A control method for an integrated big data platform for security and fire protection based on artificial intelligence

Abstract

This invention discloses a control method for an integrated big data platform for safety and fire prevention based on artificial intelligence, relating to the field of safety production and fire management technology. The method includes acquiring operational and historical data, standardizing the operational data to generate a training dataset, training and constructing a deep convolutional neural network model, extracting spatial correlation features through multi-layer convolution operations, and identifying key state information using an attention mechanism to generate a feature mapping matrix. The feature mapping matrix is ​​then input into a constructed safety assessment model to calculate a safety risk index, and risk levels are classified according to the safety risk index. Based on the risk level classification results, an inspection task list is generated and distributed to mobile terminals for execution, achieving dynamic control of safety risks. This invention overcomes the technical bottlenecks of traditional integrated safety and fire prevention platforms in data analysis, risk identification, and control, achieving intelligent and precise safety risk management and improving safety management efficiency.

Description

Technical Field

[0001] This invention relates to the field of safety production and fire management technology, and in particular to a control method for an integrated big data platform for safety and fire protection based on artificial intelligence. Background Technology

[0002] Traditional safety production and fire management models suffer from problems such as information silos, data fragmentation, and low management efficiency, making it difficult to meet the needs of modern management. In recent years, integrated safety and fire management platforms based on big data technology have gradually emerged, achieving coordinated linkage between safety management and fire management by integrating safety production data and fire monitoring data. However, existing integrated safety and fire management platforms generally suffer from insufficient data analysis capabilities, untimely risk warnings, and low levels of intelligence. Specifically, these issues manifest in the following ways: First, their ability to process massive amounts of heterogeneous data is limited, making it difficult to fully extract the deep correlations contained within the data; second, risk assessment methods rely too heavily on empirical models and expert knowledge, lacking effective capture of dynamic change characteristics; and third, system response is relatively slow, failing to achieve real-time risk perception and intelligent early warning.

[0003] To address these issues, researchers have begun to explore the integration of deep learning technology into integrated safety and fire management platforms to enhance the system's intelligence. Some studies have already applied convolutional neural networks to safety risk identification, achieving some success. However, existing solutions still suffer from the following shortcomings: first, limited feature extraction capabilities, making it difficult to effectively identify complex spatiotemporal correlation features in multidimensional data; second, insufficient model generalization ability, resulting in poor adaptability to novel risk scenarios; and third, a lack of effective risk classification mechanisms and dynamic control measures, hindering precise risk management.

[0004] To address these issues, this invention proposes a control method for an integrated big data platform for safety and fire protection based on artificial intelligence. By organically combining deep convolutional neural networks and attention mechanisms, it achieves intelligent identification and dynamic control of safety risks, providing a new technical approach to improve the level of integrated safety and fire protection management. Summary of the Invention

[0005] In view of the problems existing in the integrated big data platform for safety and fire protection in terms of feature extraction capability, model generalization performance and risk control mechanism, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to improve the intelligence level of the integrated big data platform for security and fire protection through deep learning technology, so as to achieve accurate identification and dynamic control of security risks.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide a method for controlling an integrated big data platform for safety and fire protection based on artificial intelligence. The method includes: acquiring operational data and historical data of the integrated big data platform for safety and fire protection, and standardizing the operational data to generate a training dataset; training and constructing a deep convolutional neural network model based on the training dataset, extracting spatial correlation features through multi-layer convolution operations, and identifying key state information using an attention mechanism to generate a feature mapping matrix; inputting the feature mapping matrix into a constructed safety assessment model to calculate a safety risk index, and classifying risk levels according to the safety risk index; generating an inspection task list based on the risk level classification results, and distributing it to a mobile terminal for execution, thereby achieving dynamic control of safety risks.

