An intelligent fire early warning method and system, an electronic device, and a storage medium

Abstract

The application provides a kind of intelligent fire early warning method, system, electronic equipment and storage medium, belong to fire safety and intelligent monitoring field;The method comprises: collecting the special parameter for fire detection in the key risk area of aviation refueling station;The special parameter is preprocessed to obtain preprocessed special parameter;Create improved SRU network for identifying fire state and for simulation training;The preprocessed special parameter is input into the improved SRU network to identify the fire state probability of the corresponding detection area;The fire warning level of the corresponding detection area is obtained by fusing the fire state probability and decision factor through the inaccurate reasoning theory.It can realize the processing pressure of multiple single sensor data fusion by the present application, and the improved SRU network based on adaptive excitation function is combined to accurately, quickly and stably identify the probability of smoldering fire, open fire and interference fire, and the severity of the fire can be responded in time by means of the decision factor through the identification probability.

Description

Technical Field

[0001] This invention belongs to the technical field of fire safety and intelligent monitoring, specifically relating to an intelligent fire early warning method, system, electronic device and storage medium. Background Technology

[0002] Aviation refueling stations are places where flammable and explosive oil products such as aviation fuel are stored and processed. Therefore, effective fire safety management at aviation refueling stations is crucial. Fire alarm systems have revealed problems such as slow response times, relatively high false alarm and false negative rates. Fire detection technology can solve these problems compared to fire alarm systems. Therefore, timely and accurate fire warnings to minimize fire losses have become the primary task of aviation refueling stations.

[0003] Current fire detection technologies widely use detectors based on relatively simple sensors such as smoke, combustible gas, and temperature. These detectors can only monitor a single fire parameter, often require proximity to the fire source to function, and require the fire to reach a certain level exceeding a threshold set by the sensor before triggering an alarm. However, fire is a complex physicochemical process, and different environments and combustible materials produce different products. Detecting only a single fire parameter makes it difficult to accurately determine all fire situations, failing to provide effective fire early warning and allowing a longer window for the fire to spread.

[0004] Therefore, how to fuse data from different types of single sensors, process the fused sensor data through intelligent algorithms to make accurate and stable judgment results, and predict fires in advance and issue early warning instructions based on the judgment results is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] To address at least one of the aforementioned technical problems, this invention provides an intelligent fire early warning method, system, electronic device, and storage medium, which can significantly shorten the response time to fire signals, improve the accuracy and timeliness of fire early warning, and thus enable the detection and early warning of early stages of fires.

[0006] In a first aspect, the invention provides an intelligent fire early warning method, comprising: Collect specific parameters for fire detection within the critical risk area of ​​the aviation refueling station, including temperature, smoke concentration, and CO concentration. The special parameters are preprocessed to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; An improved SRU network was created for simulation training and for identifying fire states, including smoldering fire, open flame, and interfering fire. The preprocessing parameters are input into the improved SRU network to identify the probability of fire status in the corresponding detection area; The fire warning level for the corresponding detection area is obtained by integrating the fire state probability and decision factors using inaccurate reasoning theory.

[0007] Preferably, the step of collecting specific parameters within the key risk area of ​​the aviation refueling station for fire detection specifically includes: Analyze the characteristics and development patterns of fire signals at aviation refueling stations; Temperature, smoke concentration, and CO concentration were selected as specific parameters for fire detection. A multi-source sensor network covering key risk areas of aviation refueling stations is constructed to collect the specific parameters.

[0008] Preferably, the step of preprocessing the specific parameters to obtain preprocessed specific parameters specifically includes: The specific parameters are processed using a local decision-making method based on a variable threshold. The specific parameters after local decision-making are processed using upper and lower limit amplitude values. The preprocessed special parameters are obtained by normalizing the special parameters using a linear function and then processing them using upper and lower limit amplitude methods.

[0009] Preferably, the step of creating an improved SRU network for simulation training and for identifying fire conditions specifically includes: An experimental platform based on AC power line carrier communication was built to collect simulation data of smoldering fire, open flame, and interference fire. Construct an SRU network for identifying fire status; The simulation data is input and the SRU network is trained to obtain an improved SRU network based on an adaptive activation function.

[0010] Preferably, the step of inputting the preprocessed specific parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area specifically includes: The preprocessing parameters are divided into data groups according to the same detection area and the same time. Each data group includes temperature, smoke concentration, and CO concentration. Set the relevant configuration parameters for fire status recognition iteration, wherein the configuration parameters include the number of neurons, the number of network layers, the learning class, and the loss function; Based on the configuration parameters, several data groups are input into the improved SRU network for iterative calculation to identify the probability of fire status in the corresponding detection area.

[0011] Preferably, the step of deriving the fire warning level for the corresponding detection area by fusing the fire state probability and decision factors through imprecise reasoning theory specifically includes: The protection level of aviation refueling stations and the duration of abnormal signals were selected as decision factors. The decision factors are fuzzified using membership functions; The early warning set output is obtained by fusing the fire state probability and the fuzzified decision factors using a fuzzy logic algorithm. The output of the warning set is defuzzified using the centroid method to determine the fire warning level of the corresponding detection area.

[0012] Preferably, after the step of fusing the fire state probability and decision factors through imprecise reasoning theory to derive the fire warning level for the corresponding detection area, the method further includes: Based on the obtained fire warning level, emergency strategy programs are retrieved from the pre-made fire emergency strategy library to control and execute the corresponding fire emergency measures.

[0013] Secondly, an intelligent fire early warning system includes: The data acquisition module is used to collect specific parameters in key risk areas of aviation refueling stations for fire detection, including temperature, smoke concentration, and CO concentration. The preprocessing module is used to preprocess the special parameters to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; A module is created to create an improved SRU network for simulation training and for identifying fire states, including smoldering fire, open flame, and interfering fire. The identification module is used to input the preprocessed special parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area; The decision module is used to determine the fire warning level for the corresponding detection area by fusing the fire state probability and decision factors through imprecise reasoning theory.

