A forest fire prevention monitoring method and system based on intelligent sensing network

Abstract

The application provides a forest fire prevention monitoring method and system based on intelligent sensing network, which comprises the following steps: an electronic device analyzes multi-source data through an anomaly detection engine, accurately identifies fire spread prediction results, and obtains the fire spread prediction results by using a fire spread simulation model. Based on the fire spread prediction results, a multi-objective optimization algorithm is used to solve the best emergency treatment scheme, and multiple optimization schemes are provided for decision makers. The method has the advantages of wide monitoring coverage, no restriction by terrain, high monitoring accuracy, fast response speed and the like, and significantly improves the intelligence and decision efficiency of fire prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a forest fire prevention monitoring method and system based on intelligent sensing network. BACKGROUND

[0002] Forest fires are extremely destructive natural disasters that pose a serious threat to the ecological environment, social economy, and human safety. Forest fires not only burn large areas of vegetation, causing devastating damage to the forest ecosystem, but also can trigger a series of chain reactions. After the fire, the loss of ground vegetation exposes the soil to the outside, making it vulnerable to wind and water erosion, which in turn leads to soil erosion and soil degradation. This ecological damage not only affects the natural recovery of plants, but also threatens wildlife that depend on forests for survival, destroys biodiversity, and can even alter local climate conditions. More seriously, the land after the fire may become vulnerable due to the loss of vegetation cover, increasing the risk of secondary disasters such as landslides and mudslides, further exacerbating the deterioration of the ecosystem.

[0003] Traditional methods mainly rely on manual patrols and fixed observation stations such as lookout towers for monitoring, but the coverage is limited, restricted by terrain and topography, and there are monitoring blind spots. The observation effect is affected by weather, such as lightning weather that cannot be observed, resulting in low monitoring accuracy and slow response speed. SUMMARY

[0004] In order to overcome the above-mentioned defects of limited monitoring coverage, restricted by terrain and topography, existence of monitoring blind spots, and observation effect affected by weather, resulting in low monitoring accuracy and slow response speed, the present application provides a forest fire prevention monitoring method based on intelligent sensing network, comprising:

[0005] Obtaining multi-source data collected by the intelligent sensing network of the target forest area;

[0006] Based on the multi-source data, using an anomaly detection engine for anomaly detection to obtain the fire situation monitoring result of the target forest area;

[0007] Based on the multi-source data and the fire situation monitoring result, using a fire spread simulation model for fire spread simulation to obtain a fire spread prediction result;

[0008] Based on the fire spread prediction result, using a multi-objective optimization algorithm to solve the best emergency handling scheme for the target forest area;

[0009] The anomaly detection engine is constructed based on a Long Short-Term Memory (LSTM) network combined with an Attention model; and the fire spread simulation model integrates a Fire Area Simulator (FARSITE) model and a Spatial-Temporal Graph Convolutional Network (ST-GCN).

[0010] Optionally, the anomaly detection based on the multi-source data utilizes an anomaly detection engine to perform anomaly detection to obtain a fire situation monitoring result of the target forest area.

[0011] Based on the multi-source data and the spatial positions of the collection sensors collecting the multi-source data in the intelligent sensing network, an Attention model is utilized to perform attention calculation to obtain an attention coefficient of each piece of data in the multi-source data.

[0012] Based on the multi-source data and the attention coefficient of each piece of data, each piece of data is weighted to obtain corresponding weighted data.

[0013] Based on each piece of weighted data, an LSTM network is utilized to perform anomaly detection to obtain a fire situation monitoring result of the target forest area.

[0014] Optionally, the LSTM network includes a time-scale LSTM subnetwork and a period-scale LSTM subnetwork.

[0015] The anomaly detection based on each piece of weighted data utilizing the LSTM network to obtain a fire situation monitoring result of the target forest area includes:

[0016] Based on each piece of weighted data, the time-scale LSTM subnetwork is utilized to perform feature extraction to obtain a mutation feature.

[0017] Based on the mutation feature and each piece of weighted data, the period-scale LSTM subnetwork is utilized to identify an abnormal deviation from a baseline to obtain a fire situation monitoring result of the target forest area.

[0018] Optionally, the fire spread simulation based on the multi-source data and the fire situation monitoring result utilizes a fire spread simulation model to perform fire spread simulation to obtain a fire spread prediction result.

[0019] Based on terrain slope, combustible humidity, and real-time wind direction data of the target forest area in the multi-source data, a FARSITE model is utilized to perform fire behavior prediction to obtain a fire behavior.

[0020] Based on the fire dynamic behavior, the multi-source data and the fire situation monitoring result, fire spread simulation is performed by using an ST-GCN to obtain a fire spread prediction result.

[0021] Optionally, the multi-objective optimization algorithm is solved based on the fire spread prediction result to obtain the optimal emergency handling scheme of the target forest area.

[0022] Based on the fire spread prediction result, the multi-objective optimization algorithm is solved to maximize the fire-fighting efficiency, minimize the personnel risk and minimize the resource consumption, and an optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions and distances between rescue personnel and fire field boundaries is obtained.

[0023] Optionally, the multi-objective optimization algorithm is solved based on the fire spread prediction result to maximize the fire-fighting efficiency, minimize the personnel risk and minimize the resource consumption, and an optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions and distances between rescue personnel and fire field boundaries is obtained.

[0024] The distance between the rescue personnel and the fire field boundary is greater than a preset safety distance, and the sum of the expansion speed of the fire field area and the diffusion speed of the fire is not more than the threshold of the extinguishing equipment coverage efficiency as a constraint condition;

[0025] Based on the fire spread prediction result and the constraint condition, the multi-objective optimization algorithm is solved to maximize the fire-fighting efficiency, minimize the personnel risk and minimize the resource consumption, and an optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions and distances between rescue personnel and fire field boundaries is obtained.

[0026] Optionally, the multi-objective optimization algorithm is solved based on the fire spread prediction result to obtain the optimal emergency handling scheme of the target forest area.

