Fire-fighting monitoring method, system, medium and equipment based on multi-sensor fusion

Abstract

The invention discloses a fire-fighting monitoring method and system based on multi-sensor fusion, a medium and equipment. The method comprises the following steps: acquiring equipment fire-fighting data acquired by multiple sensors in real time; performing preprocessing and abnormal data processing on the equipment fire-fighting data to obtain target fire-fighting data; performing fire prediction on the target fire-fighting data based on a space-time fusion model to obtain a fire probability; performing fire-fighting early warning judgment on the target fire-fighting data based on a risk level triggering condition; and generating a fire-fighting control instruction according to the fire probability and the fire-fighting early warning judgment result, and feeding back the fire-fighting control instruction to the equipment end. Therefore, the problem of hidden fire source recognition delay of an existing system is solved, and millisecond-level fire early warning and graded response closed-loop treatment are achieved.

Description

Technical Field

[0001] The present invention relates to the field of multi-dimensional perception computing technology, and in particular to a fire monitoring method, system, medium and equipment based on multi-sensor fusion. Background Art

[0002] Existing fire monitoring systems have significant delays in identifying hidden fire sources (such as smoldering cable insulation), as shown by:

[0003] 1. Insufficient ability to capture early signals:

[0004] In the early stage of smoldering, only trace gases are released (CO and VOCs concentrations <5ppm) and the temperature rises slowly (<3℃ / min). Traditional smoke sensors and infrared sensors cannot effectively detect them due to their low sensitivity or poor anti-interference. The temperature field distribution in hidden areas such as cable trenches is complex, and conventional point temperature measurement is prone to miss local hot spots.

[0005] 2. Data processing and decision-making delays:

[0006] The centralized architecture relies on cloud computing, and data transmission (sensor → cloud → execution end) takes >5 seconds, far exceeding the critical time window for smoldering spread (<3 seconds); the single threshold judgment method has a high false alarm rate (>20%), and multiple reviews are required, resulting in delayed response.

[0007] 3. Inefficient linkage control:

[0008] The existing system requires manual confirmation to trigger fire-fighting actions, with an average delay of 10 to 30 seconds; at the same time, it lacks a graded response mechanism and cannot implement differentiated treatments for different risk levels.

[0009] To address the above issues, it is necessary to solve the problem of delay in the existing system's recognition of hidden fire sources and achieve millisecond-level fire warning and graded response closed-loop disposal. Summary of the Invention

[0010] The present invention provides a fire monitoring method, system, medium and equipment based on multi-sensor fusion, which solves the problem of delayed identification of hidden fire sources in the prior art and realizes millisecond-level fire warning and graded response closed-loop disposal.

[0011] In a first aspect, a fire monitoring method based on multi-sensor fusion is provided, comprising the following steps:

[0012] Obtain equipment fire data collected by multiple sensors in real time;

[0013] Preprocessing the firefighting data of the equipment and performing abnormal data processing to obtain target firefighting data;

[0014] Performing fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability;

[0015] Performing fire warning judgment on the target fire data based on risk level trigger conditions;

[0016] Based on the fire probability and fire warning judgment results, a fire control instruction is generated and fed back to the device end.

[0017] According to the first aspect, in a first possible implementation of the first aspect, the step of “preprocessing the equipment fire data and processing abnormal data to obtain target fire data” specifically includes the following steps:

[0018] Normalizing the firefighting data of the equipment, and constructing a spatiotemporal matrix from the normalized data;

[0019] Based on the DS evidence theory algorithm, the spatiotemporal features extracted from the spatiotemporal matrix are converted into basic probability distributions corresponding to each sensor, and the basic probability distributions corresponding to each sensor are fused to obtain the confidence weight corresponding to each sensor;

[0020] When the confidence weight of any sensor is less than the preset threshold, the sensor is determined to be abnormal.