[0014] Preferably, the acquisition module specifically includes: The analysis unit is used to analyze the characteristics and development patterns of fire signals at aviation refueling stations. The selection unit is used to select temperature, smoke concentration, and CO concentration as specific parameters for fire detection. A construction unit is used to construct a multi-source sensor network covering key risk areas of aviation refueling stations to collect the specific parameters.

[0015] Preferably, the preprocessing module specifically includes: A threshold valve unit is used to process the specific parameter based on a variable threshold method for local decision-making. A limiting unit is used to process the localized special parameters using upper and lower limiting values. The normalization unit is used to obtain the preprocessed special parameters by normalizing the special parameters after the upper and lower limit amplitude method through linear function normalization.

[0016] Preferably, the creation module specifically includes: The building unit is used to build an experimental platform based on AC power line carrier communication to collect simulation data of smoldering fire, open flame, and interference fire. Construction unit, used to build an SRU network for identifying fire status; The training unit is used to input the simulation data and train the SRU network to obtain an improved SRU network based on an adaptive activation function.

[0017] Preferably, the identification module specifically includes: The division unit is used to divide the preprocessed parameters into data groups according to the same detection area and the same time. Each data group includes temperature, smoke concentration, and CO concentration. The setting unit is used to set the relevant configuration parameters for the fire status recognition iteration, wherein the configuration parameters include the number of neurons, the number of network layers, the learning class, and the loss function. An iterative unit is used to input several of the data groups into the improved SRU network based on the configuration parameters to perform iterative calculations and identify the probability of fire status in the corresponding detection area.

[0018] Preferably, the decision-making module specifically includes: The selection unit is used to select the protection level of the aviation refueling station and the duration of abnormal signals as decision factors; Fuzzy units are used to fuzzify the decision factors using membership functions; The fusion unit is used to fuse the fire state probability and the fuzzified decision factors through a fuzzy logic algorithm to obtain the early warning set output. The deblurring unit is used to deblur the output of the warning set using the centroid method to determine the fire warning level of the corresponding detection area.

[0019] Preferably, the intelligent fire early warning system further includes: The emergency module is used to retrieve emergency strategy programs from a pre-made fire emergency strategy library based on the acquired fire warning level, so as to control the execution of corresponding fire emergency measures.

[0020] Thirdly, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent fire early warning method as described in the first aspect.

[0021] Fourthly, this application provides a storable medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent fire early warning method as described in the first aspect.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Analyzing the occurrence mechanism and development process of fire, fire events are complex and unpredictable. Selecting temperature, smoke concentration and CO concentration as specific parameters for fire monitoring can, firstly, more accurately reflect the fire situation, and secondly, facilitate the perception of specific parameters by multiple sensors. Through sensor complementarity, the system can obtain more comprehensive, accurate and reliable information, thereby facilitating precise real-time monitoring of different stages of the fire.

[0023] 2. Because the sensors used for fire detection have local decision-making capabilities, they can perform preprocessing on their own data, such as uploading simple values, setting upper and lower limits, and normalization. Sending the preprocessed data to the data fusion process reduces the data processing pressure and improves computational efficiency.

[0024] 3. Based on the fire development law, a more realistic fire process is simulated to obtain simulation data for training the model. By inputting simulation data into training and avoiding the risk of gradient explosion, an improved SRU network based on an adaptive excitation function is formed, which realizes the rapid identification of feature quantities containing time series information.

[0025] 4. By combining the advantages of the improved SRU network in terms of recognition time, and by using the adaptive activation function to avoid gradient vanishing during iterative calculations, it is possible to quickly and stably identify the probability of smoldering fires, open flames, and interfering fires, thereby improving the accuracy of fire early warning decisions.

[0026] 5. By combining the identified probabilities of smoldering fire, open flame, and interference fire with decision factors such as the protection level of the aviation refueling station and the duration of abnormal signals, fuzzy reasoning is used at the decision level for information fusion, and the correct alarm decision is finally output; this solves the problem that using only the current probabilities of open flame and smoldering fire to characterize the current fire status of the computer room cannot determine the specific severity of the fire based solely on these probabilities. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art 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.

[0028] Figure 1 A flowchart of the intelligent fire early warning method provided in Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of an intelligent fire early warning system corresponding to the method in Embodiment 1, provided in Embodiment 2 of the present invention; Figure 3 A flowchart of the intelligent fire early warning method provided in Embodiment 3 of the present invention; Figure 4 This is a structural block diagram of an intelligent fire early warning system corresponding to the method in Embodiment 3, provided in Embodiment 4 of the present invention; Figure 5 This is a schematic diagram of the hardware structure of the computer provided in Embodiment 5 of the present invention.

[0029] Explanation of reference numerals in the attached figures: 10-Acquisition module, 11-Analysis unit, 12-Selection unit, 13-Construction unit.

[0030] 20 - Preprocessing module, 21 - Threshold valve unit, 22 - Limiting unit, 23 - Normalization unit.

[0031] 30 - Create a module, 31 - Build a unit, 32 - Construct a unit, 33 - Train a unit.

[0032] 40 - Identification module, 41 - Division unit, 42 - Setting unit, 43 - Iteration unit.

[0033] 50 - Decision module, 51 - Selection unit, 52 - Fusion unit, 53 - Fusion unit, 54 - Defuzzification unit.

[0034] 60 - Emergency Module.

[0035] 70-Bus, 71-Processor, 72-Memory, 73-Communication interface. Detailed Implementation

[0036] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0037] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., may be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring aspects of this disclosure.

[0038] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0039] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily need to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the actual situation.

