Intelligent Trigger Control Method and System for Fire Extinguishing Stick Based on Internet of Things

Abstract

The present invention discloses an intelligent triggering control method and system for fire extinguishing patches based on the Internet of Things, relating to the technical field, and comprising the following steps: constructing a fire extinguishing patch matrix for the area to be measured, and obtaining initial fire signal data; performing DWT decomposition on the initial fire signal data based on the blind source separation method and the wavelet threshold denoising method to obtain fire signal features at different scales; optimizing the fire signal features based on the particle swarm algorithm to obtain a fire signal feature subset; constructing a fire warning model based on the fire signal feature subset by means of the random forest algorithm, and controlling the triggering of the fire extinguishing patches based on the prediction result of the fire warning model; obtaining the fire location according to the location of the fire extinguishing patches after triggering, judging the fire extinguishing situation and mobilizing the fire extinguishing patches; through optimizing the intelligent control of the fire extinguishing patches and the fineness of the triggering conditions, the present invention solves the problems that the fire extinguishing patches cannot be scheduled in real time and the triggering conditions are not intelligent or accurate enough.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of fire extinguishing patches, and more specifically, to an intelligent triggering control method and system for fire extinguishing patches based on the Internet of Things. Background Art

[0002] With the acceleration of the urbanization process, the risk of fire occurrence is increasing day by day. Especially in fire-prone places such as high-rise buildings, industrial parks, and storage facilities, traditional fire-fighting means are difficult to meet the increasingly complex fire prevention and control requirements. In recent years, the rapid development of the Internet of Things technology has provided a new solution for fire monitoring and prevention. Through the combination of sensors, networks, and cloud platforms, the Internet of Things technology enables real-time, automated, and remote control of fire monitoring and response.

[0003] A fire extinguishing patch is a fire-fighting device used to prevent and extinguish incipient fires. It is usually attached to the surface or inside of a device in the form of a patch and automatically releases a fire extinguishing agent through a heat induction trigger mechanism to achieve rapid fire extinguishing.

[0004] At present, although the fire extinguishing patch has the function of self-starting fire extinguishing, the lack of the application of the Internet of Things technology and intelligent control for the fire extinguishing patch leads to the inability to accurately locate the fire position and conduct real-time scheduling, or the fire extinguishing patch may be triggered under unnecessary circumstances, resulting in waste of resources or unnecessary chemical reactions. Secondly, if the triggering conditions are not intelligent or accurate enough, the fire extinguishing patch may not be activated at the first time of the fire occurrence, delaying the fire extinguishing opportunity and increasing the risk of fire spread. In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide an intelligent triggering control method and system for fire extinguishing patches based on the Internet of Things, and optimize the intelligent control of the fire extinguishing patch and the fineness of the triggering conditions to solve the problems that the fire extinguishing patch cannot be scheduled in real time and the triggering conditions are not intelligent or accurate enough.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent triggering control method for fire extinguishing patches based on the Internet of Things constructs a fire extinguishing patch matrix for the area to be measured, and obtains initial fire signal data. The initial fire signal data is decomposed by DWT based on the blind source separation method and the wavelet threshold denoising method to obtain fire signal features at different scales; the fire signal features are optimized based on the particle swarm algorithm to obtain a fire signal feature subset; a fire early warning model is constructed based on the fire signal feature subset by the random forest algorithm, and the triggering of the fire extinguishing patch is controlled based on the prediction result of the fire early warning model; the fire position is obtained according to the position of the fire extinguishing patch after triggering, the fire extinguishing situation is judged, and the fire extinguishing patch is mobilized.

[0008] In a preferred embodiment, a fire extinguishing patch matrix of the area to be measured is constructed, specifically: obtaining the position data and status data of the fire extinguishing patches in the area to be measured; constructing the fire extinguishing patch matrix based on spatial analysis according to the position data and status data of the fire extinguishing patches; and obtaining the usage conditions of the fire extinguishing patches in the fire extinguishing patch matrix through the status data of the fire extinguishing patches.

[0009] In a preferred embodiment, initial fire signal data is obtained, and the initial fire signal data is subjected to DWT decomposition based on the blind source separation method and the wavelet threshold denoising method to obtain fire signal features of different scales, specifically: obtaining the initial fire signal data, where the initial fire signal data includes air pressure change, current fluctuation, temperature data, smoke concentration, and gas concentration; based on the blind source separation method, separating the initial fire signal data to obtain each independent signal data in the initial fire signal data; performing multi-layer decomposition on each independent signal data through DWT to obtain the low-frequency approximation coefficient and high-frequency detail coefficient of each signal data; evaluating to obtain a noise threshold based on the noise standard deviation, and selecting a hard threshold formula based on the wavelet threshold denoising algorithm to perform denoising processing on the high-frequency detail coefficient; recombining the denoised high-frequency detail coefficient and the low-frequency approximation coefficient based on the inverse wavelet transform to obtain fire signal features of different scales.

