A method and system for controlling the release of fire extinguishing microcapsules based on artificial bee colony algorithm

Abstract

This invention discloses a method and system for controlling the release of fire extinguishing microcapsules based on an artificial bee colony algorithm, belonging to the field of fire safety technology. The method includes the following steps: acquiring monitoring data from multiple types of sensors; preprocessing the monitoring data based on a multi-modal noise dictionary to obtain a valid dataset; cross-validating the valid dataset using multi-parameter verification logic to obtain a fire confirmation result; and based on the fire confirmation result, locating the fire area using an artificial bee colony algorithm and controlling the release of fire extinguishing microcapsules within the fire area. This invention addresses the problems of low data reliability, insufficient accuracy in fire identification, and lack of precision in release control during the control of fire extinguishing microcapsule release.

Description

Technical Field

[0001] This invention relates to the field of fire safety technology, and more specifically, to a method and system for controlling the release of fire extinguishing microcapsules based on an artificial bee colony algorithm. Background Technology

[0002] Fire extinguishing microcapsules, as a new type of highly efficient fire extinguishing medium, have been increasingly used in construction, warehousing and precision equipment protection in recent years due to their advantages such as high fire extinguishing efficiency, minimal damage to equipment, and ease of deployment in complex scenarios.

[0003] However, existing publicly available methods for controlling the release of fire extinguishing microcapsules have at least the following technical problems: First, the triggering method is singular, making it prone to false triggering or delayed triggering. Currently, most microcapsule releases rely on a single temperature control module or a simple threshold alarm. Once the sensor is affected by noise or environmental interference, it may lead to false triggering, or delayed response due to unreasonable threshold settings, failing to achieve a rapid and effective response to real fires. Second, the accuracy of fire confirmation is insufficient, lacking intelligent identification. Existing systems often rely solely on temperature or smoke parameters for judgment, making it difficult to comprehensively judge the complex signals from multiple sources in the early stages of a fire, resulting in serious false alarms and missed alarms, limiting the scientific validity and reliability of microcapsule release. Third, the fire location is coarse, and the release lacks specificity. Fire extinguishing microcapsules mostly adopt a regional or overall release mode, lacking intelligent positioning and tracking of the actual fire spread area, easily leading to excessive release range and waste, or failure to cover key fire points, resulting in incomplete fire extinguishing. Fourth, the release control lacks dynamic optimization, resulting in limited execution efficiency. Most existing methods use fixed control logic, lacking intelligent optimization algorithms to dynamically adjust the release strategy according to the real-time fire situation, making it difficult to ensure a consistently high fire extinguishing effect throughout the fire's development.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a method and system for controlling the release of fire extinguishing microcapsules based on an artificial bee colony algorithm. Multi-source sensor data is preprocessed using a multimodal noise dictionary and combined with a graph neural network to confirm the fire situation. Based on this, the artificial bee colony algorithm is used to gradually converge and locate the fire area, ultimately triggering a control signal to execute the targeted release of the fire extinguishing microcapsules. This solves the problems of low data reliability, insufficient accuracy in fire identification, and lack of precision in release control during fire extinguishing microcapsule release.

[0006] To achieve the above objectives, the present invention provides the following technical solution: acquiring monitoring data from multiple types of sensors; preprocessing the monitoring data from multiple types of sensors based on a multimodal noise dictionary to obtain a valid dataset; cross-validating the valid dataset using multi-parameter verification logic to obtain a fire confirmation result; and based on the fire confirmation result, locating the fire area using an artificial bee colony algorithm and controlling the release of fire extinguishing microcapsules in the fire area.

[0007] In a preferred embodiment, the preprocessing of multi-type sensor monitoring data based on a multi-modal noise dictionary to obtain effective data specifically involves: constructing a multi-modal noise dictionary; classifying the multi-type sensor monitoring data according to sensor type to obtain a first dataset; comparing feature parameters of the first dataset based on the multi-modal noise dictionary and performing noise reduction using a Kalman filter algorithm to obtain a second dataset; and based on the spatiotemporal correlation of the second dataset of multiple types of sensors in the same region, completing and integrating the missing data in the second dataset to form an effective dataset.