[0040] Fires at aviation refueling stations differ from other fires primarily due to improper operation during refueling, unloading, and metering processes, resulting in leaks from equipment and pipeline ruptures, as well as open flames. They are characterized by high concealment in their early stages, rapid spread, and significant destructive power. Current fire detection technologies, relying on single sensors, suffer from high false alarm rates and struggle to comprehensively assess the complex early signs in different risk areas such as refueling areas, unloading ports, and storage tank areas. Furthermore, they fail to combine the fire status with specific site characteristics to formulate appropriate fire alarm decisions. Example 1

[0041] Specifically, Figure 1 The diagram shown is a flowchart of the intelligent fire early warning method provided in this embodiment.

[0042] like Figure 1 As shown, the intelligent fire early warning method in this embodiment includes the following steps: S101, collects special parameters within the critical risk area of ​​aviation refueling stations for fire detection.

[0043] Specifically, fire events exhibit complexity and unpredictability. Key risk areas at aviation refueling stations include refueling areas, unloading areas, storage tank areas, and convenience store entrances / exits. Analyzing the mechanisms and development of fires reveals that, given the diversity of fire types and the variable nature of their characteristic parameters, relying on a single monitoring system is insufficient to accurately determine all fire situations. Furthermore, a single fire detector cannot fully integrate various fire parameters in the environment, leading to a high false alarm rate. This embodiment proposes an intelligent fire detection technology based on multi-sensor information fusion. Temperature, smoke concentration, and CO concentration are selected as specific parameters for fire monitoring. By integrating multiple types of sensors (e.g., temperature, smoke concentration, CO concentration, etc.) into one unit, multiple specific parameters generated during a fire can be monitored, effectively compensating for the shortcomings of traditional single-detection systems. This improves the accuracy, timeliness, and stability of fire detectors. Firstly, it provides a more realistic reflection of the fire situation; secondly, it facilitates multi-sensor perception of specific parameters. Through sensor complementarity, the system can obtain more comprehensive, accurate, and reliable information, thus enabling precise real-time monitoring of different stages of a fire.

[0044] Furthermore, the specific steps of step S101 include: S1011, Analyze the characteristics and development patterns of fire signals at aviation refueling stations; Specifically, based on temperature changes, the fire process can be divided into three stages: the initial stage, the development stage, and the extinguishing stage. The initial stage of a fire includes smoldering and spreading. In the smoldering stage, the combustible material begins to heat up, releasing a small amount of heat; the temperature is slightly higher than the surrounding area, but no open flame appears, accompanied by smoke and gas production. In the spreading stage, the combustible material continues to heat up, exceeding its ignition point, causing a rapid temperature rise, producing large amounts of smoke and gas, and eventually flames. Therefore, in the initial stage of a fire, the overall development speed is slow, and the fire's trajectory is unstable. Thus, early-stage firefighting can minimize losses and casualties.

[0045] S1012, select temperature, smoke concentration and CO concentration as specific parameters for fire detection; Specifically, the environment, cause, and form of a fire are all uncertain, making it difficult to analyze the specific parameters of a fire. Based on the fire's occurrence mechanism and development patterns, the system selects temperature, smoke, and CO as specific parameters for fire detection. First, heat is rapidly transferred to surrounding unburned combustibles through conduction, causing the ambient temperature to rise and the fire to spread rapidly. Heat is also one of the conditions for whether a fire continues to spread, and temperature is a manifestation of heat; therefore, temperature is suitable as a specific parameter for a fire. Second, the composition of smoke is related to the burning substances and can vary depending on the combustion conditions and fuel properties. Fires produce a large amount of smoke, making smoke a suitable specific parameter for a fire. Third, fires produce gaseous combustion products such as CO2, CO, gaseous sulfides, and hydrocarbons. However, not all combustion products produce gaseous sulfides or hydrocarbons, and using CO2 as a specific parameter for a fire may lead to false alarms due to poor ventilation and crowds. Therefore, using CO as a specific parameter for a fire is more appropriate.

[0046] S1013, Construct a multi-source sensor network covering the key risk areas of the aviation refueling station to collect the specific parameters.

[0047] Specifically, this embodiment uses a sensor acquisition module to collect various characteristic quantities of a fire. The sensors include a temperature sensor, a smoke sensor, and a CO sensor. In this embodiment, the temperature sensor can use an NTC thermocouple for measurement. The resistance of an NTC thermistor decreases exponentially with increasing temperature, and can be directly converted into the temperature of the measured medium by an electrical instrument. The MQ-2 smoke sensor is selected as the fire combustion smoke sensor. It utilizes SnO2 gas-sensitive material to form an N-type semiconductor on its surface. When it comes into contact with smoke in the environment, the conductivity inside the sensor structure increases with the increase of gas concentration, resulting in an increase in the output analog signal. The MQ-7 CO sensor is selected. Its working principle is based on the fact that the SnO2 gas-sensitive material has low conductivity in clean air, but when CO gas is present in the environment detected by the sensor, the conductivity of SnO2 increases with the increase of CO gas concentration. With the increase in conductivity, the voltage of the output analog signal also increases continuously. The concentration of CO can be accurately measured by observing the change in conductivity.

[0048] S102, preprocess the special parameters to obtain preprocessed special parameters.

[0049] Specifically, fire detection research accumulates a large amount of data in a short period. Directly inputting this data into the feature layer for processing would overload the feature layer, reducing processing speed. Because fire detection sensors possess local decision-making capabilities, they can perform preprocessing on their own data, such as uploading simple values, setting upper and lower limits, and normalization. Sending this preprocessed data to the data fusion process reduces processing pressure and improves computational efficiency.

[0050] Furthermore, the specific steps of step S102 include: S1021, The special parameter is processed by local decision-making based on the variable threshold method.