[0010] In a preferred embodiment, the fire signal features are optimized based on the particle swarm algorithm to obtain a fire signal feature subset, specifically: according to the fire signal features of different scales, performing selection, crossover, and mutation operations based on the genetic algorithm, and using each obtained optimal solution as the initial position of each particle in the particle swarm algorithm, where each particle represents a set of fire signal features; generating the initial velocity of the particles through uniform distribution based on a preset initial velocity range; for each particle, based on the classification model, inputting the fire signal features represented by it into the classification model, and evaluating the fitness by calculating the classification accuracy of the model, that is, the fitness function is the calculation formula of the classification accuracy of the model; each particle is updated based on the current fitness, the historical best fitness, and the global best position according to a preset position and velocity update formula; when the algorithm runs to the set maximum number of iterations, the algorithm is stopped, and at this time, the feature subset corresponding to the global best position of the particles is the optimal fire signal feature subset optimized by the particle swarm algorithm.

[0011] In a preferred embodiment, each particle is updated according to the current fitness, historical best fitness, and global best position based on a preset position and velocity update formula. Specifically: Obtain the historical best fitness of each particle. If the fitness of the current particle is better than its historical best fitness, update the personal best position of the particle based on the preset position and velocity update formula. By comparing the personal best positions of all particles, find the particle with the optimal fitness, and use the solution of this particle as the global optimal solution to update the global best position.

[0012] In a preferred embodiment, a fire warning model is constructed based on the fire signal feature subset using the random forest algorithm, and the triggering of the fire extinguishing patch is controlled based on the prediction result of the fire warning model. Specifically: Obtain historical fire data, and obtain the fire thresholds of different fire signal features through regression analysis of the historical fire data, and label the corresponding fire thresholds in the fire signal feature subset. The fire signal features include air pressure features, current fluctuation features, temperature features, smoke concentration features, and gas concentration features. Divide the labeled fire signal feature subset into a training set and a test set, and construct the corresponding number of decision trees according to the fire signal features in the training set. Train each decision tree to form an independent decision tree, and set the best splitting point based on the Gini index through impurity. Based on the best splitting point, perform root node splitting until the maximum depth of the tree is reached, stop splitting, and generate leaf nodes. If the proportion of fire samples in the leaf node is greater than the proportion of non-fire samples, the label of this leaf node is fire, otherwise it is non-fire. Integrate the generation results of all decision trees and obtain the final prediction result based on the average mechanism. When the prediction result is fire, control the triggering of the fire extinguishing patch to perform fire extinguishing treatment.

[0013] In a preferred embodiment, the fire location is obtained according to the position of the fire extinguishing patch after triggering, the fire extinguishing situation is judged, and the fire extinguishing patch is mobilized. Specifically: After the fire extinguishing patch is triggered, predict again according to the fire warning model to judge whether the fire is extinguished. If not, obtain the spread situation of the fire source and formulate mobilization rules according to the distribution of the fire extinguishing patches in the fire extinguishing patch matrix. Update the fire extinguishing patch matrix in real time according to the mobilization result.

[0014] In a preferred embodiment, the spread situation of the fire source is obtained, and mobilization rules are formulated according to the distribution of the fire extinguishing patches in the fire extinguishing patch matrix. Specifically: Based on graph theory and graph search algorithms, the fire extinguishing area and the fire extinguishing patch matrix are regarded as graphs, where each position is set as a node, and the neighborhood within a preset range is set as an edge; the neighborhood is traversed by the breadth-first search method, and according to the spread situation of the fire source, the nodes closest to the fire source and the fire extinguishing patches in the spreading direction of the fire source are preferentially traversed; through layer-by-layer traversal, it is judged whether the fire extinguishing patch can be triggered within a preset time according to the status data of the fire extinguishing patch; if it cannot be triggered, continue to traverse until a fire extinguishing patch that can be triggered and has the shortest path to the fire source is found, and the fire extinguishing patch is controlled to perform fire extinguishing treatment.