[0009] As a preferred embodiment of the AI-based integrated big data platform control method for safety and fire prevention described in this invention, the method involves: generating an inspection task list based on risk level classification results and distributing it to mobile terminals for execution, thereby achieving dynamic control of safety risks. This includes: generating an inspection task list based on risk level classification results and allocating inspection resources using a task priority ranking algorithm. The inspection task list includes a first-level inspection task list, a second-level inspection task list, and a third-level inspection task list. The inspection task list includes inspection paths, monitoring point location coordinates, and inspection frequency requirements. If the safety risk level is level four, a third-level inspection task list is generated. If the safety risk level is level four, a third-level inspection task list is generated. If the risk level is Level 3, a Level 2 inspection task list is generated; if the safety risk level is Level 2, a Level 1 inspection task list is generated. The inspection task list is then pushed to the mobile terminal devices of the inspection personnel to conduct inspections. Image data and environmental parameters of the inspection points are collected through the mobile terminal devices, which plan the optimal inspection route based on a built-in electronic map navigation module. The image data and environmental parameters are analyzed, and the analysis results are transmitted to the integrated safety and fire protection big data platform, outputting inspection feedback data. The parameters of the risk assessment model are updated using the inspection feedback data, and the risk level classification results are dynamically corrected.

[0010] As a preferred embodiment of the AI-based integrated big data platform control method for safety and fire protection described in this invention, the method includes: inputting the feature mapping matrix into a constructed safety assessment model, calculating a safety risk index, and classifying risk levels according to the safety risk index, including: constructing a nonlinear mapping space based on historical data using a radial basis function kernel function, and selecting kernel function parameters and penalty factors through cross-validation to construct a safety assessment model; calculating weight coefficients for spatial and temporal features in the feature mapping matrix respectively, and constructing a composite feature vector using a weighted combination method; inputting the composite feature vector into the safety assessment model, calculating the distance from sample points to the classification hyperplane to obtain the safety risk index; setting a grading threshold for the safety risk index based on the statistical distribution of historical data, calculating the trend of safety risk index changes at multiple consecutive time points using a sliding time window, establishing risk warning levels, and setting warning strategies.

[0011] As a preferred embodiment of the integrated big data platform control method for security and fire protection based on artificial intelligence described in this invention, the specific formula of the security assessment model is as follows:

[0012]

[0013] Where R(t) is the safety risk index at time t, and α i Let x be the support vector coefficient of the i-th sample. i For the support vector samples in the historical data, K(x) i H(t) is the improved radial basis kernel function, H(t) is the eigenmap matrix, b is the bias term, λ is the regularization coefficient, and w j F represents the weight coefficient of the j-th feature dimension. j (t) represents the temporal feature integral term of the j-th feature dimension. is the time window smoothing function, n is the number of samples, and m is the extracted feature dimension.

[0014] As a preferred embodiment of the integrated big data platform control method for security and fire protection based on artificial intelligence described in this invention, it further includes: when the security risk index R(t) < the first threshold and within a continuous time window |R(t) k )-R(t k-1 If the first threshold is less than the preset threshold, it is classified as a Level 1 risk; if the first threshold is less than or equal to the safety risk index R(t) and less than the second threshold, and the risk index R(t) at the current time k is less than or equal to the preset threshold, it is classified as a Level 1 risk. k The risk index R(t) of the previous time k-1 k-1If the second threshold is less than or equal to the safety risk index R(t) and the third threshold is less than the third threshold, and the safety risk index R(t) remains within this range within a continuous time window, then it is classified as a level 3 risk, triggering a level 2 warning. This involves adjusting the inspection route, deploying additional professional inspectors, and developing an emergency plan. If the safety risk index R(t) is greater than or equal to the third threshold or if the safety risk index R(t) remains within this range within a continuous time window, then it is classified as a level 3 risk, triggering a level 2 warning. This requires adjusting the inspection route, deploying additional professional inspectors, and developing an emergency plan. k )-R(t k-1 If the preset threshold is reached, it is determined to be a level four risk, triggering a level three warning, activating the emergency response mechanism, and implementing special control measures.