[0051] Specifically, based on the analysis of fire development, certain characteristic quantities may increase rapidly during the process. Considering the generally limited data processing capabilities of sensors, a variable threshold method is used for local fire decision-making to reduce the sensor processing load. This method adjusts the selected threshold based on the average value of the characteristic quantity within a fixed time window. The variable threshold method in this embodiment is shown below:

[0052] In the formula, Y i (m) Let m represent the th characteristic quantity m (m=1, 2, 3) i The data layer processing result of each data point; x i (m) Representing characteristic quantity m No. i The data collected this time; k Indicates the amount of data sampled within the time window; h(x) This represents the unit step function. During the detection process... Y i (m) When one or more values ​​are 1, it indicates that the feature quantity is abnormal, which may indicate that a fire has occurred. Only then will the data detected by the sensor at the current moment be processed.

[0053] S1022, the special parameters after local decision-making are processed using upper and lower limit amplitude values.

[0054] Specifically, the detection terminal module monitors fire-related parameters in the environment and processes the raw data. However, because it fuses the raw data without processing it, its anti-interference capability is poor. This embodiment uses upper and lower limit amplitude methods. In dynamic systems, these limits are not fixed values ​​but need to be adjusted in real time based on system status or external input. For example, in Simulink, the Saturation module supports setting the upper and lower limits as external input signals; the operation steps are as follows: ① Double-click the Saturation module to open the parameter setting dialog box. ② Set the values ​​of the "Upperlimit" and "Lower limit" fields to variable names or expressions. ③ Check the "Limit input" option and select "external". ④ The module will automatically add two additional input ports, used to connect the dynamic upper and lower limit signals respectively. Through the above method, real-time updates of the upper and lower limits can be achieved, adapting to changes in system status and improving timeliness and accuracy.

[0055] S1023, the preprocessed special parameters are obtained by normalizing the special parameters through a linear function and processing them using upper and lower limit amplitude methods.

[0056] Specifically, different types of sensors collect data with different units of measurement and significantly different ranges of variation. Therefore, in order to ensure that all characteristic parameters function under the same conditions, To improve the classification accuracy of the feature layer, the collected data needs to be normalized. In this embodiment, a linear function is used for normalization, and the specific linear function is as follows:

[0057] In the formula, Y norm Representing characteristic quantity m The normalized value, X max (m) Representing characteristic quantity m The maximum value, X min (m) Representing characteristic quantity m The minimum value.

[0058] In this embodiment, 100 sets of data are taken sequentially from each of the interference, smoldering, and open flame data points, and 100 sets of data are also taken sequentially from the interference data points. Since the interference data points are located differently from the open flame and smoldering flame data points, the data are separated and normalized within the data layer before being passed to the feature layer. The table below shows the results obtained after taking 5 sets of feature values ​​sequentially from the interference, smoldering, and open flame data, respectively, and processing them through the data layer.

[0059]

[0060] S103, creating an improved SRU network for simulation training and identification of fire status.

[0061] Specifically, when a fire signal is detected at an aviation refueling station, the detection values ​​of various fire sensors begin to show abnormalities. Over time, this abnormality may transform into smoldering or open flames in the engine room, meaning that the detection values ​​of the fire detectors have a certain time accumulation effect. Of course, the abnormality may also be due to the influence of certain interference signals or noise, even if there is no actual fire at the aviation refueling station; therefore, the fire states described in this embodiment include smoldering, open flames, and interference fires. In this embodiment, a more realistic fire process is simulated based on the fire development law to obtain simulation data for training the model. By inputting simulation data for training and avoiding the risk of gradient explosion, an improved SRU network based on an adaptive excitation function is formed, realizing the rapid identification of feature quantities containing time-series information.

[0062] Furthermore, the specific steps of step S103 include: S1031, Build an experimental platform based on AC power line carrier communication to collect simulation data of smoldering fire, open flame, and interference fire.

[0063] Specifically, an experimental platform for fire detection terminals based on AC power line carrier communication was constructed, including a fire alarm management system, a smart fire management platform, fire detection terminal modules, a communication switching module, and an equipment management unit. All tests were conducted in a relatively enclosed indoor environment, with combustible materials placed inside containers for fire experiments, and fire detectors installed approximately 1 meter away from the fire source. During system testing, wood and alcohol were selected as combustible materials to simulate fire experiments. The wood used in the experiment was damp, resulting in slow, smoldering combustion without an open flame in the initial stage, releasing a large amount of smoke. In contrast, alcohol combustion exhibited a clear open flame.

[0064] In the simulation experiments, the first step was a smoldering fire test on wood. The test involved a fire detection module, a communication adapter module, and an equipment management unit, connected to a PC-based fire management system and intelligent fire management platform. After the system was powered on and initialized, the combustion experiment was in the early smoldering stage. The fire management system was used to check whether the detection module issued a pre-alarm to the fire scene. The second step was an alcohol combustion test. The wiring was similar to the smoldering fire test. The equipment management unit uploaded the field data to the intelligent fire management platform via a 4G network. A web browser was used to access the management platform to check whether the mounted fire detection module responded to the experimental scene with a pre-alarm, and to check the collected fire data, whether a fire pre-alarm was issued, and whether other functions were normal. The test results for both tests were viewed through the fire management system and the intelligent fire management platform, and simulation data was obtained from the platform interface.

[0065] S1032, Construct an SRU network for identifying fire status.