[0015] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0016] 1. By constructing a fire extinguishing patch matrix, resource allocation can be optimized, the response speed can be improved, and high-risk areas can be accurately identified, thereby enhancing the fire extinguishing efficiency and safety. Secondly, the blind source separation and wavelet threshold denoising methods are used to process the fire signal, enabling the signal to be accurately analyzed in multiple scales and frequency bands, extracting clear fire characteristics, and effectively improving the accuracy and sensitivity of the fire warning system. DWT decomposition can capture the subtle changes in the fire signal, and combined with the noise removal technology, it greatly enhances the robustness of signal processing, ensuring the real-time identification of potential fire risks in complex environments.

[0017] 2. By combining the particle swarm optimization algorithm and the genetic algorithm, the feature selection process of the fire signal is optimized. The particle swarm optimization algorithm can automatically select the most relevant fire signal features, thereby removing redundant and irrelevant features, and enhancing the expression efficiency and recognition accuracy of the signal. Through the initialization process of the genetic algorithm, its global search ability is fully utilized to provide a high-quality initial solution for the particle swarm optimization algorithm, ensuring the convergence of the algorithm and avoiding the problem of local optimal solutions. Finally, the optimized feature subset can retain the discriminative information to the greatest extent, thereby improving the accuracy and reliability of the classification and recognition models. The overall solution effectively improves the performance of the fire signal processing system, ensuring efficient and accurate fire detection and warning.

[0018] 3. Through the random forest algorithm, the probability and risk of fire occurrence can be accurately predicted, ensuring that the fire extinguishing patch is triggered at the initial stage of the fire, thereby effectively shortening the reaction time and controlling the spread of the fire. Secondly, by analyzing the relationships between multiple features, the possibility of false triggering is reduced, avoiding unnecessary resource waste and equipment loss.

[0019] 4. Dynamically adjust the positions of fire extinguishing patches through the breadth - first search algorithm, and preferentially mobilize the fire extinguishing patches near the fire source and in the spreading direction, so as to achieve accurate and efficient fire - fighting response. This method ensures the rapid mobilization and reasonable allocation of fire extinguishing patches, improves the fire - fighting efficiency, and can timely judge whether the fire extinguishing patches can be triggered, avoiding resource waste and misoperation. Description of the Drawings

[0020] Figure 1 It is a schematic flowchart of the intelligent trigger control method for fire extinguishing patches based on the Internet of Things provided by the embodiment of the present application.

[0021] Figure 2 It is a schematic structural diagram of the intelligent trigger control system for fire extinguishing patches based on the Internet of Things provided by the embodiment of the present application. Detailed Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment 1 Figure 1 It is a schematic flowchart of the intelligent trigger control method for fire extinguishing patches based on the Internet of Things provided by the embodiment of the present application, including the following steps:

[0024] S1, construct a fire - extinguishing patch matrix for the area to be measured, and obtain the initial fire signal data. Decompose the initial fire signal data by DWT based on the blind source separation method and the wavelet threshold denoising method to obtain fire signal features at different scales.

[0025] The purpose of the fire - extinguishing patch matrix is to optimize and improve the efficiency of fire - fighting operations, ensure that in the event of a fire, it can respond quickly, accurately, and effectively, and minimize losses to the greatest extent. The fire - extinguishing patch matrix provides scientific and systematic decision - making support for fire - fighting work by comprehensively considering factors such as fire risks in different regions, resource distribution, and fire - fighting capabilities. Specifically, there are several key reasons for constructing the fire - extinguishing patch matrix:

[0026] Optimize resource allocation: Fire - fighting resources are often limited, so it is necessary to reasonably allocate according to the fire risks and resource requirements in different regions. This helps to ensure that resources can be put to the most needed places at critical moments, thus improving the fire - fighting efficiency.

[0027] Improve response speed: After a fire breaks out, it is crucial to carry out fire extinguishing operations promptly. The fire extinguishing sticker matrix can identify the fire extinguishing priorities and resource requirements of each area, enabling firefighters to respond quickly based on the information displayed in the matrix, shortening the reaction time, and increasing the fire extinguishing speed.

[0028] Identify high-risk areas: Due to factors such as the environment, buildings, and population density, certain areas may face a higher fire risk. By constructing a fire extinguishing sticker matrix, high-risk areas can be clearly identified, providing data support for formulating fire prevention strategies, resource allocation, and training plans, and making preparations in advance.

[0029] Install multiple micro sensors on the fire extinguishing device. The micro sensors include a temperature sensor arranged on the surface of the fire extinguishing device, a smoke sensor arranged near the ventilation hole of the fire extinguishing device, a gas sensor arranged at the edge of the fire extinguishing device, a pressure sensor arranged at the center of the fire extinguishing device, and a current sensor arranged in the area connected to the circuit. The micro sensors obtain initial fire signal data in real time and transmit it to the online control system through the communication module for data processing.