[0021] According to the first aspect, in a second possible implementation of the first aspect, the step of “performing a fire analysis on the target fire data based on the spatiotemporal fusion model and the loss function to obtain a fire probability” specifically includes the following steps:

[0022] Extracting local hotspot spatial features of the target firefighting data based on the spatial convolutional network in the spatiotemporal fusion model;

[0023] Based on the bidirectional LSTM network in the spatiotemporal fusion model, the dynamic temporal regularity characteristics in the local hotspot spatial characteristics are captured using time series data of a sliding window unit length;

[0024] Dynamically analyzing the coupling relationship in the dynamic temporal regularity features based on the cross-attention network in the spatiotemporal fusion model to generate a spatiotemporal coupling feature vector;

[0025] Fire prediction is performed on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain the fire probability.

[0026] According to the second possible implementation manner of the first aspect, in the third possible implementation manner of the first aspect, the step of “performing fire prediction on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain a fire probability” specifically includes the following steps:

[0027] The spatiotemporal coupling feature vector is input into the probability mapping layer, and the fire probability is generated through the Sigmoid activation function;

[0028] A dynamic temperature compensation mechanism is used to calibrate the fire probability output by the Sigmoid activation function.

[0029] According to the first aspect, in a fourth possible implementation manner of the first aspect, the step of “performing a fire warning judgment on the target fire data based on a risk level trigger condition” specifically includes the following steps:

[0030] When the CO concentration is detected to be greater than the first preset concentration threshold, or the temperature rise is greater than the first preset temperature rise threshold and the temperature rise lasts for a preset time, the fire warning is judged to be a pre-alarm;

[0031] When it is detected that the CO concentration is greater than the second preset concentration threshold, the VOCs concentration is greater than the third preset concentration threshold, and the temperature rise is greater than the second preset temperature rise threshold, the fire warning is judged to be a level one alarm;

[0032] When a flame signal is detected and the temperature is greater than the preset temperature threshold, the fire warning is judged to be a level 2 warning.

[0033] According to the fourth possible implementation manner of the first aspect, in the fifth possible implementation manner of the first aspect, the step of “generating a fire control instruction based on the fire probability and the fire warning judgment result and feeding it back to the device end” specifically includes the following steps:

[0034] When it is detected that the fire probability is less than the first preset probability, and / or the fire warning judgment result is the pre-alarm, a ventilation mode instruction is generated and fed back to the device end;

[0035] When it is detected that the fire probability is greater than or equal to the first preset probability and less than the second preset probability, and / or the fire warning judgment result is the first-level alarm, an inert gas filling and equipment load reduction operation mode control instruction is generated and fed back to the equipment end;

[0036] When it is detected that the fire probability is greater than or equal to the second preset probability, and / or the fire warning judgment result is the second-level alarm, a liquid fire extinguishing and equipment emergency stop mode control instruction is generated and fed back to the equipment end.

[0037] According to the first aspect, in a sixth possible implementation of the first aspect, after the step of “obtaining equipment fire data collected by multiple sensors in real time,” the following steps are specifically included:

[0038] When the data collected in real time by any target sensor is abnormal, the target sensor is controlled to switch to high-frequency sampling mode and redundant sensors within a preset range are activated;

[0039] When it is obtained that the data collected by the redundant sensor is abnormal and no flame signal is obtained, a device abnormality signal is sent to the device end or the mobile end.

[0040] Secondly, a fire monitoring system based on multi-sensor fusion is provided, comprising:

[0041] Data acquisition module, used to obtain equipment fire data collected by multiple sensors in real time;

[0042] A data processing module, in communication with the data acquisition module, for preprocessing the firefighting data of the equipment and processing abnormal data to obtain target firefighting data;

[0043] An analysis module, in communication with the data processing module, configured to perform fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability;

[0044] a judgment module, which is in communication with the data processing module and is used to perform fire warning judgment on the target fire data based on risk level trigger conditions; and

[0045] The instruction generation module is in communication with the judgment module and the analysis module, and is used to generate a fire control instruction based on the fire probability and the fire warning judgment result and feed it back to the device end.