[0066] Specifically, the SRU network simplifies the computation process at each time step, making most of the computations at each step independent and recursive, thus achieving parallel computation. Simultaneously, the SRU network ensures that the computation at each time step is only related to the previous time step and independent of other earlier time steps, thereby simplifying the computation and training process. In this embodiment, the basic SRU neuron consists of an input gate, a forget gate, a reset gate, an output gate, internal states, and residual connections. By defining a gate composed of a single-layer feedforward network and an activation function to perform dot multiplication operations at the output to merge the inputs, it optimizes the shortcomings of other recurrent neural networks, which have strong time-series dependencies. The specific calculation formula for the SRU network structure in this embodiment is as follows: f t =(σ(W f x t +b f ));r t =(σ(W r x t +b r )); c t =(f*c t-1 +(1-f t Wx t h t =(r t *tanh(c t )+(1-r t )*x t ; In the formula: x t f is the input value at time t of the current layer; t and r tc represents the outputs of the forget gate and reset gate at time t, respectively, and their activation function σ is the sigmoid function; t-1 c is the output value of the cell state at time t-1. t h represents the cell state at time t. t Let W be the hidden state vector at time t; W, W f W r b are the model weight parameters; f b r is the bias vector; * represents the Hadamard product.

[0067] S1033, Input the simulation data and train the SRU network to obtain an improved SRU network based on an adaptive activation function.

[0068] Specifically, in this embodiment, the SRU network incorporates a highway network for performance optimization, accelerating its parallel execution speed. The highway network can improve accuracy by increasing network depth, mitigating gradient vanishing caused by increasing neural network depth. Furthermore, by adding a non-saturated activation function to the activation function after connecting the highway network, the SRU network becomes more sensitive to gradient changes. However, adding a non-saturated activation function may lead to gradient explosion; this embodiment uses an adaptive activation function to address this issue. To fully utilize the advantages of both saturated and non-saturated activation functions, a weighted average of the saturated and non-saturated activation functions is used, calculated as follows: g=1 / u*tanh+1 / v*LeakReLU, u+v=1; In the formula, the unsaturated excitation function

[0069] Saturated excitation function tanh=(e x -e -x ) / (e x -e -x ).

[0070] S104, The preprocessing parameters are input into the improved SRU network to identify the probability of fire status in the corresponding detection area.

[0071] Specifically, by combining the advantages of the improved SRU network in terms of recognition time, and by using the adaptive activation function to avoid gradient vanishing during iterative computation, it is possible to quickly and stably identify the probability of smoldering fires, open flames, and interfering fires, thereby improving the accuracy of fire early warning decisions.

[0072] Furthermore, the specific steps of step S104 include: S1041, the preprocessing parameters are divided into data groups according to the same detection area and the same time, and each data group includes temperature, smoke concentration and CO concentration.

[0073] Specifically, each input group is x t =[x1 t x2 t x3 t ], where x1 t x2 t x3 t These represent the smoke concentration, CO concentration, and temperature at time t, respectively.

[0074] S1042, set the relevant configuration parameters for fire status recognition iteration.

[0075] Specifically, the first 88 groups from the 100 groups were selected as the training set, and the last 12 groups were selected as the test set for experimental simulation. The experiments were conducted on the same machine, with each neural network configured with the same parameters: the number of hidden layer neurons was set to 64, the epoch was 1, and the batch size was 64; the cross-entropy loss function was used for all networks; the learning rate was set to 0.01; and the number of layers in all networks were set to 1, 3, 5, and 7, respectively. After multiple iterations and optimization approximations, u=0.38 and v=0.62 were obtained. The cross-entropy loss function was chosen, as detailed below:

[0076] In the formula, p i y represents the probability that a sample belongs to the i-th class; when a sample belongs to the i-th class, y... i =1, otherwise 0.

[0077] S1043, Based on the configuration parameters, several data groups are input into the improved SRU network for iterative calculation to identify the probability of fire status in the corresponding detection area.

[0078] Specifically, the SRU network can handle nonlinear data relationships and has high accuracy and generalization ability. It can also learn automatically and store a large amount of input-output mapping relationships without needing to describe the mathematical equations of these relationships. During training, normalized data is input, weighted, and then the output is obtained. The SRU network can learn through the backpropagation algorithm, continuously adjusting the weights between the hidden and output layers to improve the model's capabilities by reducing the error between the expected and actual outputs. Through multiple iterative calculations, it achieves tasks such as accurate classification, regression, and prediction of the input data. In this embodiment, 60 test groups were determined. Here, 6 groups of test sample data are selected for analysis and explanation. The comparison between the network's actual output and the expected output of the samples is shown in the table below, where X, Y, and Z represent the probabilities of open flame, smoldering fire, and interfering fire, respectively.

[0079]

[0080] S105, by integrating the fire state probability and decision factors through imprecise reasoning theory, the fire warning level of the corresponding detection area is obtained.

[0081] Specifically, if the output probabilities show a clear one-sidedness, such as 0.87 for open flame, 0.1 for smoldering fire, and 0.03 for interfering fire, a fire can be directly determined to have occurred based on the probabilities. However, when the output results are often ambiguous, a direct judgment cannot be made, and the same ambiguity may appear in the next moment, making it difficult to make a final judgment on the fire. Therefore, it is necessary to use imprecise reasoning theory to obtain accurate detection results. In addition, the probabilities output by existing technologies do not specifically reflect the severity of the fire. Therefore, this implementation incorporates the duration of the detected fire signal and the protection level of the aviation refueling station into the model's decision factors, and performs fuzzy reasoning fusion with the probabilities of open flame and smoldering fire to output a graded alarm decision. This makes the alarm decision for computer room fires more reasonable and facilitates the effective use of fire-fighting resources. This effectively solves the problem that using only the current probabilities of open flame and smoldering fire to characterize the current fire status of the computer room, but not the specific severity of the fire, cannot be determined solely from these probabilities.

[0082] Furthermore, the specific steps of step S105 include: S1051, the protection level of aviation refueling stations and the duration of abnormal signals are selected as decision factors.

[0083] Specifically, during the design process of aviation refueling stations, their level is determined based on their usage, management requirements, and their importance in the economy and society, and they are assumed to be divided into three levels: A, B, and C.