[0027] Based on the fire spread prediction result, the multi-objective optimization algorithm is solved based on the fire spread prediction result to obtain the optimal emergency handling scheme of the target forest area.

[0028] The fitness values of the individuals in the initialization population are calculated according to the fitness function, and the individuals in the initialization population are optimized based on the elite retention strategy combined with the chaotic mapping technology to obtain a new population. The population individuals are iteratively updated until the maximum iteration number is reached, and the emergency handling scheme corresponding to the individuals in the new population at this time is taken as the optimal emergency handling scheme of the target forest area.

[0029] In another aspect, the present application also provides a forest fire prevention monitoring system based on an intelligent sensing network, comprising:

[0030] an acquisition module configured to acquire multi-source data collected by an intelligent sensing network of a target forest area;

[0031] a fire situation real-time monitoring module configured to perform anomaly detection based on the multi-source data by using an anomaly detection engine to obtain a fire situation monitoring result of the target forest area, wherein the anomaly detection engine is constructed based on an LSTM network combined with an Attention model;

[0032] a fire spread simulation module configured to perform fire spread simulation based on the multi-source data and the fire situation monitoring result by using a fire spread simulation model to obtain a fire spread prediction result, wherein the fire spread simulation model is integrated with a FARSITE model and an ST-GCN;

[0033] an emergency command and collaborative disposal module configured to solve an optimal emergency disposal scheme for the target forest area by using a multi-objective optimization algorithm based on the fire spread prediction result.

[0034] Optionally, the fire situation real-time monitoring module is specifically configured to perform attention degree calculation by using an Attention model based on the multi-source data and spatial positions of acquisition sensors for collecting the multi-source data in the intelligent sensing network to obtain an attention degree coefficient of each piece of data in the multi-source data;

[0035] each piece of data is weighted based on the multi-source data and the attention degree coefficient of each piece of data to obtain corresponding each piece of weighted data;

[0036] anomaly detection is performed by using an LSTM network based on each piece of weighted data to obtain the fire situation monitoring result of the target forest area.

[0037] Optionally, the fire situation real-time monitoring module is specifically configured to perform feature extraction by using the time-scale LSTM subnetwork based on each piece of weighted data to obtain a mutation feature;

[0038] the cycle-scale LSTM subnetwork is used to identify abnormal deviation from a baseline based on the mutation feature and each piece of weighted data to obtain the fire situation monitoring result of the target forest area.

[0039] Optionally, the fire spread simulation module is specifically configured to perform fire intensity prediction by using a FARSITE model based on terrain slope, combustible humidity and real-time wind direction data of the target forest area in the multi-source data to obtain fire dynamic behavior;

[0040] the ST-GCN is used to perform fire spread simulation based on the fire dynamic behavior, the multi-source data and the fire situation monitoring result to obtain the fire spread prediction result.

[0041] Optionally, the emergency command and cooperative disposal module is specifically configured to solve the multi-objective optimization algorithm based on the fire spread prediction result, so as to maximize the fire-fighting efficiency, minimize the personnel risk and minimize the resource consumption, and obtain an optimal emergency treatment scheme including multiple rescue path directions, equipment scheduling decisions and distances between rescue personnel and fire boundaries.

[0042] Optionally, the emergency command and cooperative disposal module is specifically configured to take, as a constraint condition, that the distance between the rescue personnel and the fire boundary is greater than a preset safety distance, and the sum of the expansion speed of the fire area and the spread speed of the fire does not exceed a threshold of the extinguishing equipment coverage efficiency.

[0043] Based on the fire spread prediction result and the constraint condition, the multi-objective optimization algorithm is solved to maximize the fire-fighting efficiency, minimize the personnel risk and minimize the resource consumption, and an optimal emergency treatment scheme including multiple rescue path directions, equipment scheduling decisions and distances between rescue personnel and fire boundaries is obtained.

[0044] Optionally, the emergency command and cooperative disposal module is specifically configured to initialize a population of the multi-objective optimization algorithm based on the fire spread prediction result and in combination with a probability density function of the fire prediction result, and obtain an initialized population, wherein each individual in the initialized population corresponds to an initial emergency treatment scheme.

[0045] The fitness values of the individuals in the initialized population are calculated according to a fitness function, the individuals in the initialized population are optimized based on an elite reservation strategy and in combination with a chaotic mapping technology to obtain a new population, and the population individuals are iteratively updated until a maximum iteration number is reached, and the emergency treatment scheme corresponding to each individual in the new population at this time is taken as the optimal emergency treatment scheme of the target forest area.

[0046] In another aspect, the present application also provides a computer device, characterized in that it comprises one or more processors.

[0047] The processor is configured to store one or more programs.

[0048] When the one or more programs are executed by the one or more processors, the above-mentioned forest fire monitoring method based on an intelligent sensing network is implemented.

[0049] In another aspect, the present application also provides a computer readable storage medium, characterized in that it has a computer program stored thereon, and the computer program is executed to implement the above-mentioned forest fire monitoring method based on an intelligent sensing network.

[0050] Compared with the prior art, the present application has the following beneficial effects:

[0051] The application provides a forest fire prevention monitoring method and system based on an intelligent sensing network, which comprises the following steps: an electronic device acquires multi-source data collected by the intelligent sensing network; based on the multi-source data, an anomaly detection engine is used for anomaly detection to obtain a fire spread prediction result corresponding to the multi-source data; based on the multi-source data, a fire spread simulation model is used for fire spread to obtain a fire spread prediction result; and based on the fire spread prediction result, a multi-objective optimization algorithm is used to balance the extinguishing efficiency, personnel safety and resource consumption to obtain an optimal emergency treatment scheme. The multi-source data is analyzed by the anomaly detection engine to accurately identify the fire spread prediction result, and the fire spread simulation model is used to predict the fire spread in a future preset period. Based on the fire spread prediction result, the multi-objective optimization algorithm is used to balance the extinguishing efficiency, personnel safety and resource consumption to generate the optimal emergency treatment scheme, thereby providing multiple optimization schemes for decision makers. The method has the advantages of wide monitoring coverage, high monitoring accuracy, fast response speed and the like, and significantly improves the intelligence and decision efficiency of fire prevention and control. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 FIG. 1 is a flowchart of the forest fire prevention monitoring method based on the intelligent sensing network of the application;

[0053] Figure 2 FIG. 2 is a structural diagram of the forest fire prevention monitoring system based on the intelligent sensing network of the application;

[0054] Figure 3 FIG. 3 is a structural diagram of the electronic device of the application. DETAILED DESCRIPTION

[0055] The specific embodiments of the application will be further described in detail below with reference to the accompanying drawings.