[0046] In a third aspect, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the fire monitoring method based on multi-sensor fusion as described above is implemented.

[0047] In a fourth aspect, an electronic device is provided, comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein the processor implements the fire monitoring method based on multi-sensor fusion as described above when executing the computer program.

[0048] Compared with existing technologies, the present invention offers the following advantages: It achieves precise and rapid fire warnings through a dual fire warning assurance approach: The first approach utilizes risk-level triggering conditions to determine fire warnings for the target fire data; the second approach utilizes a spatiotemporal fusion model to predict fires and determine fire probabilities. Finally, fire control instructions are generated based on the fire probabilities and the fire warning judgment results. These two combined fire warning approaches address the single control strategy inherent in existing fire warnings. Edge computing, powered by a spatiotemporal fusion model, enables local decision-making, bypassing cloud transmission and addressing decision-making delays. Furthermore, these two combined fire warning approaches enable coordinated control, achieving millisecond-level fire warnings and multi-level response closed-loop processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of an embodiment of a fire monitoring method based on multi-sensor fusion according to the present invention;

[0050] Figure 2 This is a flow chart of another embodiment of a fire monitoring method based on multi-sensor fusion according to the present invention;

[0051] Figure 3 This is a flow chart of another embodiment of a fire monitoring method based on multi-sensor fusion according to the present invention;

[0052] Figure 4 This is a flow chart of another embodiment of a fire monitoring method based on multi-sensor fusion according to the present invention;

[0053] Figure 5 It is a structural schematic diagram of a fire monitoring system based on multi-sensor fusion of the present invention. DETAILED DESCRIPTION

[0054] Reference will now be made in detail to specific embodiments of the present invention, examples of which are illustrated in the accompanying drawings. Although the present invention will be described in conjunction with specific embodiments, it will be understood that the present invention is not intended to be limited to those embodiments. On the contrary, it is intended to cover variations, modifications, and equivalents within the spirit and scope of the present invention as defined by the appended claims. It should be noted that the method steps described herein can be implemented by any functional block or functional arrangement, and any functional block or functional arrangement can be implemented as a physical entity or a logical entity, or a combination of the two.

[0055] In order to enable those skilled in the art to better understand the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Note: The following example is only a specific example and is not intended to limit the embodiments of the present invention to the following specific steps, values, conditions, data, sequence, etc. Those skilled in the art can apply the concepts of the present invention to construct more embodiments not described in this specification by reading this specification.

[0057] See also Figure 1 As shown, an embodiment of the present invention provides a fire monitoring method based on multi-sensor fusion, comprising the following steps:

[0058] S100, obtains equipment fire data collected in real time by multiple sensors;

[0059] S200, pre-processing the equipment fire data and abnormal data processing to obtain target fire data;

[0060] S300, performing fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability;

[0061] S400, performing fire warning judgment on the target fire data based on the risk level trigger condition;

[0062] S500: Generate a fire control instruction based on the fire probability and fire warning judgment result and feed it back to the device end.

[0063] Specifically, in this embodiment, the multi-sensor used in the present invention is a multi-dimensional perception layer design, which is as follows:

[0064] 1. High-precision gas detection module:

[0065] Adopting TDLAS (tunable diode laser absorption spectroscopy) technology to detect CO and VOCs concentrations with a sensitivity of 0.1ppm (50 times higher than traditional electrochemical sensors);

[0066] Anti-interference design: wavelength modulation is used to eliminate the interference of oil mist and dust on the optical path, and the false detection rate is <0.1%.

[0067] 2. Distributed optical fiber temperature measurement network:

[0068] Armored temperature measurement optical fiber is laid along the cable trench, with a spatial resolution of 0.1m and a sampling rate of 10Hz;

[0069] Deployment strategy: serpentine winding + longitudinal layering to achieve three-dimensional reconstruction of the cable surface and internal temperature field.