[0084] Data centers that meet one of the following criteria should be classified as Class A: ①A fire will cause significant economic losses; ②A fire would cause serious disruption to public order.

[0085] Data centers that meet one of the following criteria should be classified as Class B: ①A fire will cause significant economic losses; ②A fire will cause chaos in public places.

[0086] The remaining data center computer rooms that do not belong to Class A or Class B are classified as Class C. Different classes require different fire protection resources; therefore, in this embodiment, the protection level of the aviation refueling station is included in the decision factors. Simultaneously, the duration of abnormal signals currently collected by the sensors is closely related to the duration of the fire, and therefore, it is also included in the decision factors for computer room fire alarms.

[0087] S1052, The decision factors are fuzzified using membership functions.

[0088] Specifically, before fuzzy inference, the input quantities must be fuzzified in order to utilize fuzzy rules for inference. In the fuzzification process, firstly, the universe of discourse U is specified; secondly, the fuzzification level is determined based on the actual situation; and finally, based on the existing fuzzy set, a membership function is established. After fuzzification, the input and output quantities are transformed into membership degrees with values ​​in the range [0, 1]. Therefore, this embodiment uses a triangular membership function, the specific form of which is as follows:

[0089] In the formula: u(x) represents the membership function, and a and b represent the numerical values ​​of x.

[0090] S1053, the early warning set output is obtained by fusing the fire state probability and the fuzzified decision factor through a fuzzy logic algorithm.

[0091] Specifically, the core of fuzzy inference is to establish corresponding fuzzy control rules, i.e., various semantic rules, through experts. These fuzzy control rules are composed of a large number of conditional inference statements. These statements are mainly in the form of "IF…THEN…", and their main function is to perform fuzzy inference on the input variables according to the fuzzy control rules to obtain qualitatively descriptive output quantities. The specific semantic rules in this embodiment are as follows: Article 1: IF Low probability of smoldering AND Low probability of open flame AND High protection level of aviation refueling station AND Long duration of abnormal signal THEN Minor fire alarm.

[0092] Article 2: IF Smoldering fire probability is moderate AND open flame probability is low AND aviation refueling station protection level is high AND abnormal signal duration is long THEN serious fire alarm.

[0093] Article 3: If the probability of smoldering fire is high AND the probability of open flame is low AND the protection level of the computer room is low AND the duration of the abnormal signal is short, the fire signal may be an interference signal. The fire protection level of the aviation refueling station is low and the duration of the abnormal signal is short. Therefore, there is no obvious fire for this type of signal output. At this time, staff need to go and check.

[0094] Article 4: IF Low probability of smoldering AND Medium probability of open flame AND Low protection level of aviation refueling station AND Long duration of abnormal signal THEEN Minor fire alarm decision.

[0095] Article 5: If the probability of smoldering fire is high AND the probability of open flame is low AND the protection level of aviation refueling stations is high AND the duration of abnormal signals is long, then a serious fire alarm decision needs to be issued.

[0096] S1054, the output of the warning set is defuzzified using the centroid method to determine the fire warning level of the corresponding detection area.

[0097] Specifically, defuzzification methods include the centroid method and the maximum membership method. This embodiment uses the centroid method, which takes the value of the universe of discourse corresponding to the centroid of the region enclosed by the membership function curve of the fuzzy output set obtained from fuzzy inference and the horizontal axis as the result of refinement. The calculation formula for the centroid method is as follows:

[0098] In the formula: u A (z) The membership function represents the fuzzy output.

[0099] To make rational use of fire-fighting resources, different fire situations require different fire-fighting responses and resource allocations. This embodiment categorizes fire alarms into four levels, from minor to severe, each requiring a different fire-fighting response: Level 1 indicates no obvious fire, meaning there are no obvious fire characteristics in the detection area. If the fire signal originates from interference and the abnormal signal is short-lived, staff can proceed with fire investigation.

[0100] Level 2 indicates a minor fire alarm. If a detector reading in the detection area exceeds the threshold, but the abnormal signal lasts for a short period of time, an immediate inspection is required.

[0101] Level 3 indicates a moderate fire alarm, indicating that a moderate fire has been detected in the detection area, but it may not cause major casualties or there are few people on site. At this time, certain fire-fighting resources are needed to put out the fire. This is also indicated if smoldering fire occurs in an area with a high fire protection level and the abnormal signal lasts for a long time.

[0102] Level 4 indicates a serious fire alarm, meaning a fire has been confirmed in the detection area, is spreading rapidly, and poses a high risk, requiring immediate deployment of sufficient firefighting resources. This could be due to an open flame that has been present for an extended period in an area with a high fire protection level.

[0103] It should be noted that while examples have been given for various types of early warnings, there are many other combinations of fire characteristic signal changes, abnormal currents, and regional fire protection levels that may correspond to different types of fire early warnings.

[0104] In summary: First, collecting specific parameters within the key risk areas of aviation refueling stations for fire detection provides a more realistic reflection of the fire situation and facilitates multi-sensor perception of these parameters. Through sensor complementarity, the system can obtain more comprehensive, accurate, and reliable information. Second, sensors possess local decision-making capabilities, enabling them to preprocess their own data, such as uploading simple values, setting upper and lower limits, and normalizing. Preprocessing these parameters yields preprocessed parameters, which are then sent to data fusion, reducing the data processing burden during fusion. Third, by simulating real fire processes and obtaining simulation data, which is then used for training and to avoid gradient vanishing hazards, an improved SRU network based on an adaptive activation function is formed, enabling rapid identification of features containing temporal information. Fourth, combining the advantages of the improved SRU network in terms of identification time and leveraging the adaptive activation function's ability to avoid gradient vanishing during iterative computation, the probability of identifying smoldering fires, open flames, and interfering fires can be achieved quickly and stably. Finally, by integrating the fire state probability and decision factors through imprecise reasoning theory, the fire warning level of the corresponding detection area can be obtained. This can solve the problem that the current fire state of the computer room can only be characterized by the probability of open flame and smoldering fire, but the specific fire severity cannot be determined solely from this probability. Example 2