[0008] In a preferred embodiment, the multimodal noise dictionary is generated based on historical noise data collected by multiple types of sensors under normal and interference environments, combined with the inherent characteristics of each sensor.

[0009] In a preferred embodiment, the missing data in the second dataset, based on the spatiotemporal correlation of multiple types of sensors in the same region, is completed and integrated to form a valid dataset. Specifically, this involves: extracting the timestamps and spatial coordinates of each data point in the second dataset, and calculating the temporal series correlation and spatial proximity between the data points; establishing a spatiotemporal correlation matrix based on the temporal series correlation and spatial proximity; constructing a spatiotemporal correlation model based on the spatiotemporal correlation matrix, and completing the missing data using Kriging spatial interpolation and LSTM time series prediction methods; aligning the completed data according to a unified spatiotemporal benchmark, and integrating the aligned data to generate a valid dataset.

[0010] In a preferred embodiment, the alignment includes: resampling time-series data at fixed time intervals and mapping spatial coordinates to standardized grid cells.

[0011] In a preferred embodiment, the cross-validation of the effective dataset through multi-parameter validation logic specifically involves: constructing a parameter correlation graph based on the effective dataset and inputting it into a pre-trained graph neural network model; processing the parameter correlation graph through the graph neural network model to mine nonlinear implicit correlations between multiple types of monitoring parameters and identify abnormal correlation patterns; comparing the abnormal correlation patterns with the fire feature database to select abnormal correlation combinations that conform to fire precursors; and generating a fire confirmation result by comprehensively considering the duration of the correlation combination and the change amplitude of multiple types of monitoring parameters.

[0012] In a preferred embodiment, the fire feature database contains parameter association patterns for typical fire occurrences and is dynamically updated based on historical fire data.

[0013] In a preferred embodiment, the step of locating the fire area using an artificial bee colony algorithm based on the fire confirmation result specifically involves: activating the sensor network within the monitoring area based on the fire confirmation result, dividing the monitoring area into several overlapping grid cells; generating multiple candidate fire areas by scout bees applying an adaptive search strategy within each grid cell, and calculating the fitness of these candidate fire areas based on multi-type sensor data; using candidate fire areas with fitness higher than a first preset threshold as the search starting point for leader bees, and conducting a local search within the neighborhood of that area, dynamically updating the candidate fire areas based on the fitness of the neighborhood; following bees, based on the updated candidate fire areas explored by the leader bees, selecting candidate fire areas with fitness higher than a second preset threshold as the optimal position, and iteratively updating the optimal position through depth search; when the change in the optimal position during iteration meets the convergence condition, outputting the optimal position as the coordinates of the fire area.

[0014] In a preferred embodiment, the convergence conditions include determining the convergence region and the convergence rate.

[0015] Secondly, this application provides a fire extinguishing microcapsule release control system based on a temperature control alarm module, including: a multi-source sensor data acquisition module for acquiring monitoring data from multiple types of sensors; a multi-modal noise dictionary preprocessing module for preprocessing the monitoring data from multiple types of sensors based on a multi-modal noise dictionary to obtain a valid dataset; a multi-parameter verification and fire confirmation module for cross-validating the valid dataset through multi-parameter verification logic to obtain a fire confirmation result; and a swarm positioning and microcapsule release control module for locating the fire area based on the fire confirmation result using an artificial swarm algorithm and controlling the release of fire extinguishing microcapsules in the fire area.

