamically adjust the level of trust in information from other nodes based on the received information. Dynamically adjust monitoring strategies and early warning thresholds; threshold adjustment formula: ； The new monitoring threshold; The old monitoring threshold; The current environmental risk assessment value is provided by the sensor node; The target risk assessment value is usually set by the system; To adjust the factor and control the magnitude of the threshold change; Distributed decision model decision output formula: ； This is the final decision output for node i; Let i be the set of neighboring nodes; Let i be the self-state of node i; f is a comprehensive function that makes decisions based on information from neighboring nodes and its own state.

Images