[0056] Embodiment 1

[0057] The forest fire prevention monitoring method based on the intelligent sensing network provided by the application has the flowchart as shown in FIG. 1, which comprises the following steps: Figure 1

[0058] Step 101: acquiring multi-source data collected by the intelligent sensing network in a target forest area;

[0059] Step 102: based on the multi-source data, using an anomaly detection engine for anomaly detection to obtain a fire situation monitoring result of the target forest area;

[0060] Step 103: based on the multi-source data and the fire situation monitoring result, using a fire spread simulation model for fire spread simulation to obtain a fire spread prediction result;

[0061] ​Step 104: based on the fire spread prediction result, the optimal emergency treatment scheme of the target forest area is solved by using a multi-objective optimization algorithm; the anomaly detection engine is constructed based on an LSTM network combined with an Attention model; the fire spread simulation model is integrated with the FARSITE model and the ST-GCN.

[0062] The forest fire prevention monitoring method based on the intelligent perception network provided by the embodiment of the application is applied to an electronic device, which can be a personal computer (PC), a server, or other intelligent devices.

[0063] In order to accurately and effectively perform forest fire prevention monitoring based on the intelligent perception network, the electronic device acquires multi-source data collected by the intelligent perception network of the target forest area.

[0064] Specifically, the intelligent perception network can be deployed by arranging multiple types of sensor nodes according to gradient density in the target forest area; a temperature and humidity composite sensor with a measurement accuracy of ±0.5℃, a laser scattering smoke detector with a sensitivity of 0.05% obs / m, and a three-dimensional ultrasonic anemometer with a range of 0-30 m / s are deployed in the core protection area; unmanned aerial vehicle (UAV) nodes are deployed in the peripheral area; a micro weather station and a solar power supply unit are carried on the UAV; satellite monitoring equipment accesses high-resolution five high-spectrum data and Fengyun four meteorological satellite data; and hour-level data updating is realized through space-ground collaborative calibration.

[0065] Since the multi-source data usually comes from different sensors or monitoring devices, it may have different timestamps, spatial resolutions, or data formats. In order to accurately and effectively perform forest fire prevention monitoring, the electronic device can first align the multi-source data.

[0066] The electronic device inputs the multi-source data into a spatio-temporal multi-task learning (ST-MTL) algorithm, which aligns the multi-source data. The ST-MTL algorithm considers the correlation in the time and space dimensions and maps the multi-source data into a unified spatio-temporal framework. Specifically, the ST-MTL algorithm captures common features and differential features between different data sources through the combination of shared representation learning and task-specific learning, thereby achieving spatio-temporal alignment of the data. The aligned data has consistent time steps and spatial resolution, providing a unified basis for subsequent analysis. The improved ST-MTL algorithm can align satellite images (with a time resolution of 30 minutes), UAV thermal infrared data (with a spatial resolution of 0.1 m), and ground sensor data.

[0067] The electronic device can calculate the temperature gradient change rate based on multi-source data using a sliding window technique combined with Z-score standardization. The sliding window technique extracts local data segments for analysis by moving a fixed-size window over the multi-source data. Z-score standardization is used to eliminate the influence of data dimension, allowing temperature data of different scales to be compared uniformly. By calculating the gradient change rate of temperature data within each window, local trends in temperature changes can be captured, providing important evidence for fire warning.

[0068] The electronic device can construct a continuous concentration field of smoke concentration using a cubic spline interpolation method. Cubic spline interpolation is a smoothing interpolation technique that can generate a continuous function curve between discrete monitoring points, thus more accurately describing the spatial distribution of smoke concentration. Based on the interpolated continuous concentration field, the second derivative of smoke concentration, i.e., the spatial diffusion acceleration, is calculated. This process can reveal the rate of change of smoke concentration in space, providing key parameters for dynamic simulation of fire spread.

[0069] The electronic device can convert three-dimensional wind speed data in multi-source data into dynamic propagation direction weights using quaternion transformation. Quaternion transformation is a mathematical tool that can effectively handle rotation and direction changes in three-dimensional space. By converting wind speed data into dynamic propagation direction weights, the combined effect of wind speed vectors can be more accurately described, providing support for predicting the direction of fire spread.

[0070] After obtaining the temperature gradient change rate, the second derivative of smoke concentration, and the combined value of wind speed vectors, the calculated temperature gradient change rate, the second derivative of smoke concentration, and the combined value of wind speed vectors can be replaced with the original multi-source data.

[0071] After obtaining the multi-source data, the electronic device performs in-depth analysis of the data through an anomaly detection engine based on the multi-source data to identify potential fire spread prediction results of fire. The anomaly detection engine can combine machine learning algorithms and rule libraries to monitor and pattern match key parameters such as temperature, humidity, smoke concentration, and wind speed, quickly locate abnormal areas, and generate fire spread prediction results. The anomaly detection engine is based on an LSTM network combined with an Attention model.

[0072] Based on the multi-source data and the fire spread prediction result, the electronic device utilizes a fire spread simulation model to accurately predict the dynamic evolution of the fire, and generates the fire spread prediction result. The fire spread simulation model can comprehensively consider key parameters such as terrain slope, combustible humidity, real-time wind direction and wind speed, and simulate the space-time distribution of fire spread. The physical layer of the model integrates a fire area simulator (FARSITE) model for accurately calculating the physical diffusion process of the fire; the data-driven layer learns the mapping relationship between the historical fire field spread pattern and the current environmental parameters by training a spatial-temporal graph convolutional network (ST-GCN). In addition, the model uses a dynamic correction mechanism to fuse satellite fire point monitoring data and simulated fire spread prediction results every preset time interval by using Kalman filtering to continuously correct prediction errors and improve prediction accuracy.