[0030] By performing DWT decomposition on the initial fire signal data, the early warning of fires has the following advantages:

[0031] Multi-scale analysis: DWT can analyze signals at different time scales, which is very helpful for capturing subtle changes in fire signals. For example, the signal may change slowly in the initial stage of a fire, while higher-frequency fluctuations may occur during the fire spread stage. DWT can effectively extract the features at these different scales.

[0032] Time-frequency localization characteristics: DWT has good time-frequency localization characteristics, that is, it can accurately analyze the local changes of signals in time and frequency. This makes it particularly effective in dealing with non-stationary signals (such as sudden fire signals), and can accurately locate the occurrence time of the fire and its intensity changes.

[0033] Efficient feature extraction: Through DWT decomposition, multi-level and multi-band features can be extracted from fire signals. These features can be used as inputs for the fire early warning system to help the system better distinguish fires from other interference signals, and improve the accuracy and sensitivity of early warning.

[0034] Strong real-time performance: The calculation efficiency of DWT is relatively high, which can process and analyze fire signals in real time, and timely detect potential fire risks. Especially in the early stage of a fire, timely signal recognition and processing are very crucial.

[0035] Strong adaptability: DWT has strong adaptability and can be combined with other signal processing technologies (such as neural networks, support vector machines, etc.) to further improve the performance and robustness of the fire warning system.

[0036] The construction of the fire extinguishing patch matrix for the area to be measured is specifically as follows:

[0037] Obtain the position data and status data of the fire extinguishing patches in the area to be measured;

[0038] Based on the position data and status data of the fire extinguishing patches, construct a fire extinguishing patch matrix through spatial analysis;

[0039] And obtain the usage situation of the fire extinguishing patches in the fire extinguishing patch matrix through the status data of the fire extinguishing patches.

[0040] It should be noted that the usage situation of the fire extinguishing patches can be understood as whether the fire extinguishing patches are used. If they are used, 0 is displayed in the fire extinguishing patch matrix; if not, 1 is displayed in the fire extinguishing patch matrix. The status data of the fire extinguishing patches includes the usage status of the fire extinguishing patches, that is, whether they are being used, whether they have been activated, the validity period, the number of trigger times, and the usage period.

[0041] The acquisition of the initial fire signal data, and the DWT decomposition of the initial fire signal data based on the blind source separation method and the wavelet threshold denoising method to obtain the fire signal features at different scales is specifically as follows:

[0042] Obtain the initial fire signal data, and the initial fire signal data includes air pressure change, current fluctuation, temperature data, smoke concentration, and gas concentration;

[0043] Based on the blind source separation method, separate the initial fire signal data to obtain each independent signal data in the initial fire signal data;

[0044] Perform multi-layer decomposition on each independent signal data through DWT to obtain the low-frequency approximation coefficients and high-frequency detail coefficients of each signal data;

[0045] Evaluate based on the noise standard deviation to obtain the noise threshold, and select the hard threshold formula based on the wavelet threshold denoising algorithm to perform denoising processing on the high-frequency detail coefficients;

[0046] Recombine the denoised high-frequency detail coefficients and low-frequency approximation coefficients based on the inverse wavelet transform to obtain the fire signal features at different scales.

[0047] The noise threshold, the specific calculation formula is as follows:

[0048]

[0049] The hard threshold formula, the specific calculation formula is as follows:

[0050]

[0051] In the formula, is the noise threshold, is the standard deviation of the noise, is the length of the signal, is the th high-frequency detail coefficient.

[0052] It should be noted that for high-frequency detail coefficients, when their absolute value is less than the noise threshold, they are set to zero; when it is greater than the noise threshold, the coefficient remains unchanged. After signal decomposition, the high-frequency part is more vulnerable to noise interference. Especially in the detail coefficients, noise usually appears in these high-frequency components. Screening the high-frequency part during denoising can effectively reduce the influence of noise. Denoising can preserve the main structure of the signal: Wavelet transform decomposes the signal, enabling the low-frequency part of the signal to retain most of the information, while the high-frequency part is often noise. By removing the noise part, the main structure of the signal is better preserved. Threshold denoising can enhance subsequent feature extraction: When performing subsequent feature extraction (such as using DWT coefficients for machine learning training), the denoised signal usually provides clearer and more reliable features, enhancing the performance of the model. Since unimportant detail information is removed, the reconstructed signal should be smoother than the original signal while retaining most of the useful features. Multi-sensor data may contain mixed signals. For example, signals from different sources (temperature, smoke, gas concentration, etc.) during a fire are mixed together and are difficult to separate and process independently. Through blind source separation technology, different components of the signal are separated. Before DWT processing, blind source separation is performed first to ensure effective separation of each sensor data source, thereby improving the decomposition accuracy.