[0105] This embodiment provides a structural block diagram of a system corresponding to the method described in Embodiment 1. Figure 2 This is a structural block diagram of the intelligent fire early warning system according to this embodiment, as follows: Figure 2 As shown, the system includes: The data acquisition module 10 is used to acquire specific parameters within the critical risk area of ​​the aviation refueling station for fire detection, including temperature, smoke concentration, and CO concentration. Preprocessing module 20 is used to preprocess the special parameters to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; Create module 30 to create an improved SRU network for simulation training and for identifying fire states, where fire states include smoldering fire, open flame, and interfering fire; The identification module 40 is used to input the preprocessed special parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area; The decision module 50 is used to determine the fire warning level of the corresponding detection area by fusing the fire state probability and decision factors through imprecise reasoning theory.

[0106] Furthermore, the acquisition module 10 specifically includes: Analysis unit 11 is used to analyze the characteristics and development patterns of fire signals at aviation refueling stations; Selection unit 12 is used to select temperature, smoke concentration and CO concentration as specific parameters for fire detection; Construction unit 13 is used to construct a multi-source sensor network covering the key risk areas of the aviation refueling station to collect the specific parameters.

[0107] Furthermore, the preprocessing module 20 specifically includes: Threshold valve unit 21 is used for local decision-making processing of the special parameter based on the variable threshold method; Limiting unit 22 is used to process the localized special parameters using upper and lower limit values; Normalization unit 23 is used to obtain preprocessed special parameters by normalizing the special parameters after processing them by upper and lower limit amplitude methods through linear function normalization.

[0108] Furthermore, the creation module 30 specifically includes: Unit 31 is used to build an experimental platform based on AC power line carrier communication to collect simulation data of smoldering fire, open flame, and interference fire. Construction unit 32 is used to construct an SRU network for identifying fire status; Training unit 33 is used to input the simulation data and train the SRU network to obtain an improved SRU network based on an adaptive activation function.

[0109] Furthermore, the identification module 40 specifically includes: The division unit 41 is used to divide the preprocessed special parameters into data groups according to the same detection area and the same time. Each data group includes temperature, smoke concentration and CO concentration. The setting unit 42 is used to set the relevant configuration parameters for the fire status recognition iteration, wherein the configuration parameters include the number of neurons, the number of network layers, the learning class, and the loss function. The iteration unit 43 is used to input several data groups into the improved SRU network based on the configuration parameters to perform iterative calculations and identify the probability of fire status in the corresponding detection area.

[0110] Furthermore, the decision module 50 specifically includes: Selecting unit 51 is used to select the protection level of the aviation refueling station and the duration of abnormal signals as decision factors; Fuzzy unit 52 is used to fuzzify the decision factors using a membership function; Fusion unit 53 is used to fuse the fire state probability and the fuzzified decision factors through a fuzzy logic algorithm to obtain the early warning set output. The deblurring unit 54 is used to deblur the output of the warning set using the centroid method to determine the fire warning level of the corresponding detection area.

[0111] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination. Example 3

[0112] Specifically, the figure shows a flowchart of the intelligent fire early warning method provided in this embodiment.

[0113] like Figure 3 As shown, the intelligent fire early warning method in this embodiment includes the following steps: S201, collects special parameters within the critical risk area of ​​aviation refueling stations for fire detection.

[0114] S202, preprocess the special parameters to obtain preprocessed special parameters.

[0115] S203, creating an improved SRU network for simulation training and identification of fire status.

[0116] S204, The preprocessed parameters are input into the improved SRU network to identify the probability of fire status in the corresponding detection area.

[0117] S205, by integrating the fire state probability and decision factors through imprecise reasoning theory, the fire warning level of the corresponding detection area is obtained.

[0118] S206, based on the obtained fire warning level, retrieve the emergency strategy program from the pre-made fire emergency strategy library to control and execute the corresponding fire emergency measures.

[0119] In summary, the technical advantage of this embodiment, based on embodiment 1, is that it can automatically generate corresponding execution programs according to different warning levels, thereby achieving full automation of emergency measures. Example 4

[0120] This embodiment provides a structural block diagram of a system corresponding to the method described in Embodiment 3. Figure 4 This is a structural block diagram of the intelligent fire early warning system according to this embodiment, as follows: Figure 4 As shown, the system includes: The data acquisition module 10 is used to acquire specific parameters within the critical risk area of ​​the aviation refueling station for fire detection, including temperature, smoke concentration, and CO concentration. Preprocessing module 20 is used to preprocess the special parameters to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; Create module 30 to create an improved SRU network for simulation training and for identifying fire states, where fire states include smoldering fire, open flame, and interfering fire; The identification module 40 is used to input the preprocessed special parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area; The decision module 50 is used to determine the fire warning level of the corresponding detection area by fusing the fire state probability and decision factors through imprecise reasoning theory.

[0121] The emergency module 60 is used to retrieve emergency strategy programs from a pre-made fire emergency strategy library based on the acquired fire warning level, so as to control the execution of corresponding fire emergency measures.

[0122] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination. Example

[0123] Combination Figure 1 and Figure 3 The intelligent fire early warning method described can be implemented by a computer. Figure 5 This is a schematic diagram of the hardware structure of a computer according to this embodiment.

[0124] The computer may include a processor 71 and a memory 72 storing computer program instructions.

[0125] Specifically, the processor 71 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement this application.

[0126] The memory 72 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 72 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 72 may include removable or non-removable (or fixed) media. Where appropriate, the memory 72 may be internal or external to a data processing device. In a particular embodiment, the memory 72 is non-volatile memory. In a particular embodiment, the memory 72 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0127] The memory 72 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 71.