[0073] Based on the fire spread prediction result, a multi-objective optimization algorithm is used to obtain the best emergency treatment scheme. The multi-objective optimization algorithm optimizes the deployment and scheduling scheme of fire extinguishing resources by combining the spatial and temporal characteristics of fire spread through a dynamic reference point strategy and an elite preservation mechanism. In the optimization process, the multi-objective optimization algorithm fully considers the uncertainty of fire spread, the safety of rescue paths, and the economy of resource utilization, and provides multiple non-dominated solutions as decision options. These solution sets can achieve the best balance between different objectives, providing flexible and scientific fire fighting strategies for decision makers, and significantly improving the efficiency and effectiveness of emergency response.

[0074] When performing anomaly detection using the anomaly detection engine to obtain the fire spread prediction result corresponding to the multi-source data, the fire spread simulation model can make decisions based on multi-parameter coupling. In one example, when the following conditions are met, a fire warning is triggered, that is, it is determined that the fire spread prediction result is that there is a fire: the temperature gradient change rate exceeds the baseline value 3σ (σ is the standard deviation of the same period), the smoke concentration second-order derivative is continuously increasing for 3 consecutive sampling periods, and the angle between the wind speed vector synthesis direction and the main axis of the combustible distribution is less than 45°. It can be understood that based on the historical environmental parameter distribution (confidence interval 95%) and the real-time meteorological warning level (such as the drought index), the decision threshold can be automatically scaled.

[0075] To accurately obtain the fire monitoring result, based on the above embodiments, in the embodiments of the present application, based on the multi-source data, the anomaly detection engine is used for anomaly detection to obtain the fire monitoring result of the target forest area, including:

[0076] Based on the multi-source data and the spatial positions of the collection sensors collecting the multi-source data in the intelligent sensing network, attention degree calculation is performed by using an Attention model to obtain an attention degree coefficient of each piece of data in the multi-source data.

[0077] Based on the multi-source data and the attention degree coefficient of each piece of data, each piece of data is subjected to weighted processing to obtain corresponding each piece of weighted data.

[0078] Based on each piece of weighted data, an LSTM network is used for anomaly detection to obtain a fire condition monitoring result of the target forest area.

[0079] In the embodiment of the present application, based on the multi-source data and the spatial positions of the collection sensors collecting the multi-source data in the intelligent sensing network, attention degree calculation is performed by using an Attention model to obtain an attention degree coefficient of each piece of data in the multi-source data. The Attention model dynamically allocates weights to focus on the data collected by the sensors at key spatial positions, thereby improving the accuracy of anomaly detection.

[0080] Based on the multi-source data and the corresponding attention degree coefficient, each piece of data is subjected to weighted processing to obtain weighted data. The weighted data can highlight the importance of the data collected by the key sensors and at the same time suppress the interference of irrelevant or noise data.

[0081] Based on the weighted data, an LSTM network is used for anomaly detection. The LSTM network captures long-term dependencies in time series data, identifies abnormal patterns, and generates a fire condition monitoring result of the target forest area.

[0082] In an example, the anomaly detection engine can assign different attention degree coefficients to nodes at different spatial positions in the sensor networking, such as a 30% weight increase for nodes in the flammable vegetation area.

[0083] The present embodiment proposes a dynamic heterogeneous network optimization algorithm, which can still maintain a data integrity rate of more than 85% in the communication interruption scene; develops a hybrid fire prediction model, which can improve the prediction accuracy by 42% compared with a single physical model; designs a multi-agent collaborative decision-making framework, which can shorten the emergency response time to within 8 minutes.

[0084] In order to accurately obtain the fire condition monitoring result, on the basis of the above embodiments, in the embodiment of the present application, the LSTM network comprises a time scale LSTM subnetwork and a period scale LSTM subnetwork.

[0085] Based on each piece of weighted data, an LSTM network is used for anomaly detection to obtain a fire condition monitoring result of the target forest area, which comprises:

[0086] Based on each piece of weighted data, a time scale LSTM subnetwork is used for feature extraction to obtain a mutation feature.

[0087] Based on the mutation features and each weighted data, the periodic scale LSTM subnetwork is used to identify abnormal deviation from the baseline, and the fire monitoring result of the target forest area is obtained.

[0088] The LSTM network includes a time scale LSTM subnetwork and a periodic scale LSTM subnetwork. The time scale LSTM subnetwork is responsible for extracting mutation features from multi-source data. The time scale LSTM subnetwork captures short-term fluctuations and mutation points in the data by analyzing the time series data of key parameters such as temperature, humidity, smoke concentration, and wind speed. The time scale LSTM subnetwork uses sliding window technology and deep learning models such as Convolutional Neural Network (CNN) or Long Short-Term Memory Network (LSTM) to extract local features and time series patterns in the data. These mutation features can reflect abnormal signals in the early stages of fire occurrence, providing important basis for subsequent anomaly detection.

[0089] The periodic scale network identifies abnormal deviation from the baseline based on the mutation features extracted by the time scale network and multi-source data. The periodic scale network establishes a baseline model under normal conditions by analyzing the periodic changes of the data. The network uses methods such as Fourier transform or AutoRegressive Integrated Moving Average (ARIMA) to capture the periodicity of the data. Then, by comparing real-time data with the baseline model, it identifies abnormal points that deviate from the normal pattern, and obtains the fire spread prediction result.

[0090] In order to accurately and effectively obtain the fire spread prediction result, based on the above embodiments, in the embodiments of the present application, based on the multi-source data and the fire monitoring result, a fire spread simulation model is used to simulate the fire spread, and the fire spread prediction result is obtained, including:

[0091] Based on the terrain slope, combustible humidity, and real-time wind direction data of the target forest area in the multi-source data, the FARSITE model is used to predict the fire, and the fire dynamic behavior is obtained.