[0053] S2, optimizing the fire signal features based on the particle swarm algorithm to obtain a fire signal feature subset.

[0054] By optimizing the features after DWT using the particle swarm algorithm, the frequency bands and feature subsets most relevant to fire signal recognition can be automatically selected, thereby retaining meaningful features, removing redundant or irrelevant parts, and improving the efficiency of signal representation. The optimized feature subset can retain the most discriminative information and enhance the classification or recognition accuracy.

[0055] Optimizing the fire signal features based on the particle swarm algorithm to obtain a fire signal feature subset specifically means:

[0056] According to the fire signal features at different scales, perform selection, crossover, and mutation operations based on the genetic algorithm, and use each obtained optimal solution as the initial position of each particle in the particle swarm algorithm, where each particle represents a set of fire signal features;

[0057] Generate the initial velocity of the particles by uniform distribution based on a preset initial velocity range;

[0058] For each particle, based on the classification model, input the fire signal features it represents into the classification model, and evaluate the fitness by calculating the classification accuracy of the model. That is, the fitness function is the calculation formula of the classification accuracy of the model;

[0059] Each particle is updated according to the current fitness, historical best fitness, and global best position based on a preset position and velocity update formula;

[0060] When the algorithm runs to the set maximum number of iterations, stop the algorithm. At this time, the feature subset corresponding to the global best position of the particles is the optimal fire signal feature subset optimized by the particle swarm algorithm.

[0061] The specific calculation formula for the initial velocity of the particles is as follows:

[0062] Where ,

[0063] The specific calculation formula for the classification accuracy is as follows:

[0064]

[0065] In the formula, is the initial velocity of particle at time , is the preset maximum velocity, is the preset minimum velocity, is a random number, is the classification accuracy, is the number of fire samples that are correctly classified as fires, is the number of non-fire samples that are correctly classified as non-fires, is the number of non-fire samples that are misclassified as fires, is the number of fire samples that are misclassified as non-fires.

[0066] The update of each particle according to the current fitness, historical best fitness, and global best position based on a preset position and velocity update formula is specifically as follows:

[0067] Obtain the historical best fitness of each particle. If the fitness of the current particle is better than its historical best fitness, update the personal best position of the particle based on a preset position and velocity update formula;

[0068] By comparing the personal best positions of all particles, the particle with the optimal fitness is found, and the solution of this particle is taken as the global optimal solution to update the global best position.

[0069] The specific calculation formula of the velocity update formula is as follows:

[0070]

[0071] The specific calculation formula of the position update formula is as follows:

[0072]

[0073] In the formula, is the updated velocity, is the particle at time initial velocity, is the particle at time position, is the particle at time personal best position, is the global best position, and is random number between is the inertia weight, is the updated position, is the individual learning factor, is the swarm learning factor.

[0074] It should be noted that by combining the selection, crossover, and mutation operations of the genetic algorithm to generate the initial particle swarm of the particle swarm algorithm, the global search ability of the genetic algorithm can be fully utilized to provide a high-quality initial solution for the particle swarm algorithm, thereby accelerating the search process and avoiding being trapped in a local optimal solution in the early stage. The initial solution generated by the genetic algorithm not only has a high fitness but also can better guide the particle swarm algorithm to effectively explore in the solution space.

[0075] S3. Construct a fire warning model based on the fire signal feature subset using the random forest algorithm, and control the triggering of the fire extinguishing patch based on the prediction result of the fire warning model.

[0076] When the online control system determines that the fire signal is a fire after data processing, it controls the trigger circuit on the fire extinguishing device to start quickly, puncture the fire extinguishing microcapsules and mix with the adhesive to form a fire extinguishing patch for fire extinguishing.

[0077] Constructing a fire warning model using the random forest algorithm and controlling the triggering of the fire extinguishing patch based on the prediction result of the fire warning model has the following advantages:

[0078] Improve response efficiency: With the accurate prediction of the random forest algorithm, the probability and risk of fire occurrence can be more precisely evaluated. The fire extinguishing patch can be triggered at the initial stage of the fire, respond quickly, effectively reduce the time of fire spread, and control the fire.

[0079] Reduce false triggers: By analyzing the relationships between multiple features, the random forest can relatively accurately judge the probability of fire occurrence, thereby reducing false triggers caused by misjudging environmental changes. This can avoid resource waste (such as the consumption of fire extinguishing patches) and excessive human intervention.