[0070] 3. Auxiliary perception unit:

[0071] Ultraviolet / infrared dual-mode flame sensor: covers 200-280nm (ultraviolet) and 8-14μm (infrared) bands, identifying the critical state of smoldering to open flame;

[0072] Vibration sensor: monitors abnormal cable vibration (frequency > 500Hz), which is associated with the risk of mechanical friction overheating.

[0073] Therefore, the multi-dimensional perception design of the sensor can make up for the problem of insufficient ability to capture early hidden fire source signals.

[0074] The present invention also achieves precise and rapid fire warnings through a dual fire warning assurance approach. Specifically, the first approach uses risk-level triggering conditions to determine fire warnings for the target fire data. The second approach uses a spatiotemporal fusion model to predict fires and determine fire probabilities. Finally, fire control instructions are generated based on the fire probabilities and the fire warning judgment results. These two combined fire warning approaches address the single control strategy issue of existing fire warnings. Edge computing, powered by a spatiotemporal fusion model, enables local decision-making, bypassing cloud transmission and addressing decision-making delays. Furthermore, these two combined fire warning approaches enable coordinated control, achieving millisecond-level fire warnings and multi-level response closed-loop processing.

[0075] Preferably, in another embodiment of the present application, the step of "S200, pre-processing the equipment fire data and performing abnormal data processing to obtain target fire data" specifically includes the following steps:

[0076] S210, normalizing the equipment fire data and constructing a spatiotemporal matrix from the normalized data;

[0077] S220, based on the DS evidence theory algorithm, converting the spatiotemporal features extracted from the spatiotemporal matrix into basic probability distributions corresponding to each sensor, and fusing the basic probability distributions corresponding to each sensor to obtain a confidence weight corresponding to each sensor;

[0078] S230: When the confidence weight of any sensor is less than a preset threshold, the sensor is determined to be abnormal.

[0079] Specifically, in this embodiment, the data of multiple sensors in time series are integrated into a matrix M T×S , where T is the number of time points, S is the number of sensors, and element m t,s represents the normalized data of sensor s at time t.

[0080] Spatiotemporal feature extraction includes extracting spatiotemporal feature evidence and spatial feature evidence.

[0081] Time dimension: Calculate the mean, variance, and trend (such as the slope of the linear fit within the sliding window) of each sensor.

[0082] Spatial dimension: Calculate the similarity between different sensors at the same time point (such as cosine similarity, Euclidean distance).

[0083] The spatiotemporal features extracted from the spatiotemporal matrix are then converted into a basic probability allocation (BPA) for each sensor based on the DS evidence theory algorithm. The BPA is typically determined based on the characteristics of the data, such as whether it deviates from the normal range or whether it is consistent with data from other sensors. For example, if a sensor's data differs significantly from others, it may be assigned a lower BPA, while if it differs significantly from others, it may be assigned a higher BPA.

[0084] Taking sensor s as an example, its basic probability distribution formula is as follows:

[0085] m s (H 正常 )=1 / (1+α·Var(s)+β·Distance(s,others)).

[0086] Among them, α and β are weight parameters, which need to be adjusted according to the scenario.

[0087] After that, the BPA from the spatiotemporal features is fused. If the two pieces of evidence conflict (e.g. the temporal feature is considered normal, while the spatial feature is considered abnormal), the conflict factor K is calculated:

[0088]

[0089] Fusion BPA:

[0090]

[0091] Assign weights based on the fused BPA:

[0092]

[0093] Set a threshold θ. If Ws<θ, mark sensor s as abnormal.