[0016] To overcome the shortcomings of existing technologies, this invention provides a method and system for controlling the release of fire extinguishing microcapsules based on an artificial bee colony algorithm. Multi-source sensor data is preprocessed using a multimodal noise dictionary and combined with a graph neural network to confirm the fire situation. Based on this, the artificial bee colony algorithm is used to gradually converge and locate the fire area, ultimately triggering a control signal to execute the targeted release of the fire extinguishing microcapsules. This solves the problems of low data reliability, insufficient accuracy in fire identification, and lack of precision in release control during fire extinguishing microcapsule release. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of the fire extinguishing microcapsule release control method based on the artificial bee colony algorithm of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of the fire extinguishing microcapsule release control system based on the temperature control alarm module of the present invention.

[0019] Figure 3 Flowchart for locating fire areas using the artificial bee colony algorithm. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1, Figure 1 The present invention provides a method for controlling the release of fire extinguishing microcapsules based on an artificial bee colony algorithm, comprising the following steps:

[0022] S1 acquires monitoring data from multiple types of sensors.

[0023] In this embodiment, acquiring multi-type sensor monitoring data refers to collecting environmental parameters of the monitoring area by deploying temperature sensors, smoke sensors, gas concentration sensors, and infrared sensors within the monitoring area to obtain multi-type sensor monitoring data including temperature, smoke concentration, gas component concentration, and infrared radiation intensity, which will then serve as input for subsequent multi-modal noise dictionary preprocessing.

[0024] S2, based on a multimodal noise dictionary, preprocesses monitoring data from multiple types of sensors to obtain an effective dataset.

[0025] In this embodiment, the preprocessing of multi-type sensor monitoring data based on a multimodal noise dictionary to obtain effective data specifically involves:

[0026] Construct a multimodal noise dictionary;

[0027] The monitoring data from multiple types of sensors are classified according to sensor type to obtain the first dataset;

[0028] The first dataset is obtained by comparing feature parameters based on a multimodal noise dictionary and then using a Kalman filter algorithm for noise reduction.

[0029] Based on the spatiotemporal correlation of a second dataset from multiple types of sensors in the same region, missing data in the second dataset are completed and integrated to form a valid dataset.

[0030] Specifically:

[0031] First, by statistically analyzing historical sample data collected by different types of sensors under normal and interference environments, and combining the inherent characteristic parameters of each sensor, common noise modes and interference characteristic parameters of each sensor are extracted, and stored according to sensor category to form a dictionary, which is the multimodal noise dictionary.

[0032] Then, the acquired multi-type sensor monitoring data are grouped according to sensor type, so that temperature, smoke, infrared and gas concentration observations form corresponding subsets, collectively referred to as the first dataset. Grouping by type helps to select matching dictionary entries for comparison in subsequent steps.

[0033] Subsequently, within a preset sliding time window (the window size can be set according to the sampling frequency), feature parameters for noise identification are extracted from each subset. The feature parameters include: mean, standard deviation, rate of change, peak-valley amplitude, and frequency domain energy or dominant frequency component, etc.

[0034] Then, it calls the noise entries in the multimodal noise dictionary that match the subset type and performs window-by-window comparison and similarity evaluation on the above feature parameters:

[0035] When the characteristics of the observation sequence within the sliding time window deviate significantly from the normal range or closely match a known noise pattern, the observation data within that time window is determined to be of high noise level; otherwise, it is determined to be of low noise level or normal level.

[0036] Based on this, a Kalman filter algorithm is introduced to reduce noise in the observation sequence. The measurement noise covariance during the filtering process... Adaptive settings based on dictionary comparison results:

[0037] When the observed data has a high noise level When the observed data has a high noise level, ;

[0038] Through the above adaptive adjustment, the filter suppresses the influence of abnormal observations when the noise is large, and outputs a smoothed estimation sequence, which constitutes the second dataset after noise reduction;

[0039] In this embodiment, based on the spatiotemporal correlation of the second dataset from multiple types of sensors in the same region, missing data in the second dataset is completed and integrated to form a valid dataset, specifically as follows:

[0040] Extract the timestamps and spatial coordinates of each data point in the second dataset, and calculate the temporal series correlation and spatial proximity between the data points.