[0015] As a preferred embodiment of the AI-based integrated big data platform control method for safety and fire protection described in this invention, the method for generating the feature mapping matrix is ​​as follows: A deep convolutional neural network model is established based on a deep learning framework, and the training dataset is divided into a training set and a validation set according to a predetermined ratio. Simultaneously, the deep convolutional neural network model is optimized and trained using mini-batch stochastic gradient descent. The deep convolutional neural network model includes an input layer, multiple convolutional layers, pooling layers, fully connected layers, and an output layer. Spatial correlation features are extracted through multi-layer convolutional operations. A self-attention mechanism is introduced to calculate the weight coefficients between different channels, and the channel attention weights and spatial attention weights are multiplied to obtain fused attention weights. The fused attention weights and the original feature map are weighted and summed, and the weighted features are then dimensionality-reduced through a fully connected layer to generate a feature mapping matrix. The feature mapping matrix has a dimension of n×m, where n represents the number of samples and m represents the extracted feature dimension.

[0016] As a preferred embodiment of the AI-based integrated big data platform control method for safety and fire protection described in this invention, the method for generating the training dataset is as follows: A multi-source data acquisition channel is established, and operational data of the integrated big data platform for safety and fire protection is collected through IoT sensors. This operational data includes monitoring data, fire protection facility distribution data, and safety inspection data. The monitoring data is cleaned to remove outliers and duplicates, and missing values ​​are supplemented using linear interpolation. Simultaneously, the fire protection facility distribution data undergoes coordinate unification and scale standardization, and the safety inspection data undergoes time series alignment and format unification. Finally, the maximum and minimum value normalization method is used to map all processed data to the [0,1] interval to generate the training dataset.

[0017] Secondly, embodiments of the present invention provide an artificial intelligence-based integrated big data platform control system for safety and fire prevention, comprising: an acquisition module for acquiring operational data and historical data of the integrated big data platform for safety and fire prevention, and standardizing the operational data to generate a training dataset; a feature extraction module for training and constructing a deep convolutional neural network model based on the training dataset, extracting spatial correlation features through multi-layer convolution operations, and identifying key state information by combining an attention mechanism to generate a feature mapping matrix; a safety assessment and risk calculation module for inputting the feature mapping matrix into the constructed safety assessment model, calculating a safety risk index, and classifying risk levels according to the safety risk index; and an inspection task generation and distribution module for generating an inspection task list based on the risk level classification results and distributing it to a mobile terminal for execution, thereby realizing dynamic control of safety risks.

[0018] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program instructions, when executed by the processor, implement the steps of the artificial intelligence-based integrated big data platform control method for security and fire protection as described in the first aspect of the present invention.

[0019] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the artificial intelligence-based integrated security and fire protection big data platform control method described in the first aspect of the present invention.

[0020] The beneficial effects of this invention are as follows: It solves the problem of unifying heterogeneous data by standardizing the processing of operational and historical data; it improves the accuracy of risk identification by using a deep convolutional neural network model combined with an attention mechanism to achieve deep feature extraction and key information identification from multidimensional data; it enables quantitative assessment and hierarchical management of risks through a feature mapping matrix-based security assessment model, providing a scientific basis for risk control; it achieves closed-loop control of risk management by intelligently generating and distributing inspection tasks in real time; and the overall solution breaks through the technical bottlenecks of traditional integrated security and fire protection platforms in data analysis, risk identification, and control, achieving intelligent and precise security risk management and significantly improving security management efficiency. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0022] Figure 1This is a flowchart of the control method of the integrated big data platform for safety and fire protection based on artificial intelligence, as described in Example 1. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1

[0027] Reference Figure 1 This is the first embodiment of the present invention, which provides a control method for an integrated big data platform for security and fire protection based on artificial intelligence, including:

[0028] S1: Obtain the operational data and historical data of the integrated security and fire protection big data platform, and perform standardized processing on the operational data to generate a training dataset.

[0029] Specifically, the training dataset is generated by establishing a multi-source data acquisition channel and collecting operational data from the integrated safety and fire protection big data platform through IoT sensors. The operational data includes monitoring data, fire protection facility distribution data, and safety inspection data.

[0030] It should be noted that the monitoring data includes temperature data collected by temperature sensors, smoke concentration data collected by smoke sensors, harmful gas concentration data collected by gas sensors, and pressure data collected by pressure sensors; simultaneously, data on the distribution of fire protection facilities is collected, including data on the location of fire hydrants, fire extinguishers, emergency lighting equipment, and fire lanes; and safety inspection data uploaded by inspection personnel is collected, including inspection route data, inspection point data, inspection item data, and inspection result data.