[0094] See also Figure 3 As shown, preferably, in another embodiment of the present application, the step of "S300, performing fire analysis on the target fire data based on the spatiotemporal fusion model and the loss function to obtain the fire probability" specifically includes the following steps:

[0095] S310, extracting local hotspot spatial features from the target firefighting data based on a spatial convolutional network in a spatiotemporal fusion model;

[0096] S320, based on the bidirectional LSTM network in the spatiotemporal fusion model, and using the time series data of the sliding window unit length to capture the dynamic temporal regularity characteristics in the local hotspot spatial characteristics;

[0097] S330, dynamically analyzing the coupling relationship in the dynamic temporal regularity features based on the cross-attention network in the spatiotemporal fusion model to generate a spatiotemporal coupling feature vector;

[0098] S340 , performing fire prediction on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain a fire probability.

[0099] Specifically, in this embodiment,

[0100] Spatial convolution module: A two-dimensional convolution layer with a convolution kernel size of 3×3 is used to extract the spatial features of local hotspots in the sensor network. The output of the spatial convolution (local hotspot distribution) is used as the input of a bidirectional LSTM network, enabling the bidirectional LSTM network to retain spatial context information when analyzing time series trends.

[0101] Time series feature extraction module: Through the bidirectional LSTM network, the temperature rise rate and periodic trend are captured using time series data with a sliding window length of T; the formula is as follows:

[0102]

[0103] Cross-modal attention module: Dynamically correlates the coupling relationship between gas concentration and temperature changes through the cross-attention mechanism to generate spatiotemporal coupling feature vectors, which can solve the problem of modal heterogeneity.

[0104] The step of “S340, performing fire prediction on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain a fire probability” specifically includes the following steps:

[0105] The spatiotemporal coupling feature vector is input into the probability mapping layer, and the fire probability is generated through the Sigmoid activation function;

[0106] A dynamic temperature compensation mechanism is used to calibrate the fire probability output by the Sigmoid activation function.

[0107] Specifically, in this embodiment, the spatiotemporal coupling feature vector F∈R D Input to the probability mapping layer and generate the fire probability value fire∈[0,1] through the Sigmoid activation function.

[0108] The formula for the basic Sigmoid probability output is as follows:

[0109]

[0110] Where W∈R D ,b∈R is a trainable parameter.

[0111] Then the dynamic temperature compensation mechanism ΔT(t) is used to calibrate the Sigmoid function output.

[0112]

[0113] Where,

[0114] T env (i) is the ambient temperature at time point i (from the mean value of the temperature sensor network);

[0115] T ref is the base temperature (e.g. annual average temperature 20°C);

[0116] γ is the static compensation coefficient (default 0.05, optimized through training);

[0117] η is the dynamic change rate weight (default 0.01, controls the sensitivity of the temperature rise rate);

[0118] τ is the sliding window length;

[0119] Dt / dT is the current temperature rise rate (calculated by linear regression of the last 10 minutes of data).

[0120] The fire probability output after compensation is as follows:

[0121]

[0122] Wherein, temperature parameter k is a dynamic constraint term, which is optimized by gradient descent to satisfy k∈[1.0,3.0].

[0123] See also Figure 4 As shown, preferably, in another embodiment of the present application, the step of "S400, performing fire warning judgment on the target fire data based on the risk level trigger condition" specifically includes the following steps:

[0124] S410, when it is detected that the CO concentration is greater than a first preset concentration threshold, or the temperature rise is greater than a first preset temperature rise threshold and the temperature rise lasts for a preset time, the fire warning is determined to be a pre-alarm;

[0125] S420, when it is detected that the CO concentration is greater than the second preset concentration threshold, the VOCs concentration is greater than the third preset concentration threshold, and the temperature rise is greater than the second preset temperature rise threshold, the fire warning is determined to be a level one alarm;

[0126] S430: When a flame signal is detected and the temperature is greater than a preset temperature threshold, the fire warning is determined to be a level 2 warning.