[0080] Reduce costs: Through accurate prediction, effective measures can be taken before the fire fully occurs, thereby reducing the economic losses caused by the fire. This also avoids frequent triggering of fire extinguishing patches, thus saving the costs of equipment and fire extinguishing materials.

[0081] Improve safety: With the accurate prediction of the fire warning model, the safety of personnel can be effectively protected, especially in places with a large number of people or key facilities. The accuracy of the model prediction ensures that the fire extinguishing patch can be activated in a timely manner when a fire occurs, preventing the spread of the fire and ensuring life safety.

[0082] Intelligent management: The fire warning system constructed using the random forest algorithm can perform real-time data analysis and learning, continuously optimizing the warning results. This intelligent management can continuously improve the accuracy of the model according to different environments, weather conditions, and historical data, further enhancing the reliability of the fire extinguishing patch trigger.

[0083] Enhance the personalization of fire protection strategies: With the accumulation of data, the random forest model can customize personalized fire extinguishing strategies according to different regions and different types of fires. The triggered fire extinguishing patches can more specifically and efficiently respond to different types of fires, reducing the occurrence of misoperations.

[0084] Constructing a fire warning model based on the fire signal feature subset using the random forest algorithm and controlling the trigger of the fire extinguishing patch based on the prediction results of the fire warning model is specifically as follows:

[0085] Obtain historical fire data, obtain the fire thresholds of different fire signal features through the historical fire data based on the regression analysis method, and label the corresponding fire thresholds in the fire signal feature subset. The fire signal features include air pressure features, current fluctuation features, temperature features, smoke concentration features, and gas concentration features;

[0086] Divide the labeled fire signal feature subset into a training set and a test set, and construct the corresponding number of decision trees according to the fire signal features in the training set;

[0087] Train each decision tree to form independent decision trees, and set the best splitting point based on the Gini index through impurity;

[0088] Based on the best splitting point, perform root node splitting until the maximum depth of the tree is reached, stop splitting, and generate leaf nodes;

[0089] If the proportion of fire samples in the leaf node is greater than the proportion of non-fire samples, the label of this leaf node is fire, otherwise it is non-fire;

[0090] Integrate the generation results of all decision trees and obtain the final prediction result based on the averaging mechanism;

[0091] When the prediction result is fire, control the fire extinguishing patch to trigger and perform fire extinguishing;

[0092] The specific method for setting the best splitting point based on the Gini index through impurity is as follows:

[0093] The Gini index is an indicator to measure the purity of a data set. During the training of a decision tree, the algorithm will traverse all features and calculate the Gini index of each feature at the splitting point between fire and non-fire. If the splitting through a certain feature can divide the data set into two relatively pure subsets, then this feature and its splitting point are considered optimal.

[0094] S4. Obtain the fire location according to the position of the fire extinguishing patch after triggering, judge the fire extinguishing situation and mobilize the fire extinguishing patch, specifically:

[0095] After the fire extinguishing patch is triggered, make a prediction again according to the fire warning model to judge whether the fire is extinguished;

[0096] If it is not extinguished, obtain the spread situation of the fire source and formulate a mobilization rule according to the distribution of the fire extinguishing patches in the fire extinguishing patch matrix;

[0097] Update the fire extinguishing patch matrix in real time according to the mobilization result;

[0098] It should be noted that the real-time update of the fire extinguishing patch matrix according to the mobilization result can be understood as real-time updating the elements 0 and 1 in the fire extinguishing patch matrix representing the usage situation after the mobilization of the fire extinguishing patch.

[0099] The specific method for obtaining the spread situation of the fire source and formulating a mobilization rule according to the distribution of the fire extinguishing patches in the fire extinguishing patch matrix is as follows:

[0100] Based on graph theory and graph search algorithms, regard the fire extinguishing area and the fire extinguishing patch matrix as a graph, where each position is set as a node, and the neighborhood within a preset range is an edge;

[0101] Traverse the neighborhood through the breadth-first search method. According to the spread of the fire source, preferentially traverse the nodes closest to the fire source and the fire extinguishing patches in the direction of the fire source spread;

[0102] Through layer-by-layer traversal, judge whether the fire extinguishing patch can be triggered within the preset time according to the status data of the fire extinguishing patch;

[0103] If it cannot be triggered, continue to traverse until a fire extinguishing patch that can be triggered and has the shortest path to the fire source is found, and control the fire extinguishing patch to perform fire extinguishing treatment.