[0031] Furthermore, the monitoring data is cleaned to remove outliers and duplicates, and missing values ​​are supplemented using linear interpolation. At the same time, the distribution data of fire protection facilities is processed to unify coordinates and scale, and the safety inspection data is processed to align time series and unify format.

[0032] Furthermore, the maximum-minimum normalization method is used to map all processed data to the [0,1] interval to generate a training dataset.

[0033] S2: Based on the training dataset, train and construct a deep convolutional neural network model, extract spatial correlation features through multi-layer convolution operations, and identify key state information by combining an attention mechanism to generate a feature mapping matrix.

[0034] Specifically, the method for generating the feature mapping matrix is ​​as follows: a deep convolutional neural network model is established based on the deep learning framework, and the training dataset is divided into a training set and a validation set according to the expected ratio; at the same time, the mini-batch stochastic gradient descent method is used to optimize and train the deep convolutional neural network model, which includes an input layer, multiple convolutional layers, pooling layers, fully connected layers, and an output layer.

[0035] It should be noted that the input layer receives the training dataset and reconstructs the data into a three-dimensional tensor form; the first convolutional layer uses a 3×3 convolutional kernel to extract features from the input data, extracting low-level spatial features; the second convolutional layer uses a 5×5 convolutional kernel to perform feature mapping, extracting mid-level semantic features; the third convolutional layer uses a 7×7 convolutional kernel to combine features, extracting high-level abstract features; and a max pooling layer is set between each convolutional layer to compress the feature dimension through a 2×2 sliding window, reducing computational complexity.

[0036] Furthermore, spatial correlation features are extracted through multi-layer convolution operations, a self-attention mechanism is introduced to calculate the weight coefficients between different channels, and the channel attention weights and spatial attention weights are multiplied to obtain the fused attention weights. The fused attention weights and the original feature maps are weighted and summed, and the weighted features are then dimensionality-reduced and mapped through a fully connected layer to generate a feature mapping matrix. The feature mapping matrix has a dimension of n×m, where n represents the number of samples and m represents the dimension of the extracted features.

[0037] S3: Input the feature mapping matrix into the constructed security assessment model, calculate the security risk index, and classify the risk level according to the security risk index.

[0038] Specifically, based on historical data, a nonlinear mapping space is constructed using a radial basis function kernel, and a safety assessment model is built by selecting kernel function parameters and penalty factors through cross-validation.

[0039] Furthermore, the specific formula for the security assessment model is as follows:

[0040]

[0041] Where R(t) is the safety risk index at time t, and α i Let x be the support vector coefficient of the i-th sample. i For the support vector samples in the historical data, K(x) i H(t) is the improved radial basis kernel function, H(t) is the eigenmap matrix, b is the bias term, λ is the regularization coefficient, and w j Let be the weight coefficient of the j-th feature dimension.

[0042] F j (t) represents the temporal feature integral term of the j-th feature dimension. is the time window smoothing function, n is the number of samples, and m is the extracted feature dimension.

[0043] Furthermore, the specific formula for the improved radial basis kernel function is as follows:

[0044]

[0045] Where β is the feature correlation adjustment coefficient and θ is the angle between the feature vectors.

[0046] It should be noted that the specific formulas for the time window smoothing function and the integral term of the time series features are as follows:

[0047]

[0048] Where W is the time window value, γ is the time decay coefficient, τ is the integration time range, and R(t) k R(t) represents the risk index at the current time k. k-1 ) represents the risk index of k-1 at the previous moment.

[0049] Specifically, weight coefficients are calculated for the spatial and temporal features in the feature mapping matrix, and a composite feature vector is constructed using a weighted combination method. The composite feature vector is then input into the security assessment model to calculate the distance from the sample point to the classification hyperplane, thereby obtaining the security risk index. Based on the statistical distribution of historical data, a grading threshold for the security risk index is set. A sliding time window is used to calculate the trend of the security risk index changes over multiple consecutive time points, thereby establishing a risk warning level and setting a warning strategy.