[0092] Based on the fire dynamic behavior, multi-source data, and fire monitoring result, the ST-GCN is used to simulate the fire spread, and the fire spread prediction result is obtained.

[0093] In the embodiments of the present application, the electronic device can use the FARSITE model to predict the fire based on the terrain slope, fuel moisture and real-time wind direction data of the target forest area in the multi-source data, and obtain the dynamic behavior of the fire. The dynamic behavior can include fire line spread speed, fire intensity, burning area, etc. The FARSITE model can accurately predict the spatio-temporal evolution law of the fire through physically driven fire spread simulation.

[0094] Based on the dynamic behavior of the fire, the multi-source data and the fire monitoring results, the ST-GCN is used for fire spread simulation. The ST-GCN captures the spatio-temporal features of fire spread by combining the Graph Convolutional Network (GCN) and the Temporal Convolutional Network (TCN), and generates the fire spread prediction results. The ST-GCN can compensate for the limitations of traditional physical models in complex environments, and further improve the accuracy and reliability of fire spread simulation.

[0095] In order to accurately obtain the best emergency treatment scheme, based on the above embodiments, in the embodiments of the present application, based on the fire spread prediction results, a multi-objective optimization algorithm is used to solve the best emergency treatment scheme for the target forest area, including:

[0096] Based on the fire spread prediction results, the multi-objective optimization algorithm is solved to maximize the extinguishing efficiency, minimize the personnel risk and minimize the resource consumption, and the best emergency treatment scheme including the direction of the multiple rescue paths, the device scheduling decision, and the distance between the rescue personnel and the fire boundary is obtained.

[0097] The fire spread prediction results in the embodiments of the present application provide the spatio-temporal distribution information of the fire spread, including the boundary of the fire scene, the intensity of the fire and the future spread trend. The electronic device can use the fire spread prediction results as the input of the multi-objective optimization algorithm to guide the allocation and scheduling of rescue resources.

[0098] The core objective of the multi-objective optimization algorithm is to optimize the three key indicators simultaneously: maximizing the extinguishing efficiency, minimizing the personnel risk and minimizing the resource consumption. In order to achieve this goal, the multi-objective optimization algorithm can first construct an optimization model containing multiple decision variables. These decision variables include the direction of the rescue path, the scheduling decision of the device and the distance between the rescue personnel and the fire boundary.

[0099] During the optimization process, a series of possible rescue strategies are generated through iterative search. Each strategy makes a trade-off between extinguishing efficiency, personnel safety and resource consumption. Through non-dominated sorting and crowding degree calculation, the algorithm filters out those strategies that achieve the best balance between multiple objectives to form the best emergency treatment scheme.

[0100] Specifically, the determination of the rescue path direction takes into account the speed and direction of fire spread to ensure that rescue personnel can quickly reach the fire scene and effectively control the fire. The equipment scheduling decision is based on the availability, location and performance of the equipment to optimize the deployment and use of the equipment to improve the efficiency of the fire fighting. The distance of the rescue personnel from the fire scene boundary is included in the optimization model to ensure the safety of the rescue personnel and avoid them entering dangerous areas.

[0101] The formula f1 for maximizing the efficiency of the fire fighting is:

[0102]

[0103] where t i is the arrival time of the i th rescue team, n is the number of rescue teams, d i is the distance of the rescue personnel of the i th rescue team from the fire scene boundary, w i is the weight coefficient of the i th rescue team, and a is the distance attenuation factor, which can be 0.15.

[0104] The formula f2 for minimizing the risk of personnel is:

[0105]

[0106] where R j is the risk coefficient positively correlated with the intensity of the fire, m is the number of directions of fire spread, is the direction vector of the j th rescue path, is the direction vector of the direction of fire spread.

[0107] The formula f3 for minimizing the consumption of resources is:

[0108]

[0109] where p is the number of equipment scheduling, x k is the equipment scheduling decision variable of the k th equipment, c k is the single-machine operating cost of the k th equipment, Q is the resource capacity threshold, and l is the penalty coefficient, which can be 10 4 .

[0110] In order to accurately obtain the optimal emergency treatment scheme, on the basis of the above embodiments, in the embodiments of the present application, based on the fire spread prediction result, the multi-objective optimization algorithm is solved to maximize the efficiency of the fire fighting, minimize the risk of personnel and minimize the consumption of resources, and the optimal emergency treatment scheme including the multiple rescue path directions, the equipment scheduling decision and the distance of the rescue personnel from the fire scene boundary is obtained, including:

[0111] The constraint condition is that the distance between the rescue personnel and the fire boundary is greater than a preset safety distance, and the sum of the expansion speed of the fire area and the diffusion speed of the fire does not exceed a threshold of the coverage efficiency of the fire extinguishing equipment.

[0112] Based on the fire diffusion prediction result and the constraint condition, a multi-objective optimization algorithm is solved to obtain an optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions, and distances between the rescue personnel and the fire boundary, with the objectives of maximizing the firefighting efficiency, minimizing the personnel risk, and minimizing the resource consumption.

[0113] In order to accurately obtain the optimal emergency handling scheme, two key constraint conditions are introduced in the process of determining the optimal emergency handling scheme.

[0114] The distance between the rescue personnel and the fire boundary is greater than a preset safety distance: by calculating the distance between the current position of the rescue personnel and the fire boundary, it is ensured that the rescue personnel act within a safe range, avoiding being trapped in danger due to sudden changes or too fast spread of the fire.

[0115] The formula for the distance between the rescue personnel and the fire boundary being greater than a preset safety distance is:

[0116]

[0117] wherein the above formula represents that the minimum distance between the rescue personnel and the fire boundary in the jth rescue path is greater than a preset safety distance, P human is the position of the rescue personnel in the jth rescue path, P fire (t) is the position of the fire boundary at time t, β is a safety factor, which can be 0.12, and S(t) is the fire area at time t.