[0127] Preferably, in another embodiment of the present application, the step of "S500, generating a fire control instruction and feeding it back to the device end based on the fire probability and the fire warning judgment result" specifically includes the following steps:

[0128] When it is detected that the fire probability is less than the first preset probability, and / or the fire warning judgment result is the pre-alarm, a ventilation mode instruction is generated and fed back to the device end;

[0129] When it is detected that the fire probability is greater than or equal to the first preset probability and less than the second preset probability, and / or the fire warning judgment result is the first-level alarm, an inert gas filling and equipment load reduction operation mode control instruction is generated and fed back to the equipment end;

[0130] When it is detected that the fire probability is greater than or equal to the second preset probability, and / or the fire warning judgment result is the second-level alarm, a liquid fire extinguishing and equipment emergency stop mode control instruction is generated and fed back to the equipment end.

[0131] Specifically, in this embodiment, the multi-level response trigger mechanism is as follows:

[0132]

[0133] Preferably, in another embodiment of the present application, after the step of “S100, obtaining equipment fire data collected by multiple sensors in real time”, the following steps are specifically included:

[0134] When the data collected in real time by any target sensor is abnormal, the target sensor is controlled to switch to high-frequency sampling mode and redundant sensors within a preset range are activated;

[0135] When it is obtained that the data collected by the redundant sensor is abnormal and no flame signal is obtained, a device abnormality signal is sent to the device end or the mobile end.

[0136] Specifically, in this embodiment,

[0137] Normal mode: The sensor samples at a low frequency of 1Hz and the power consumption is less than 10W;

[0138] Trigger mode: When any sensor detects an anomaly (e.g., CO > 2ppm or temperature rise rate > 1°C / s), it automatically switches to 10Hz high-frequency sampling and simultaneously activates redundant sensor groups within 3 meters of the adjacent area, forming a cross-validation network. Data is simultaneously recorded, and cameras are activated to capture the scene. The system analyzes the presence of personnel and fire sources. If so, a voice reminder is issued to warn of fire hazards, but the wind turbine emergency management center is not alerted. If only personnel are detected without a fire source, the system alerts personnel to the wind turbine equipment anomaly and warns of safety precautions.

[0139] See also Figure 5 As shown, an embodiment of the present invention provides a fire monitoring system 100 based on multi-sensor fusion, including:

[0140] The data acquisition module 110 is used to obtain equipment fire data collected by multiple sensors in real time;

[0141] The data processing module 120 is in communication with the data acquisition module 110 and is used to pre-process the firefighting data of the equipment and process abnormal data to obtain target firefighting data;

[0142] An analysis module 130 is in communication with the data processing module 120 and is configured to perform fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability;

[0143] The judgment module 140 is in communication with the data processing module 120 and is configured to perform fire warning judgment on the target fire data based on a risk level trigger condition; and

[0144] The instruction generation module 150 is in communication with the judgment module 140 and the analysis module 130 and is used to generate a fire control instruction based on the fire probability and the fire warning judgment result and feed it back to the device end.

[0145] The data processing module 120 is used to normalize the fire protection data of the equipment and construct a space-time matrix with the normalized data; based on the DS evidence theory algorithm, the space-time features extracted from the space-time matrix are converted into basic probability distributions corresponding to each sensor, and the basic probability distributions corresponding to each sensor are fused to obtain the confidence weight corresponding to each sensor; when the confidence weight of any sensor is less than a preset threshold, the sensor is determined to be abnormal.

[0146] The analysis module 130 is used to extract local hotspot spatial features of the target fire data based on the spatial convolutional network in the spatiotemporal fusion model; capture the dynamic temporal regularity features in the local hotspot spatial features based on the bidirectional LSTM network in the spatiotemporal fusion model and the time series data of the unit length of the sliding window; dynamically analyze the coupling relationship in the dynamic temporal regularity features based on the cross-attention network in the spatiotemporal fusion model to generate a spatiotemporal coupling feature vector; and perform fire prediction on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain a fire probability.