[0050] Furthermore, when the safety risk index R(t) < the first threshold and within a continuous time window |R(t) k )-R(t k-1If the first threshold is less than the preset threshold, it is classified as a Level 1 risk; if the first threshold is less than or equal to the safety risk index R(t) and less than the second threshold, and the risk index R(t) at the current time k is less than or equal to the preset threshold, it is classified as a Level 1 risk. k The risk index R(t) of the previous time k-1 k-1 If the second threshold is less than or equal to the safety risk index R(t) and the third threshold is less than the third threshold, and the safety risk index R(t) remains within this range within a continuous time window, then it is classified as a level 3 risk, triggering a level 2 warning. This involves adjusting the inspection route, deploying additional professional inspectors, and developing an emergency plan. If the safety risk index R(t) is greater than or equal to the third threshold or if the safety risk index R(t) remains within this range within a continuous time window, then it is classified as a level 3 risk, triggering a level 2 warning. This requires adjusting the inspection route, deploying additional professional inspectors, and developing an emergency plan. k )-R(t k-1 If the preset threshold is reached, it is determined to be a level four risk, triggering a level three warning, activating the emergency response mechanism, and implementing special control measures.

[0051] It should be noted that the first threshold is determined based on the normal fluctuation range of safe operation status in historical data; the second threshold is determined based on the distribution value of minor abnormal states in historical data statistical analysis; the third threshold is determined based on the risk accumulation threshold in the early stages of historical major safety accidents; the preset threshold is determined based on the risk accumulation effect and risk diffusion trend within a continuous time window; measures for Level 1 warning include increasing the monitoring frequency of key areas and adjusting daily inspection plans; measures for Level 2 warning include adjusting inspection routes, dispatching more professional inspection personnel, and formulating emergency plans; measures for Level 3 warning include activating the emergency response mechanism and implementing special control measures, including evacuating personnel, cutting off hazardous sources, and activating fire-fighting facilities.

[0052] S4: Generate an inspection task list based on the risk level classification results and send it to the mobile terminal for execution, so as to realize dynamic control of safety risks.

[0053] Specifically, based on the risk level classification results, an inspection task list is generated, and inspection resources are allocated using a task priority ranking algorithm. The inspection task list is then pushed to the mobile terminal devices of the inspection personnel to carry out the inspection work. The mobile terminal devices plan the optimal inspection route based on the built-in electronic map navigation module and collect image data and environmental parameters of the inspection points through the mobile terminal devices.

[0054] It should be noted that the inspection task list includes a Level 1 inspection task list, a Level 2 inspection task list, and a Level 3 inspection task list. Each task list includes inspection routes, monitoring point location coordinates, and required inspection frequency. The Level 1 inspection task list corresponds to Level 2 risk situations, primarily involving increasing the frequency of daily inspections, focusing on abnormal areas based on existing inspection routes, appropriately increasing the density of monitoring point locations, and raising the required inspection frequency from once per day to twice per day. The Level 2 inspection task list corresponds to Level 3 risk situations, requiring adjustments and optimization of inspection routes, adding temporary monitoring points in key areas, ensuring monitoring point location coordinates cover all risky and hazard areas, increasing the required inspection frequency to four times per day, and requiring the participation of professional inspection personnel. The Level 3 inspection task list corresponds to the highest level, Level 4 risk situations, requiring the replanning of emergency inspection routes, deploying dense monitoring points in high-risk areas, ensuring full coverage of monitoring point location coordinates and increased density in key areas, increasing the required inspection frequency to once every two hours, and having professional emergency personnel perform the inspection tasks.

[0055] Furthermore, if the safety risk level is level four, a third inspection task list is generated; if the safety risk level is level three, a second-level inspection task list is generated; and if the safety risk level is level two, a first-level inspection task list is generated.

[0056] Furthermore, the image data and environmental parameters are analyzed, and the analysis results are transmitted to the integrated safety and fire protection big data platform to output inspection feedback data. The parameters of the risk assessment model are updated based on the inspection feedback data, and the risk level classification results are dynamically corrected.