[0041] A spatiotemporal correlation matrix is ​​established based on time series correlation and spatial proximity;

[0042] A spatiotemporal correlation model is constructed based on the spatiotemporal correlation matrix, and missing data is filled in using the Kriging spatial interpolation method and the LSTM time series prediction method.

[0043] The completed data is aligned according to a unified spatiotemporal benchmark, and the aligned data is then integrated to generate a valid dataset.

[0044] Specifically:

[0045] Obtain the timestamp of each observation record from the second dataset. with spatial coordinates Based on this, the temporal correlation and spatial proximity between observation sequences are calculated:

[0046] The time correlation was measured using the Pearson correlation coefficient:

[0047]

[0048] in, and These represent the sensors. With sensors A set of time series observations; The covariance between two time series; and They are respectively and Standard deviation;

[0049] The spatial proximity is determined using Euclidean distance:

[0050]

[0051] in, and Sensors With sensors The planar coordinate position;

[0052] Subsequently, combining temporal correlation and spatial proximity, a spatiotemporal correlation matrix is ​​defined. :

[0053] ,

[0054] in, , This is a weighting factor used to balance the importance of temporal correlation and spatial proximity; The spatial attenuation coefficient controls the influence of distance on the correlation degree.

[0055] matrix elements As a weight for information propagation between adjacent sensors, it is used to construct a spatiotemporal weighted graph model. ,in For a set of sensor nodes, The spatiotemporal weighted graph model serves as the connection edge between sensors; it provides weight constraints for the missing data completion process.

[0056] Furthermore, based on the spatiotemporal weighted graph model, the Kriging spatial interpolation method is used to predict the missing data in the spatial dimension, and the LSTM time series method is used to predict the missing data in the temporal dimension, thereby achieving the completion of missing data;

[0057] The completed data will be aligned according to a unified spatiotemporal reference.

[0058] In the time dimension, with a preset fixed sampling interval All sensor data are resampled. That is, for any sensor's time series... By using interpolation or mean methods, they are uniformly mapped to time series. This eliminates the deviation caused by inconsistencies in sampling frequencies or sampling times of different sensors;

[0059] In the spatial dimension, the planar position coordinates of the sensor Standardized mapping is performed. Specifically, the monitoring area is divided into regularized grid cells, each cell representing a standardized spatial location, and sensor coordinates are assigned to the corresponding grid cell. If multiple sensor data exist within the same grid cell, they can be fused using weighted averaging or priority rules.

[0060] Through the aforementioned temporal resampling and spatial gridding mapping, all sensor data are aligned to a unified spatiotemporal reference, and then integrated to generate an effective dataset.

[0061] The multimodal noise dictionary is generated based on historical noise data collected by various types of sensors under normal and interference conditions, combined with the inherent characteristics of each sensor.

[0062] The alignment includes: resampling time-series data at fixed time intervals and mapping spatial coordinates to standardized grid cells;

[0063] It should be noted that the missing data is usually caused by sensor failure, communication delay, sampling loss or environmental interference, resulting in some incomplete records in the second dataset.

[0064] S3 uses multi-parameter validation logic to cross-validate the valid dataset to obtain the fire confirmation result.

[0065] In this embodiment, the cross-validation of the valid dataset using multi-parameter validation logic specifically involves:

[0066] Construct a parameter association graph based on an effective dataset and input it into a pre-trained graph neural network model;

[0067] By processing the parameter correlation graph using a graph neural network model, we can uncover the nonlinear implicit correlations between various types of monitoring parameters and identify abnormal correlation patterns.

[0068] The abnormal association patterns are compared with the fire feature database to filter out abnormal association combinations that match the precursory signs of a fire.

[0069] The fire confirmation result is generated by combining the duration of the associated parameters and the variation range of multiple types of monitoring parameters.