[0128] The processor 71 reads and executes the computer program instructions stored in the memory 72 to implement the intelligent fire early warning methods of Embodiments 1 and 3 described above.

[0129] In some embodiments, the computer may further include a communication interface 73 and a bus 70. For example, Figure 5 As shown, the processor 71, memory 72, and communication interface 73 are connected through bus 70 and complete communication with each other.

[0130] The communication interface 73 is used to enable communication between the various modules, devices, units, and / or equipment in this application. The communication interface 73 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0131] Bus 70 includes hardware, software, or both, that couples computer components together. Bus 70 includes, but is not limited to, at least one of the following: Data Bus, Address Bus, Control Bus, Expansion Bus, and Local Bus. For example, and not as a limitation, bus 70 may include Accelerated Graphics Port (AGP) or other graphics buses, Extended Industry Standard Architecture (EISA) buses, Front Side Bus (FSB), HyperTransport (HT) interconnects, Industry Standard Architecture (ISA) buses, InfiniBand interconnects, and Low Pin Count (LPC) interconnects. Bus 70 may include a wire, memory bus, MicroChannel Architecture (MCA) bus, Peripheral Component Interconnect (PCI) bus, PCI-Express (PCI-X) bus, Serial Advanced Technology Attachment (SATA) bus, Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 70 may include one or more buses. Although specific buses are described and illustrated in this application, this application contemplates any suitable bus or interconnect.

[0132] The computer can access the intelligent fire early warning system and execute the intelligent fire early warning methods of Embodiments 1 and 3.

[0133] In addition, in conjunction with the intelligent fire early warning methods in Embodiments 1 and 3 above, this application can provide a storage medium for implementation. This storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement the intelligent fire early warning method of Embodiment 1 above.

[0134] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart fire early warning method, characterized in that, include: Collect specific parameters for fire detection within the critical risk area of ​​the aviation refueling station, including temperature, smoke concentration, and CO concentration. The special parameters are preprocessed to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; An improved SRU network was created for simulation training and for identifying fire states, including smoldering fire, open flame, and interfering fire. The preprocessing parameters are input into the improved SRU network to identify the probability of fire status in the corresponding detection area; The fire warning level for the corresponding detection area is obtained by integrating the fire state probability and decision factors using inaccurate reasoning theory.

2. The intelligent fire early warning method according to claim 1, characterized in that, The steps for collecting specific parameters within the critical risk area of ​​the aviation refueling station for fire detection specifically include: Analyze the characteristics and development patterns of fire signals at aviation refueling stations; Temperature, smoke concentration, and CO concentration were selected as specific parameters for fire detection. A multi-source sensor network covering key risk areas of aviation refueling stations is constructed to collect the specific parameters.

3. The intelligent fire early warning method according to claim 1, characterized in that, The step of preprocessing the specific parameters to obtain preprocessed specific parameters specifically includes: The specific parameters are processed using a local decision-making method based on a variable threshold. The specific parameters after local decision-making are processed using upper and lower limit amplitude values. The preprocessed special parameters are obtained by normalizing the special parameters using a linear function and then processing them using upper and lower limit amplitude methods.

4. The intelligent fire early warning method according to claim 1, characterized in that, The steps for creating the improved SRU network for simulation training and fire status identification specifically include: An experimental platform based on AC power line carrier communication was built to collect simulation data of smoldering fire, open flame, and interference fire. Construct an SRU network for identifying fire status; The simulation data is input and the SRU network is trained to obtain an improved SRU network based on an adaptive activation function.

5. The intelligent fire early warning method according to claim 1, characterized in that, The step of inputting the preprocessed special parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area specifically includes: The preprocessing parameters are divided into data groups according to the same detection area and the same time. Each data group includes temperature, smoke concentration, and CO concentration. Set the relevant configuration parameters for fire status recognition iteration, wherein the configuration parameters include the number of neurons, the number of network layers, the learning class, and the loss function; Based on the configuration parameters, several data groups are input into the improved SRU network for iterative calculation to identify the probability of fire status in the corresponding detection area.

6. The intelligent fire early warning method according to claim 1, characterized in that, The specific steps for determining the fire warning level for the corresponding detection area by fusing the fire state probability and decision factors using imprecise reasoning theory include: The protection level of aviation refueling stations and the duration of abnormal signals were selected as decision factors. The decision factors are fuzzified using membership functions; The early warning set output is obtained by fusing the fire state probability and the fuzzified decision factors using a fuzzy logic algorithm. The output of the warning set is defuzzified using the centroid method to determine the fire warning level of the corresponding detection area.

7. The intelligent fire early warning method according to claim 1, characterized in that, After the step of fusing the fire state probability and decision factors through imprecise reasoning theory to derive the fire warning level for the corresponding detection area, the method further includes: Based on the obtained fire warning level, emergency strategy programs are retrieved from the pre-made fire emergency strategy library to control and execute the corresponding fire emergency measures.

8. An intelligent fire early warning system, characterized in that, include: The data acquisition module is used to collect specific parameters in key risk areas of aviation refueling stations for fire detection, including temperature, smoke concentration, and CO concentration. The preprocessing module is used to preprocess the special parameters to obtain preprocessed special parameters, wherein the preprocessing includes local decision-making, upper and lower limit amplitude values, and normalization processing; A module is created to create an improved SRU network for simulation training and for identifying fire states, including smoldering fire, open flame, and interfering fire. The identification module is used to input the preprocessed special parameters into the improved SRU network to identify the probability of fire status in the corresponding detection area; The decision module is used to determine the fire warning level for the corresponding detection area by fusing the fire state probability and decision factors through imprecise reasoning theory.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent fire early warning method as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the intelligent fire early warning method as described in any one of claims 1 to 7.

Images