[0057] Furthermore, this embodiment also provides an AI-based integrated big data platform control system for safety and fire prevention, comprising: an acquisition module for acquiring operational data and historical data of the integrated big data platform for safety and fire prevention, and standardizing the operational data to generate a training dataset; a feature extraction module for training and constructing a deep convolutional neural network model based on the training dataset, extracting spatial correlation features through multi-layer convolution operations, and identifying key state information by combining an attention mechanism to generate a feature mapping matrix; a safety assessment and risk calculation module for inputting the feature mapping matrix into the constructed safety assessment model, calculating a safety risk index, and classifying risk levels according to the safety risk index; and an inspection task generation and distribution module for generating an inspection task list based on the risk level classification results and distributing it to a mobile terminal for execution, thereby realizing dynamic control of safety risks.

[0058] This embodiment also provides a computer device applicable to the control method of an integrated big data platform for security and fire protection based on artificial intelligence, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the control method of an integrated big data platform for security and fire protection based on artificial intelligence as proposed in the above embodiment.

[0059] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0060] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program performs the following steps: acquiring operational and historical data from an integrated security and fire safety big data platform, standardizing the operational data to generate a training dataset; training and constructing a deep convolutional neural network model based on the training dataset, extracting spatial correlation features through multi-layer convolution operations, and identifying key state information using an attention mechanism to generate a feature mapping matrix; inputting the feature mapping matrix into a constructed security assessment model to calculate a security risk index, and classifying risk levels according to the security risk index; generating an inspection task list based on the risk level classification results, and sending it to a mobile terminal for execution, thereby achieving dynamic control of security risks.

[0061] In summary, this invention addresses the issue of unifying heterogeneous data by standardizing the processing of operational and historical data; it improves the accuracy of risk identification by using a deep convolutional neural network model combined with an attention mechanism to extract deep features and identify key information from multidimensional data; it enables quantitative assessment and hierarchical management of risks through a feature mapping matrix-based security assessment model, providing a scientific basis for risk control; and it achieves closed-loop control of risk management by intelligently generating and distributing inspection tasks in real time. The overall solution overcomes the technical bottlenecks of traditional integrated security and fire protection platforms in data analysis, risk identification, and control, achieving intelligent and precise security risk management and significantly improving security management efficiency.

[0062] Example 2

[0063] Referring to Table 1, which is the second embodiment of the present invention, this embodiment provides a control method for an integrated big data platform for security and fire protection based on artificial intelligence. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0064] Specifically, an integrated big data platform system for fire safety and security was deployed in a large commercial complex covering an area of ​​50,000 square meters. 1200 IoT sensor nodes were installed within the building, including 500 temperature sensors (model: TS-2000, measurement range: -20℃ to 120℃, accuracy: ±0.1℃), 300 smoke sensors (model: SM-500, detection range: 0-500ppm, response time <10s), 200 hazardous gas sensors (model: GD-300, capable of detecting CO, NO2, etc., detection range: 0-100ppm), and 200 pressure sensors (model: PS-100, range: 0-10MPa). Simultaneously, 297 fire safety facilities within the building were precisely located, including 120 fire hydrants, 97 fire extinguisher boxes, 50 emergency lighting devices, and 30 fire escape routes.

[0065] Furthermore, 90 days of operational data were collected, with a sampling frequency of once every 5 minutes. Through data cleaning, approximately 3.2% of outliers and 1.8% of duplicates were removed, and approximately 2.5% of missing values ​​were supplemented using linear interpolation. For the distribution data of fire protection facilities, the WGS84 coordinate system was used for standardization, and the data was processed at a 1:100 scale. For safety inspection data, all timestamps were unified to the UTC+8 time zone, and the data format was standardized.

[0066] Furthermore, a deep convolutional neural network model was constructed, with the training and validation sets divided in an 80%:20% ratio. The deep convolutional neural network model was trained using mini-batch stochastic gradient descent with a batch size of 64, and the initial learning rate was set to 0.001, dynamically adjusted using cosine annealing. After achieving 95.8% accuracy on the validation set, practical application testing began.