[0070] Specifically:

[0071] First, a parameter association graph is constructed based on the valid dataset;

[0072] This parameter correlation diagram uses multiple types of monitoring parameters as nodes, such as temperature, smoke concentration, infrared intensity, and harmful gas concentration; and the spatiotemporal coupling relationship between different monitoring parameters as edges, where the spatiotemporal coupling relationship is measured by temporal correlation and spatial proximity. The resulting parameter correlation diagram can reflect the inherent relationship between multiple source parameters within the monitoring area. The fire feature database contains parameter correlation patterns during typical fire occurrences and is dynamically updated based on historical fire data.

[0073] Then, the parameter association graph is input into the pre-trained graph neural network model. Through graph convolution and other mechanisms, the nodes and edges are propagated and aggregated to explore the nonlinear implicit associations between various types of monitoring parameters at the graph structure level. For example, the coupling change pattern of temperature and smoke, and the linkage between infrared signal and gas concentration can all be learned by the model.

[0074] The model output results can identify parameter combinations with abnormal characteristics, i.e. abnormal correlation patterns, and compare the identified abnormal correlation patterns with the fire feature database.

[0075] The fire feature database contains parameter association patterns for typical fire occurrences and is dynamically updated based on historical fire data, such as typical combinations like a continuous rise in temperature accompanied by a slow increase in smoke concentration, a sharp increase in infrared radiation, and the release of harmful gases. By comparison, abnormal association combinations that are highly consistent with fire precursors can be screened out, thereby reducing false alarms.

[0076] Finally, the duration of the abnormal association combination and the magnitude of the changes in the monitored parameters are comprehensively determined. When an abnormal combination continues for more than a preset threshold in time, and the magnitude of its parameter changes reaches or exceeds a set range, the system generates a fire confirmation result for subsequent use in the control of fire extinguishing microcapsule release.

[0077] S4, based on the fire confirmation results, uses an artificial bee colony algorithm to locate the fire area and controls the release of fire extinguishing microcapsules in the fire area.

[0078] In this embodiment, the step of locating the fire area using an artificial bee colony algorithm based on the fire confirmation result specifically involves:

[0079] Based on the fire confirmation results, the sensor network in the monitoring area is activated, and the monitoring area is divided into several overlapping grid units;

[0080] By applying an adaptive search strategy within each grid cell, the reconnaissance bee generates multiple candidate fire zones and calculates the fitness of these candidate fire zones based on multi-type sensor data.

[0081] Candidate fire areas with fitness values ​​higher than a first preset threshold are used as the starting point for the leader bee's search, and a local search is performed in the neighborhood of that area. The candidate fire areas are dynamically updated based on the fitness values ​​of the neighborhood.

[0082] The follower bee explores updated candidate fire areas based on the leader bee's exploration, selects candidate fire areas with fitness higher than the second preset threshold as the optimal location, and iteratively updates the optimal location through deep search;

[0083] When the variation of the optimal position during the iteration meets the convergence condition, the optimal position is output as the coordinates of the fire area.

[0084] Specifically:

[0085] First, based on the fire confirmation results, activate the multi-type sensor network within the corresponding monitoring area, and then proceed according to a fixed spatial step size. , The monitoring area is divided into several overlapping network units with equal intervals.

[0086] The scout bee searches within each grid cell using an adaptive search strategy;

[0087] The adaptive search strategy refers to the reconnaissance bee dynamically adjusting its search strategy based on sensor data (such as temperature, smoke concentration, etc.) and fire confirmation results. For example, areas with higher temperature and smoke concentration will be searched first, and the reconnaissance bee will generate multiple candidate fire areas in these areas. The reconnaissance bee uses the adaptive search strategy to optimize the search direction and step size to ensure effective searching within potential fire areas.

[0088] The scout bee generates multiple candidate fire regions within each grid cell by searching; the candidate fire regions are represented as candidate fire region vectors.