[0118] The sum of the expansion speed of the fire area and the diffusion speed of the fire does not exceed a threshold of the coverage efficiency of the fire extinguishing equipment: the expansion speed and the diffusion speed of the fire are calculated by a fire diffusion model, and it is ensured that the coverage efficiency of the fire extinguishing equipment can effectively control the spread of the fire. This constraint condition dynamically adjusts the equipment scheduling and resource allocation to ensure that the fire is within a controllable range.

[0119]

[0120] wherein S is the fire area in the fire diffusion prediction result, is the fire spread speed quality in the fire diffusion prediction result, A equip is the coverage efficiency of the fire extinguishing equipment, and 𝜏 is a preset proportion coefficient, is the change rate of the fire area S with time t, i.e., the expansion speed of the fire, is a divergence operator, which represents divergence calculation on a vector field.

[0121] The electronic device can solve the multi-objective optimization algorithm based on the fire spread prediction result and the constraint condition described above, to maximize the fire-fighting efficiency, minimize the personnel risk, and minimize the resource consumption, to obtain an optimal emergency handling scheme including multiple rescue path directions, device scheduling decisions, and distances between rescue personnel and fire field boundaries. Wherein, optimizing the travel route of the rescue team can ensure quick arrival at the fire field and avoidance of high-risk areas; reasonably allocating fire-fighting equipment (such as fire trucks and unmanned aerial vehicles) can improve resource utilization; dynamically adjusting the safety distance between the rescue personnel and the fire field boundary can ensure personnel safety.

[0122] In order to accurately obtain the optimal emergency handling scheme, on the basis of the above embodiments, in the embodiment of the application, the optimal emergency handling scheme of the target forest area is solved based on the fire spread prediction result by using a multi-objective optimization algorithm, which includes:

[0123] Based on the fire spread prediction result, the multi-objective optimization algorithm is initialized by combining the probability density function of the fire prediction result to obtain an initialized population, and each individual in the initialized population corresponds to an initial emergency handling scheme;

[0124] The fitness value of each individual in the initialized population is calculated according to the fitness function, and each individual in the initialized population is optimized based on the elite reservation strategy combined with the chaotic mapping technology to obtain a new population; the population individuals are iteratively updated until the maximum iteration number is reached, and the emergency handling scheme corresponding to the individual in the new population at this time is taken as the optimal emergency handling scheme of the target forest area.

[0125] The electronic device can introduce a dynamic reference point strategy based on the fire spread prediction result, and initialize the multi-objective optimization algorithm by using the probability density function of the fire prediction result to generate an initialized population. Wherein, each individual in the initial population corresponds to an initial emergency handling scheme, including rescue path direction, device scheduling decision and safety distance between rescue personnel and fire field boundary.

[0126] The electronic device can determine the initial value in the initialized population by the following formula:

[0127]

[0128] Wherein, x init is a determined initial solution, argmax x represents that an x value is to be found, so that the expression behind it reaches the maximum value, is the conditional expectation value, and the expectation value of the objective function f1 under the conditional probability density function of the given fire spread model, is the probability density function, and the specific function of the objective function f1 has been described in the above embodiments, and will not be described here.

[0129] The electronic device can calculate the fitness values of individuals in the initial population according to a fitness function, which comprehensively considers the goals of rescue efficiency, personnel risk and resource consumption. Based on an elite reservation strategy and a chaotic mapping technology, the individuals in the initial population are optimized to generate a new population. The elite reservation strategy ensures the overall quality of the population by reserving individuals with higher fitness values. The chaotic mapping technology enhances the diversity of the population by introducing a chaotic sequence, thereby avoiding the algorithm from falling into a local optimum. The population individuals are updated through multiple iterations until the maximum number of iterations is reached. Finally, the emergency treatment scheme corresponding to the individuals in the new population is taken as the best emergency treatment scheme for the target forest area. This scheme can achieve multi-objective optimization of maximizing rescue efficiency, minimizing personnel risk and minimizing resource consumption under complex constraints, thereby providing a scientific basis for fire emergency decision-making.

[0130] In the embodiment of the present application, the electronic device can also construct a three-dimensional target space surface and extract a typical scheme using an epsilon-constraint method, wherein the rapid attack scheme is f1≥0.85f1 max , f2≤1.2f2 min , and the steady control scheme is This scheme calculates the entropy weight value of each objective function where H(f i )=-∑p k logp k , p k is the value of the objective function f i in the kth state. Wherein, represents the maximum value of the rescue efficiency among all emergency treatment schemes, represents the minimum value of the personnel risk among all emergency treatment schemes, represents the divergence calculation of the formula f3 for minimizing resource consumption, Δx represents the change rate of the device scheduling decision variable, w i represents the entropy weight value of each objective function, H(f i ) represents the entropy of the objective function f i , f i =f1, f2, f3, ∑H(f i ) represents the entropy of all objective functions.

[0131] Embodiment 2:

[0132] Based on the same inventive concept, the present application also provides a forest fire prevention monitoring system based on an intelligent perception network, a structural schematic diagram of which is shown in Figure 2 , comprising:

[0133] An acquisition module 201 is configured to acquire multi-source data collected by an intelligent perception network in a target forest area.

[0134] The fire situation real-time monitoring module 202 is configured to perform anomaly detection on the multi-source data by using an anomaly detection engine to obtain a fire situation monitoring result of the target forest area; and the anomaly detection engine is constructed based on an LSTM network combined with an Attention model.

[0135] The fire spread simulation module 203 is configured to perform fire spread simulation on the multi-source data and the fire situation monitoring result by using a fire spread simulation model to obtain a fire spread prediction result; and the fire spread simulation model is integrated with a FARSITE model and an ST-GCN.

[0136] The emergency command and collaborative disposal module 204 is configured to solve an optimal emergency processing scheme for the target forest area by using a multi-objective optimization algorithm based on the fire spread prediction result.

[0137] In a specific implementation, the fire situation real-time monitoring module 202 is specifically configured to perform attention degree calculation on the multi-source data and spatial positions of collection sensors collecting the multi-source data in the intelligent sensing network by using an Attention model to obtain an attention degree coefficient of each piece of data in the multi-source data.