[0067] Specifically, as shown in Table 1, the average risk index of different functional areas exhibits a reasonable distribution of differences. The average risk index of the equipment room area is the highest at 0.578, which is consistent with its characteristics of dense equipment and high power load. In contrast, the risk index of the commercial office area is the lowest at 0.312, reflecting the relatively stable safety status of this area. This distribution characteristic verifies that the risk assessment model can accurately capture the inherent risk characteristics of different areas.

[0068] Table 1. Experimental Data Table

[0069]

[0070] Furthermore, the system demonstrated excellent overall performance in terms of early warning accuracy, achieving over 95% accuracy in all areas, with the highest accuracy of 98.1% in the equipment room area. Particularly noteworthy is the system's generally low false alarm rate, around 1%, while the false alarm rate remained low as well, with the highest rate in the underground parking lot area at only 1.5%, significantly reducing unnecessary waste of human resources. The average response time was controlled within 2.5 seconds, with the equipment room area requiring only 1.9 seconds to complete risk assessment and early warning issuance. The accuracy of risk level determination remained above 96% across all areas, fully demonstrating the superiority of deep convolutional neural network models in feature extraction and risk level classification.

[0071] Furthermore, outstanding results have been achieved in terms of inspection task completion rate, with all areas reaching a completion rate of over 98.5%. This is attributed to the intelligent task allocation mechanism based on risk level and the precise navigation function of mobile terminals. Although the catering area has a relatively high risk index of 0.521, through the dynamic adjustment of the system, its early warning accuracy rate reached 97.2%, and the inspection task completion rate reached 99.3%, fully demonstrating the system's precise control capabilities in high-risk areas.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based integrated security and safety big data platform regulation method, characterized in that: Comprising, obtaining operation data and historical data of the safety and elimination integrated big data platform, and performing standardization processing on the operation data to generate a training data set; Based on the training data set, a deep convolutional neural network model is trained and constructed, spatial correlation features are extracted through multi-layer convolution operation, and key state information is identified combined with attention mechanism to generate a feature mapping matrix; The feature mapping matrix is input into the constructed safety evaluation model to calculate a safety risk index, and the safety risk index is used to divide risk levels; According to the risk level division result, a patrol task list is generated and sent to a mobile terminal for execution to realize dynamic regulation and control of safety risks; Based on historical data, a nonlinear mapping space is constructed using a radial basis kernel function, and kernel function parameters and penalty factors are selected through cross-validation method to construct a safety evaluation model; The weight coefficients of the spatial features and the time sequence features in the feature mapping matrix are calculated respectively, and a composite feature vector is constructed by using a weighted combination method; The composite feature vector is input into the safety evaluation model to calculate the distance from the sample point to the classification hyperplane to obtain the safety risk index; According to the statistical distribution of historical data, the classification threshold of the safety risk index is set, the safety risk index change trend of continuous multiple time points is calculated using a sliding time window, a risk warning level is established, and a warning strategy is set; The specific formula of the safety evaluation model is as follows: ； wherein, is the safety risk index at time t, is the support vector coefficient of the i-th sample, is the support vector sample in the historical data, is the improved radial basis kernel function, is the feature mapping matrix, is the bias term, is the regularization coefficient, is the weight coefficient of the j-th feature dimension, is the time series feature integral term of the j-th feature dimension, is the time window smoothing function, is the number of samples, is the extracted feature dimension.