[0089]

[0090] in, Indicates the first The coordinates of each grid position;

[0091] Each candidate fire zone represents the location within that area that the scout bee believes is most likely to be a fire;

[0092] Subsequently, based on monitoring data from multiple types of sensors within the grid location, the fitness of the candidate fire area is calculated. :

[0093]

[0094] in, This indicates the difference in temperature between the candidate fire area and the normal temperature. This indicates anomalies in smoke concentration or magnitude of smoke gradient in the candidate fire area; This indicates the abnormal gas concentration value in the candidate fire area; The radiation intensity anomaly value for the candidate fire area; , , , This is a weighting factor used to balance the importance of different sensor parameters;

[0095] When fitness When the fire rate exceeds the first preset threshold, the corresponding candidate fire area is selected as the search starting point for the leader bee. Then, the leader bee performs a local search within the neighborhood of the search starting point and dynamically updates the candidate fire area based on the fitness of the neighborhood.

[0096] The step of dynamically updating the candidate fire area based on the fitness of the neighborhood is as follows: when the fitness of the searched neighborhood location is higher than the current fitness, the neighborhood location is updated as a new candidate fire area; otherwise, the original candidate fire area is retained.

[0097] Furthermore, the follower bee selects the candidate fire area with a fitness higher than the second preset threshold from the candidate fire area updated by the leader bee as the optimal location, and performs a deep search within the optimal location area;

[0098] During the depth search, the bee continuously updates the optimal position based on fitness feedback. In each iteration, the bee compares the fitness difference between the current optimal position and the previous position; if the current searched position has a higher fitness, the optimal position is updated.

[0099] When the variation of the optimal position during the iteration meets the convergence condition, the search is considered to have converged, and the optimal position at this time is output as the coordinates of the fire area.

[0100] The convergence conditions include the determination of the convergence region range and the convergence speed. The convergence region range refers to the change amplitude of the optimal position being less than a predetermined change threshold, and the convergence speed refers to the rate of change of the optimal position gradually decreasing to below a set convergence speed threshold.

[0101] Once the coordinates of the fire area are obtained, a control signal is immediately triggered to control the release of fire extinguishing microcapsules in the fire area.

[0102] It should be noted that the second preset threshold is higher than the first preset threshold.

[0103] In its specific implementation, the execution flow of the artificial bee colony algorithm is as follows: Figure 3 As shown, this process gradually converges to obtain the optimal coordinates of the fire area through adaptive search by scout bees, local search and update by guide bees, and deep search iteration by follower bees, thereby triggering the release of fire extinguishing microcapsules.

[0104] Figure 2 As shown, this embodiment discloses a fire extinguishing microcapsule release control system based on a temperature control alarm module, including:

[0105] Multi-source sensor data acquisition module, used to acquire monitoring data from multiple types of sensors;

[0106] The multimodal noise dictionary preprocessing module is used to preprocess monitoring data from multiple types of sensors based on the multimodal noise dictionary to obtain an effective dataset;

[0107] The multi-parameter verification and fire confirmation module is used to cross-validate the valid dataset through multi-parameter verification logic to obtain the fire confirmation result;

[0108] The swarm positioning and microcapsule release control module is used to locate the fire area based on the fire confirmation result using an artificial swarm algorithm, and to control the release of fire extinguishing microcapsules in the fire area.