[0147] The judgment module 140 is configured to determine that the fire warning is a pre-alarm when it is detected that the CO concentration is greater than a first preset concentration threshold, or the temperature rise is greater than the first preset temperature rise threshold and the temperature rise continues for a preset time; determine that the fire warning is a first-level alarm when it is detected that the CO concentration is greater than a second preset concentration threshold, the VOCs concentration is greater than a third preset concentration threshold, and the temperature rise is greater than the second preset temperature rise threshold; and determine that the fire warning is a second-level alarm when a flame signal is detected and the temperature is greater than a preset temperature threshold.

[0148] The instruction generation module 150 is used to generate a ventilation mode instruction and feed it back to the device end when it is detected that the fire probability is less than the first preset probability, and / or the fire warning judgment result is the pre-alarm; when it is detected that the fire probability is greater than or equal to the first preset probability and less than the second preset probability, and / or the fire warning judgment result is the first-level alarm, generate an inert gas filling and equipment load reduction operation mode control instruction and feed it back to the device end; when it is detected that the fire probability is greater than or equal to the second preset probability, and / or the fire warning judgment result is the second-level alarm, generate a liquid fire extinguishing and equipment emergency stop mode control instruction and feed it back to the device end.

[0149] In summary, the beneficial effects of the present invention are as follows:

[0150] 1. Significantly reduced latency:

[0151] Smoldering detection delay ≤ 0.5 seconds (traditional system ≥ 8 seconds), and response speed increased by 16 times;

[0152] Edge computing enables local decision-making, bypassing cloud transmission links and saving more than 3 seconds of delay.

[0153] 2. Improved detection accuracy:

[0154] The accuracy rate of concealed fire source identification is ≥98.5% (traditional solutions ≤70%);

[0155] The false alarm rate is less than 1% (the traditional solution is ≥20%), reducing invalid downtime losses.

[0156] 3. Energy efficiency and reliability optimization:

[0157] The dynamic sampling strategy reduces system power consumption by 60% (from 120W to 45W).

[0158] Specifically, this embodiment corresponds one-to-one to the above method embodiment, and the functions of each module have been described in detail in the corresponding method embodiment, so they will not be repeated here.

[0159] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, all or part of the method steps of the above method are implemented.

[0160] The present invention implements all or part of the process in the above method, and can also be completed by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0161] Based on the same inventive concept, an embodiment of the present application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program running on the processor, and when the processor executes the computer program, all or part of the method steps in the above method are implemented.

[0162] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of a computer device, connecting all parts of the entire computer device using various interfaces and lines.

[0163] The memory can be used to store computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, video data, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0164] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, servers, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage) containing computer-usable program code.

[0165] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), servers, and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0166] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0167] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0168] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A fire monitoring method based on multi-sensor fusion, characterized in that: The following steps are involved: Obtain equipment fire data collected by multiple sensors in real time; Preprocessing the firefighting data of the equipment and performing abnormal data processing to obtain target firefighting data; Performing fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability; Performing fire warning judgment on the target fire data based on risk level trigger conditions; Based on the fire probability and fire warning judgment results, a fire control instruction is generated and fed back to the device end.

2. The fire monitoring method based on multi-sensor fusion according to claim 1, characterized in that: The step of "preprocessing the equipment fire data and processing abnormal data to obtain target fire data" specifically includes the following steps: Normalizing the firefighting data of the equipment, and constructing a spatiotemporal matrix from the normalized data; Based on the DS evidence theory algorithm, the spatiotemporal features extracted from the spatiotemporal matrix are converted into basic probability distributions corresponding to each sensor, and the basic probability distributions corresponding to each sensor are fused to obtain the confidence weight corresponding to each sensor; When the confidence weight of any sensor is less than a preset threshold, the sensor is determined to be abnormal.