2. The artificial intelligence-based security and safety integrated big data platform regulation method of claim 1, wherein: According to the risk level division result, a patrol task list is generated and sent to a mobile terminal for execution to realize dynamic regulation and control of safety risks, comprising: Based on the risk level division result, a patrol task list is generated, and a task priority sorting algorithm is used to allocate patrol resources, wherein the patrol task list includes a first-level patrol task list, a second-level patrol task list and a third-level patrol task list; The patrol task list includes a patrol path, a monitoring point position coordinate and a patrol frequency requirement; If the safety risk level is four, a third-level patrol task list is generated; if the safety risk level is three, a second-level patrol task list is generated; if the safety risk level is two, a first-level patrol task list is generated; The patrol task list is pushed to the mobile terminal device of the patrol personnel to carry out patrol work, and image data and environmental parameters of the patrol point are collected through the mobile terminal device, wherein the mobile terminal device plans an optimal patrol route based on the built-in electronic map navigation module; The image data and the environmental parameters are analyzed, and the analysis results are transmitted to the safety and elimination integrated big data platform to output patrol feedback data; The parameters of the risk evaluation model are updated through the patrol feedback data, and the risk level division result is dynamically corrected. 3.The AI-based security and safety integrated big data platform regulation method of claim 1, wherein: Further comprising, When the security risk index < the first threshold value and within a continuous time window < the preset threshold value, it is determined as a first-level risk. If the first threshold ≤ the safety risk index < the second threshold and the risk index at the current time k > the risk index at the previous time k-1 a secondary risk is determined, and a primary warning is triggered. If the second threshold value ≤ the safety risk index < the third threshold value and the safety risk index in the continuous time window If the safety risk index is maintained in this interval, it is determined as a third-level risk, a second-level early warning is triggered, the inspection route is adjusted, professional inspection personnel are added, and an emergency plan is developed. When the security risk index ≥ the third threshold or in a continuous time window > the preset threshold, it is determined as a fourth-level risk, a third-level early warning is triggered, an emergency response mechanism is started, and special control measures are implemented. 4.The AI-based security and safety integrated big data platform regulation method of claim 1, wherein: The generation method of the feature mapping matrix is, A deep convolutional neural network model is established according to a deep learning framework, and the training data set is divided into a training set and a validation set according to an expected proportion; Meanwhile, the deep convolutional neural network model is optimized and trained by using a small batch random gradient descent method, wherein the deep convolutional neural network model includes an input layer, multiple convolution layers, a pooling layer, a full connection layer and an output layer; The spatial correlation features are extracted through multi-layer convolution operation, the self-attention mechanism is introduced to calculate the weight coefficients between different channels, and the fusion attention weight is obtained by multiplying the channel attention weight and the spatial attention weight; The fusion attention weight and the original feature map are weighted and summed, and the weighted features are mapped and reduced through a fully connected layer to generate a feature mapping matrix, wherein the dimension of the feature mapping matrix is n*m, wherein n represents the number of samples, and m represents the dimension of the extracted features. 5.The AI-based security and safety integrated big data platform regulation method of claim 4, wherein: The generation method of the training data set is, A multi-source data acquisition channel is established, and operation data of the integrated safety and fire-fighting big data platform is collected through Internet of Things sensors, wherein the operation data includes monitoring data, fire-fighting facility distribution data and safety inspection data; The monitoring data is cleaned to eliminate abnormal values and repeated values, and the linear interpolation method is used to supplement missing values; At the same time, the fire-fighting facility distribution data is processed by coordinate unification and scale standardization, and the safety inspection data is processed by time series alignment and format unification; All the processed data is mapped to the [0, 1] interval by using the maximum and minimum value normalization method to generate a training data set.

6. An AI-based security and safety integrated big data platform regulation system based on the AI-based security and safety integrated big data platform regulation method of any one of claims 1-5. It comprises, An acquisition module is configured to acquire operation data and historical data of the integrated safety and fire-fighting big data platform, and to standardize the operation data to generate a training data set; A feature extraction module is configured to train and build a deep convolutional neural network model based on the training data set, to extract spatial correlation features through multi-layer convolution operation, and to identify key state information in combination with an attention mechanism to generate a feature mapping matrix; A safety evaluation and risk calculation module is configured to input the feature mapping matrix into a built safety evaluation model, to calculate a safety risk index, and to divide a risk level according to the safety risk index; An inspection task generation and delivery module is configured to generate an inspection task list according to the risk level division result, and to deliver it to a mobile terminal for execution to realize dynamic regulation and control of safety risks.

7. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the integrated safety and fire-fighting big data platform regulation and control method based on artificial intelligence according to any one of claims 1-5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the integrated safety and fire-fighting big data platform regulation and control method based on artificial intelligence according to any one of claims 1-5.

Images