[0109] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0110] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0111] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0112] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0114] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the release of fire-extinguishing microcapsules based on an artificial bee colony algorithm, characterized in that, The method comprises the following steps: acquiring multi-type sensor monitoring data; preprocessing the multi-type sensor monitoring data based on a multi-modal noise dictionary to obtain an effective data set, specifically: constructing a multi-modal noise dictionary; classifying the multi-type sensor monitoring data according to sensor types to obtain a first data set; comparing feature parameters of the first data set based on the multi-modal noise dictionary, and performing noise reduction using a Kalman filtering algorithm to obtain a second data set; completing and integrating missing data in the second data set based on the spatio-temporal correlation of the second data set of the same area multi-type sensors to form an effective data set; cross-validating the effective data set through multi-parameter verification logic to obtain a fire condition confirmation result, specifically: constructing a parameter correlation graph based on the effective data set and inputting a pre-trained graph neural network model; processing the parameter correlation graph through the graph neural network model to mine the non-linear implicit correlation between the multi-type monitoring parameters and identify abnormal correlation patterns; comparing the abnormal correlation patterns with a fire condition feature library to screen abnormal correlation combinations that meet the fire condition precursors; generating a fire condition confirmation result by comprehensively considering the duration of the correlation combinations and the change range of the multi-type monitoring parameters; locating a fire condition area based on the fire condition confirmation result, specifically: based on the fire condition confirmation result, activating the sensor network in the monitoring area, and dividing the monitoring area into a plurality of overlapping grid units; generating a plurality of candidate fire condition areas by applying an adaptive search strategy in each grid unit through a scout bee, and calculating the fitness of these candidate fire condition areas based on multi-type sensor data; selecting a candidate fire condition area with a fitness higher than a first preset threshold as a search starting point of a leader bee, and performing local search in the neighborhood of the area, and dynamically updating the candidate fire condition area according to the fitness of the neighborhood; following the leader bee to explore the updated candidate fire condition area, selecting a candidate fire condition area with a fitness higher than a second preset threshold as an optimal position, and iteratively updating the optimal position through deep search; when the optimal position in the iteration changes within a range satisfying a convergence condition, outputting the optimal position as the coordinates of the fire condition area; and controlling the fire condition area to release fire-extinguishing microcapsules.

2. The artificial bee colony algorithm-based release control method of fire extinguishing microcapsules according to claim 1, characterized by, The multi-modal noise dictionary is generated based on historical noise data collected by multi-type sensors in normal and interference environments, and combined with the inherent characteristics of each sensor.

3. The artificial bee colony algorithm-based release control method of fire extinguishing microcapsules according to claim 2, characterized by, The missing data in the second data set is completed and integrated based on the spatio-temporal correlation of the second data set of the same area multi-type sensors to form an effective data set, specifically: extracting the time stamp and spatial coordinates of each data in the second data set, and calculating the time sequence correlation and spatial proximity between the data; establishing a spatio-temporal correlation matrix based on the time sequence correlation and spatial proximity; completing the missing data by constructing a spatio-temporal correlation model according to the spatio-temporal correlation matrix, and using the Kriging spatial interpolation method and the LSTM time series prediction method; aligning the completed data according to a unified spatio-temporal reference, and integrating the aligned data to generate an effective data set.

4. The artificial bee colony algorithm-based release control method of fire extinguishing microcapsules according to claim 3, characterized by, The alignment includes: resampling the time series data at a fixed time interval, and mapping the spatial coordinates to standardized grid units.

5. The artificial bee colony algorithm-based release control method of fire extinguishing microcapsules according to claim 4, characterized by, The fire feature library comprises a parameter correlation mode when a typical fire occurs, and is dynamically updated based on historical fire data.

6. The artificial bee colony algorithm-based release control method of fire extinguishing microcapsules according to claim 5, characterized by, The convergence condition comprises a convergence area range and a convergence speed determination.

7. A system using the artificial bee colony algorithm-based fire-extinguishing microcapsule release control method of any one of claims 1-6, comprising: a multi-source sensing data acquisition module for acquiring multi-type sensor monitoring data; a multi-modal noise dictionary preprocessing module for preprocessing the multi-type sensor monitoring data based on a multi-modal noise dictionary to obtain an effective data set; a multi-parameter verification and fire confirmation module for cross- verifying the effective data set through multi-parameter verification logic to obtain a fire confirmation result; a bee colony positioning and microcapsule release control module for positioning a fire area through an artificial bee colony algorithm and controlling release of fire-extinguishing microcapsules in the fire area based on the fire confirmation result.

Images