[0104] It should be noted that judging whether the fire extinguishing patch can be triggered within the preset time according to the status data of the fire extinguishing patch can be understood as judging according to the validity period, trigger times and service life of the fire extinguishing patch. If the trigger times do not exceed the preset safe trigger times and the fire extinguishing patch is within the safe and effective range, the fire extinguishing patch can be triggered.

[0105] Embodiment 2, Figure 2 It is a schematic structural diagram of an intelligent trigger control system for fire extinguishing patches based on the Internet of Things provided by an embodiment of the present application, including a matrix construction and data processing module, a feature optimization module, a fire warning module, and a fire extinguishing patch mobilization module. There are connections between the modules:

[0106] The matrix construction and data processing module is used to construct a fire extinguishing patch matrix for the area to be measured, obtain initial fire signal data, and perform DWT decomposition on the initial fire signal data based on the blind source separation method and the wavelet threshold denoising method to obtain fire signal features of different scales;

[0107] The feature optimization module is used to optimize the fire signal features based on the particle swarm algorithm to obtain a fire signal feature subset;

[0108] The fire warning module is used to construct a fire warning model based on the random forest algorithm for the fire signal feature subset, and control the trigger of the fire extinguishing patch based on the dynamic threshold adjustment method;

[0109] The fire extinguishing patch mobilization module is used to obtain the fire location according to the position of the fire extinguishing patch after triggering, judge the fire extinguishing situation, and mobilize the fire extinguishing patch.

[0110] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0111] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0112] Those of ordinary skill in the art will recognize that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0113] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0114] As described above, the above are only specific embodiments of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0115] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A fire extinguishing sticker intelligent trigger control method based on the Internet of Things, characterized in that: The steps include: Construct a fire extinguishing sticker matrix for the area to be tested, obtain the initial fire signal data, perform DWT decomposition on the initial fire signal data based on the blind source separation method and wavelet threshold denoising method, and obtain fire signal features of different scales; The fire signal features are optimized based on a particle swarm algorithm to obtain a fire signal feature subset; A fire warning model is constructed based on the random forest algorithm using a subset of fire signal features, and the triggering of the fire extinguishing sticker is controlled based on the prediction results of the fire warning model; The fire location is obtained according to the position of the fire extinguishing sticker after being triggered, the fire extinguishing situation is judged and the fire extinguishing sticker is mobilized. The mobilization is performed by treating the fire extinguishing area and the fire extinguishing sticker matrix as a graph, traversing layer by layer, and judging whether the fire extinguishing sticker can be triggered within the preset time according to the status data of the fire extinguishing sticker. If it cannot be triggered, continue to traverse until a triggerable fire extinguishing sticker with the shortest path to the fire source is found, and the fire extinguishing sticker is controlled to extinguish the fire.

2. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 1 is characterized in that: The fire extinguishing sticker matrix of the area to be tested is constructed as follows: Obtain the fire extinguisher location data and fire extinguisher status data of the area to be tested; Construct a fire extinguishing sticker matrix based on spatial analysis according to the fire extinguishing sticker location data and fire extinguishing sticker status data; And the usage of the fire extinguishing stickers in the fire extinguishing sticker matrix is ​​obtained through the fire extinguishing sticker status data.

3. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 1 is characterized in that: The initial fire signal data is obtained, and the initial fire signal data is subjected to DWT decomposition based on a blind source separation method and a wavelet threshold denoising method to obtain fire signal features of different scales, specifically: Acquiring initial fire signal data, the initial fire signal data including air pressure changes, current fluctuations, temperature data, smoke concentration, and gas concentration; Based on the blind source separation method, the initial fire signal data is separated to obtain each independent signal data in the initial fire signal data; Through DWT, each independent signal data is decomposed in multiple layers to obtain the low-frequency approximate coefficient and high-frequency detail coefficient of each signal data; The noise threshold is obtained by evaluating based on the noise standard deviation, and the hard threshold formula is selected based on the wavelet threshold denoising algorithm to denoise the high-frequency detail coefficients; The denoised high-frequency detail coefficients and low-frequency approximate coefficients are recombined based on inverse wavelet transform to obtain fire signal characteristics of different scales.

4. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 1 is characterized in that: The fire signal feature subset is obtained by optimizing the fire signal feature based on the particle swarm algorithm, specifically: According to the fire signal characteristics of different scales, selection, crossover and mutation operations are performed based on the genetic algorithm, and each optimal solution is used as the initial position of each particle in the particle swarm algorithm, where each particle represents a set of fire signal characteristics; Based on the preset initial velocity range, the initial velocity of particles is generated by uniform distribution; For each particle, based on the classification model, the fire signal characteristics it represents are input into the classification model, and the fitness is evaluated by calculating the classification accuracy of the model, that is, the fitness function is the classification accuracy calculation formula of the model; Each particle is updated based on the current fitness, the best historical fitness, and the global best position based on the preset position and speed update formula; When the algorithm reaches the set maximum number of iterations, the algorithm stops. At this time, the feature subset corresponding to the global optimal position of the particle is the optimal fire signal feature subset after optimization by the particle swarm algorithm.

5. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 4 is characterized in that: Each particle is updated according to the current fitness, the best historical fitness, and the global best position based on the preset position and speed update formula, specifically: Get the historical best fitness of each particle. If the current particle's fitness is better than its historical best fitness, update the particle's personal best position based on the preset position and speed update formula. By comparing the personal best positions of all particles, the particle with the best fitness is found, the solution of this particle is taken as the global optimal solution, and the global best position is updated.

6. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 1 is characterized in that: The fire signal feature subset is used to construct a fire warning model based on the random forest algorithm, and the triggering of the fire extinguishing sticker is controlled based on the prediction result of the fire warning model, specifically: Acquire historical fire data, obtain fire thresholds of different fire signal characteristics based on the historical fire data based on a regression analysis method, and mark the corresponding fire thresholds into a fire signal characteristic subset, wherein the fire signal characteristics include air pressure characteristics, current fluctuation characteristics, temperature characteristics, smoke concentration characteristics, and gas concentration characteristics; The annotated fire signal feature subsets are divided into a training set and a test set, and a corresponding number of decision trees are constructed according to the fire signal features in the training set; Train each decision tree to form an independent decision tree, and set the optimal split point based on the impurity of the Gini index; The root node is divided based on the optimal division point until the maximum depth of the tree is reached, the division stops, and the leaf node is generated; If the proportion of fire samples in a leaf node is greater than the proportion of non-fire samples, the label of the leaf node is fire, otherwise it is non-fire; Integrate the generated results of all decision trees and obtain the final prediction result based on the average mechanism; When the prediction result is fire, the fire extinguishing sticker is controlled to be triggered to carry out fire extinguishing.

7. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 1 is characterized in that: The fire location is obtained according to the position of the fire extinguishing sticker after the trigger, the fire extinguishing situation is judged and the fire extinguishing sticker is mobilized, specifically: After the fire extinguishing sticker is triggered, the fire warning model is used to make another prediction to determine whether the fire has been extinguished; If it is not extinguished, the fire extinguishing sticker will be mobilized; Each position in the fire extinguishing sticker matrix is ​​set as a node, and the neighborhood within a preset range is an edge; The traversal traverses the neighborhood by a breadth-first search method, and according to the spread of the fire source, the nodes closest to the fire source and the fire extinguishing posts in the direction of the fire source spread are preferentially traversed; The fire extinguishing sticker matrix is ​​updated in real time according to the mobilization results.

8. The method for intelligent triggering and controlling of fire extinguishing stickers based on the Internet of Things according to claim 4 is characterized in that: The specific calculation formula of the initial velocity of the particle is as follows: in , The classification accuracy calculation formula is as follows: In the formula, For particles At the moment The initial velocity, is the preset maximum speed. is the preset minimum speed. is a random number, is the classification accuracy, is the number of samples that are fire and are correctly classified as fire, The number of samples that are non-fire and correctly classified as non-fire, The number of non-fire samples that were misclassified as fire, The number of samples that were fire but were misclassified as non-fire.

9. A system using the method for intelligent triggering and controlling a fire extinguishing sticker based on the Internet of Things as claimed in any one of claims 1 to 8, characterized in that: It includes matrix construction and data processing module, feature optimization module, fire warning module and fire extinguishing sticker mobilization module. There are connections between the modules: The matrix construction and data processing module is used to construct the fire extinguishing sticker matrix of the area to be tested and obtain the initial fire signal data. The initial fire signal data is decomposed by DWT based on the blind source separation method and wavelet threshold denoising method to obtain fire signal characteristics of different scales. A feature optimization module, used for optimizing the fire signal features based on a particle swarm algorithm to obtain a fire signal feature subset; The fire warning module is used to construct a fire warning model based on the random forest algorithm using a subset of fire signal features, and to control the triggering of fire extinguishing stickers based on a dynamic threshold adjustment method; The fire extinguishing sticker mobilization module is used to obtain the fire location according to the position of the fire extinguishing sticker after triggering, judge the fire extinguishing situation and mobilize the fire extinguishing sticker.

Images