[0138] Each piece of data is weighted based on the multi-source data and the attention degree coefficient of each piece of data to obtain each piece of weighted data.

[0139] The anomaly detection is performed on each piece of weighted data by using an LSTM network to obtain the fire situation monitoring result of the target forest area.

[0140] In a specific implementation, the fire situation real-time monitoring module 202 is specifically configured to perform feature extraction on each piece of weighted data by using the time-scale LSTM subnetwork to obtain a mutation feature.

[0141] The anomaly deviation from the baseline is identified based on the mutation feature and each piece of weighted data by using the cycle-scale LSTM subnetwork to obtain the fire situation monitoring result of the target forest area.

[0142] In a specific implementation, the fire spread simulation module 203 is specifically configured to perform fire intensity prediction on terrain slope, combustible humidity and real-time wind direction data of the target forest area in the multi-source data by using a FARSITE model to obtain a fire intensity dynamic behavior.

[0143] The fire spread simulation is performed based on the fire intensity dynamic behavior, the multi-source data and the fire situation monitoring result by using an ST-GCN to obtain a fire spread prediction result.

[0144] In a specific implementation, the emergency command and coordinated disposal module 204 is specifically used to solve the multi-objective optimization algorithm based on the fire spread prediction results, with the goals of maximizing firefighting efficiency, minimizing personnel risks, and minimizing resource consumption, to obtain the optimal emergency response plan that includes multiple rescue path directions, equipment scheduling decisions, and the distance between rescue personnel and the fire scene boundary.

[0145] In a specific implementation, the emergency command and coordinated handling module 204 is specifically configured to set as constraints that the distance between the rescue personnel and the fire boundary is greater than a preset safety distance, and that the sum of the fire area expansion speed and the fire spread speed does not exceed a threshold of the fire extinguishing equipment coverage efficiency;

[0146] Based on the fire spread prediction results and the constraints, the multi-objective optimization algorithm is solved with the goals of maximizing firefighting efficiency, minimizing personnel risks, and minimizing resource consumption to obtain the optimal emergency response plan that includes multiple rescue path directions, equipment scheduling decisions, and the distance between rescue personnel and the fire scene boundary.

[0147] In a specific implementation, the emergency command and coordinated handling module 204 is specifically configured to initialize the population of the multi-objective optimization algorithm based on the fire spread prediction result and the probability density function of the fire intensity prediction result to obtain an initialized population, wherein the individuals in the initialized population correspond to the initial emergency handling plan;

[0148] The fitness values ​​of the individuals in the initialized population are calculated according to the fitness function, and the individuals in the initialized population are optimized based on the elite retention strategy combined with the chaotic mapping technology to obtain a new population; the individuals in the population are updated iteratively until the maximum number of iterations is reached, and the emergency treatment plan corresponding to the individuals in the new population at this time is used as the optimal emergency treatment plan for the target forest area.

[0149] Example 3:

[0150] like Figure 3 As shown, the present invention also provides an electronic device, which may be a computer, a single-chip microcomputer, a smart mobile device, or the like. The electronic device in this embodiment may include a processor, a memory, a transceiver component, and the like. The memory, processor, and transceiver component are connected via a bus; the memory may be used to store an execution program, which may include instructions; and the processor may be used to execute the instructions stored in the memory. The memory may also be used to store data, which may be accessed and / or modified during the execution of the instructions.

[0151] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of the forest fire monitoring method based on intelligent sensing network in the above embodiment.

[0152] Embodiment 4:

[0153] Based on the same inventive concept, the application further provides a readable storage medium, specifically an electronic device readable storage medium (Memory). The electronic device readable storage medium is a memory device in the electronic device, and is used for storing programs and data. It can be understood that the storage medium herein can include a built-in storage medium in the electronic device, and of course can also include an expansion storage medium supported by the electronic device. The storage medium provides a storage space, and the storage space stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more execution programs (including program codes). It should be noted that the storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium, and the steps of the forest fire monitoring method based on intelligent sensing network in the above embodiment can be implemented.

[0154] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0155] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0156] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0157] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0158] Finally, it should be noted that the above-described embodiments are merely intended for describing the technical solutions of the present application, but not to limit the protective scope of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that: after reading the present application, those skilled in the art can make various changes, modifications or equivalent replacements to the specific embodiments described above, but these changes, modifications or equivalent replacements are all within the protective scope of the pending claims.

Claims

1. A forest fire prevention monitoring method based on an intelligent sensing network, characterized in that, The method comprises: acquiring multi-source data collected by an intelligent sensing network of a target forest area; based on the multi-source data, using an anomaly detection engine for anomaly detection to obtain a fire situation monitoring result of the target forest area; based on the multi-source data and the fire situation monitoring result, using a fire spread simulation model for fire spread simulation to obtain a fire spread prediction result; based on the fire spread prediction result, using a multi-objective optimization algorithm to solve an optimal emergency handling scheme for the target forest area; the anomaly detection engine is constructed based on a long short-term memory (LSTM) network combined with an attention (Attention) model; the fire spread simulation model integrates a fire behavior simulation system (FARSITE) model and a space-time graph convolution network (ST-GCN); the method of using a multi-objective optimization algorithm to solve an optimal emergency handling scheme for the target forest area based on the fire spread prediction result comprises: taking the condition that the distance between rescue personnel and the fire field boundary is greater than a preset safety distance, and the sum of the expansion speed of the fire field area and the diffusion speed of the fire is not more than the threshold of the coverage efficiency of the fire extinguishing equipment as a constraint condition; based on the fire spread prediction result and the constraint condition, taking maximizing the extinguishing efficiency, minimizing the personnel risk, and minimizing the resource consumption as the target, the multi-objective optimization algorithm is solved to obtain an optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions, and the distance between rescue personnel and the fire field boundary; the formula for the distance between rescue personnel and the fire field boundary being greater than a preset safety distance is: wherein the above formula represents that the minimum distance between the rescue personnel and the fire boundary in the jth rescue path is greater than the preset safety distance, P human P is the position of the rescue personnel in the jth rescue path, fire (t) is the position of the fire boundary at time t, β is a safety factor, and S(t) is the fire area at time t; the formula for the sum of the expansion speed of the fire field area and the diffusion speed of the fire being not more than the threshold of the coverage efficiency of the fire extinguishing equipment is: Wherein, S is the fire field area in the fire spread prediction result, is the fire spread speed quality in the fire spread prediction result, A equip is the fire extinguishing equipment coverage efficiency, and 𝜏 is a preset proportion coefficient, is the change rate of the fire field area S with time t, that is, the expansion speed of the fire, is a divergence operator, indicating divergence calculation on a vector field.