3. The fire monitoring method based on multi-sensor fusion according to claim 1, characterized in that: The step of "performing fire analysis on the target fire data based on the spatiotemporal fusion model and the loss function to obtain the fire probability" specifically includes the following steps: Extracting local hotspot spatial features of the target firefighting data based on the spatial convolutional network in the spatiotemporal fusion model; Based on the bidirectional LSTM network in the spatiotemporal fusion model, the dynamic temporal regularity characteristics in the local hotspot spatial characteristics are captured using time series data of the unit length of the sliding window; Dynamically analyzing the coupling relationship in the dynamic temporal regularity features based on the cross-attention network in the spatiotemporal fusion model to generate a spatiotemporal coupling feature vector; Fire prediction is performed on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain the fire probability.

4. The fire monitoring method based on multi-sensor fusion according to claim 3, characterized in that: The step of "predicting fire on the spatiotemporal coupling feature vector based on the Sigmoid activation function in the spatiotemporal fusion model to obtain the fire probability" specifically includes the following steps: The spatiotemporal coupling feature vector is input into the probability mapping layer, and the fire probability is generated through the Sigmoid activation function; A dynamic temperature compensation mechanism is used to calibrate the fire probability output by the Sigmoid activation function.

5. The fire monitoring method based on multi-sensor fusion according to claim 1, characterized in that: The step of "performing fire warning judgment on the target fire data based on the risk level trigger condition" specifically includes the following steps: When the CO concentration is detected to be greater than the first preset concentration threshold, or the temperature rise is greater than the first preset temperature rise threshold and the temperature rise lasts for a preset time, the fire warning is judged to be a pre-alarm; When it is detected that the CO concentration is greater than the second preset concentration threshold, the VOCs concentration is greater than the third preset concentration threshold, and the temperature rise is greater than the second preset temperature rise threshold, the fire warning is judged to be a level one alarm; When a flame signal is detected and the temperature is greater than the preset temperature threshold, the fire warning is judged to be a level 2 warning.

6. The equipment fire monitoring method based on multi-sensor fusion according to claim 5, characterized in that: The step of "generating a fire control instruction and feeding it back to the device end based on the fire probability and the fire warning judgment result" specifically includes the following steps: When it is detected that the fire probability is less than the first preset probability, and / or the fire warning judgment result is the pre-alarm, a ventilation mode instruction is generated and fed back to the device end; When it is detected that the fire probability is greater than or equal to the first preset probability and less than the second preset probability, and / or the fire warning judgment result is the first-level alarm, an inert gas filling and equipment load reduction operation mode control instruction is generated and fed back to the equipment end; When it is detected that the fire probability is greater than or equal to the second preset probability, and / or the fire warning judgment result is the second-level alarm, a liquid fire extinguishing and equipment emergency stop mode control instruction is generated and fed back to the equipment end.

7. The fire monitoring method based on multi-sensor fusion according to claim 1, characterized in that: After the step of "obtaining equipment fire data collected by multiple sensors in real time", the following steps are specifically included: When the data collected in real time by any target sensor is abnormal, the target sensor is controlled to switch to high-frequency sampling mode and redundant sensors within a preset range are activated; When it is obtained that the data collected by the redundant sensor is abnormal and no flame signal is obtained, a device abnormality signal is sent to the device end or the mobile end.

8. A fire monitoring system based on multi-sensor fusion, characterized in that: include: Data acquisition module, used to obtain equipment fire data collected by multiple sensors in real time; A data processing module, in communication with the data acquisition module, for preprocessing the firefighting data of the equipment and processing abnormal data to obtain target firefighting data; an analysis module, communicatively connected to the data processing module, for performing fire prediction on the target fire data based on a spatiotemporal fusion model to obtain a fire probability; a judgment module, in communication with the data processing module, for performing fire warning judgment on the target fire data based on a risk level trigger condition; as well as, The instruction generation module is in communication with the judgment module and the analysis module, and is used to generate a fire control instruction based on the fire probability and the fire warning judgment result and feed it back to the device end.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fire monitoring method based on multi-sensor fusion according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein: When the processor runs the computer program, the fire monitoring method based on multi-sensor fusion according to any one of claims 1 to 7 is implemented.

Images