2. The method of claim 1, wherein, the method of using an anomaly detection engine to perform anomaly detection based on the multi-source data to obtain a fire situation monitoring result of the target forest area comprises: based on the multi-source data and the spatial position of the collection sensor collecting the multi-source data in the intelligent sensing network, using an Attention model to calculate the attention degree to obtain the attention degree coefficient of each data in the multi-source data; based on the multi-source data and the attention degree coefficient of each data, each data is weighted to obtain corresponding each weighted data; based on each weighted data, using an LSTM network for anomaly detection to obtain a fire situation monitoring result of the target forest area.

3. The method of claim 2, wherein, The LSTM network comprises a time scale LSTM subnetwork and a period scale LSTM subnetwork; the method of using an LSTM network to perform anomaly detection based on each weighted data to obtain a fire situation monitoring result of the target forest area comprises: based on each weighted data, using the time scale LSTM subnetwork to extract features to obtain mutation features; based on the mutation features and each weighted data, using the period scale LSTM subnetwork to identify abnormal deviation from the baseline to obtain a fire situation monitoring result of the target forest area.

4. The method of claim 1, wherein, the method of using a fire spread simulation model to perform fire spread simulation based on the multi-source data and the fire situation monitoring result to obtain a fire spread prediction result comprises: The FARSITE model is used for fire behavior prediction based on terrain slope, fuel moisture and real-time wind direction data of the target forest area in the multi-source data, to obtain fire dynamic behavior; The ST-GCN is used for fire spread simulation based on the fire dynamic behavior, the multi-source data and the fire situation monitoring result, to obtain a fire spread prediction result.

5. The method of claim 1, wherein, The multi-objective optimization algorithm is used for solving the optimal emergency handling scheme of the target forest area based on the fire spread prediction result, which includes: The multi-objective optimization algorithm is initialized based on the fire spread prediction result and the probability density function of the fire behavior prediction result, to obtain an initialized population, and an individual in the initialized population corresponds to an initial emergency handling scheme; The fitness value of the individual in the initialized population is calculated according to a fitness function, and the individual in the initialized population is optimized based on an elite reservation strategy and a chaotic mapping technology, to obtain a new population; the population individual is iteratively updated until a maximum iteration number is reached, and the emergency handling scheme corresponding to the individual in the new population at this time is taken as the optimal emergency handling scheme of the target forest area.

6. A forest fire prevention monitoring system based on intelligent sensing network, characterized in that, It includes: An acquisition module is configured to acquire multi-source data collected by an intelligent sensing network of a target forest area; A fire situation real-time monitoring module is configured to perform anomaly detection based on the multi-source data by using an anomaly detection engine, to obtain a fire situation monitoring result of the target forest area; The anomaly detection engine is constructed based on a long short-term memory (LSTM) network and an attention (Attention) model; A fire spread simulation module is configured to perform fire spread simulation based on the multi-source data and the fire situation monitoring result by using a fire spread simulation model, to obtain a fire spread prediction result; the fire spread simulation model integrates a fire behavior simulation system (FARSITE) model and a spatio-temporal graph convolution network (ST-GCN); An emergency command and collaborative disposal module is configured to solve an optimal emergency handling scheme of the target forest area by using a multi-objective optimization algorithm based on the fire spread prediction result; The emergency command and collaborative disposal module is specifically configured to take, as a constraint condition, a case that a distance between a rescue personnel and a fire field boundary is greater than a preset safety distance, and a sum of an expansion speed of a fire field area and a diffusion speed of a fire behavior is not more than a threshold value of a fire extinguishing equipment coverage efficiency; The multi-objective optimization algorithm is solved based on the fire spread prediction result and the constraint condition, to maximize a rescue efficiency, minimize a personnel risk and minimize a resource consumption, to obtain the optimal emergency handling scheme including multiple rescue path directions, equipment scheduling decisions and distances between the rescue personnel and the fire field boundary. The formula that the distance between the rescue personnel and the fire field boundary is greater than the preset safety distance is: wherein the above formula represents that the minimum distance between the rescue personnel and the fire boundary in the jth rescue path is greater than the preset safety distance, P human P is the position of the rescue personnel in the jth rescue path, fire (t) is the position of the fire boundary at time t, β is a safety factor, and S(t) is the fire area at time t. The formula that the sum of the expansion speed of the fire field area and the diffusion speed of the fire behavior is not more than the threshold value of the fire extinguishing equipment coverage efficiency is: wherein S is a fire field area in a fire spread prediction result, is a fire spread speed quality in the fire spread prediction result, A equip is a fire extinguishing equipment coverage efficiency, τ is a preset proportional coefficient, is a change rate of the fire field area S with time t, that is, a fire spread speed, is a divergence operator, indicating a divergence calculation on a vector field.

7. An electronic device, comprising: It includes: At least one processor and a memory; The memory and the processor are connected through a bus; The memory is configured to store one or more programs; When the one or more programs are executed by the at least one processor, the forest fire monitoring method based on the intelligent sensing network is implemented as claimed in any one of claims 1-5.

8. A readable storage medium, characterized by, A computer program product, having a computer program stored thereon, which, when executed, implements the forest fire monitoring method based on intelligent sensing network according to any one of claims 1-5.

Images