Intelligent fire hydrant fire extinguishing system and control method

Abstract

The invention discloses an intelligent fire hydrant fire extinguishing system and a control method, and relates to the technical field of fire engineering, the system comprises a sensing module, a decision control module, an execution module, a communication and data management module and a self-diagnosis and health management module; the method comprises the steps that the sensing module collects multi-source heterogeneous data, the decision control module fuses the data to generate a control instruction, and the execution module adjusts the fire hydrant valve and the water supply flow according to the instruction. According to the scheme, comprehensive, accurate and real-time fire situation perception can be realized, the accuracy of fire extinguishing response and the resource utilization efficiency are improved, and the function degradation safety and the overall reliability of the system when the performance of part of components is degraded are ensured; the problems that in the prior art, an intelligent fire hydrant fire extinguishing system is simple in control logic, poor in functional module collaboration and weak in self-diagnosis capacity, and a control method is not suitable for complex city scenes, so that reliability is insufficient, and adaptability is limited are solved.

Description

Technical Field

[0001] This invention relates to the field of fire protection engineering technology, specifically to an intelligent fire hydrant extinguishing system and control method. Background Technology

[0002] With the continuous improvement of urban public safety systems, the level of intelligence in fire protection infrastructure has become a crucial link in safeguarding people's lives and property. Fire hydrants, as the most basic and critical fire-fighting facility, directly determine the success or failure of initial fire suppression due to their response speed, water supply reliability, and ease of operation. Traditional fire hydrant systems are mostly purely mechanical or semi-automatic, relying on manual inspection to determine equipment status. They cannot monitor key parameters such as water pressure, flow rate, valve opening and closing, and ambient fire conditions in real time, often resulting in missed opportunities for optimal fire suppression during sudden fires due to equipment malfunctions, insufficient water pressure, or operational delays. Intelligent fire hydrant systems aim to achieve early fire identification and rapid response through sensing, communication, and automatic control technologies. These systems typically integrate functions such as water pressure monitoring, remote opening and closing, and location reporting to enhance the proactive defense capabilities of fire protection facilities.

[0003] However, existing technologies still have significant shortcomings: First, the control logic is too simple, with most systems only able to execute binary decisions that activate upon detecting an anomaly, lacking a comprehensive assessment of the fire scale, environmental risks, and pipeline capacity; second, the coordination between functional modules is poor, with sensor data, control commands, and water supply scheduling failing to form a closed-loop linkage, easily leading to false activation or delayed response; third, the system's self-diagnostic capabilities are weak, unable to continuously assess its own health status and provide early warnings of potential faults in non-fire alarm states; and finally, the control methods are not adapted to complex urban scenarios, such as high-density building clusters or aging pipeline areas, making it difficult to dynamically adjust fire suppression strategies to balance efficiency and resource constraints.

[0004] The aforementioned problems render existing intelligent fire hydrant systems unreliable and limited in adaptability in practical applications, necessitating a solution from an intelligent fire hydrant extinguishing system and control method that possesses multi-source information fusion, intelligent decision-making, and closed-loop control capabilities. Summary of the Invention

[0005] Based on this, and in response to the above problems, this invention proposes an intelligent fire hydrant extinguishing system and control method, which solves the problems of simple control logic, poor functional module coordination, weak self-diagnosis capability, and unsuitability of the control method for complex urban scenarios in current intelligent fire hydrant extinguishing systems, resulting in insufficient reliability and limited adaptability.

[0006] The technical solution of this invention is: An intelligent fire hydrant extinguishing system includes fire hydrants connected to a water supply network, and further includes: The sensing module is used to collect fire hydrant status parameters, environmental fire parameters, and water supply network pressure parameters. The decision control module communicates with the perception module and is used to receive and integrate multi-source heterogeneous data collected by the perception module, and generate control commands based on a preset intelligent decision model. The execution module, which communicates with the decision control module, is used to receive control commands and execute the opening and closing control of fire hydrant valves and the regulation of water supply flow. The communication and data management module is connected to the sensing module, decision control module and execution module respectively, and is used to realize data interaction and command transmission between the modules within the system and data synchronization with the external fire command center. The self-diagnosis and health management module is integrated into the decision control module and is used to perform periodic status assessments and fault warnings for each component of the system in non-fire alarm states.

[0007] Preferably, the sensing module includes a fire hydrant status sensor group, an environmental fire sensor group, and a pipeline pressure sensor group; the fire hydrant status sensor group includes at least an opening sensor installed at the fire hydrant valve, a flow sensor installed at the fire hydrant outlet, and a water pressure sensor installed inside the fire hydrant; the environmental fire sensor group includes at least a temperature sensor, a smoke sensor, and an infrared thermal imaging sensor distributed at multiple monitoring points on the fire hydrant protective cover and in the adjacent area; the pipeline pressure sensor group is installed on the municipal water supply main connected to the fire hydrant to monitor the pressure stability of the water supply source.

[0008] Preferably, the decision control module includes a data fusion unit, a risk assessment unit, and an instruction generation unit. The data fusion unit receives the raw data stream from the sensing module and uses a Kalman filter-based adaptive weighted fusion algorithm to perform spatiotemporal alignment and noise reduction on the multi-source data, generating a fused feature vector with a unified timestamp. The risk assessment unit has a built-in fire risk assessment model, which takes the aforementioned fused feature vector as input and outputs a quantified fire risk level. .

[0009] Preferably, the instruction generation unit connects the risk assessment unit and the self-diagnosis and health management module. The instruction generation unit has a pre-built hierarchical response strategy library, which includes fire risk levels. This is mapped to a specific set of control commands. These commands include at least the target valve opening, target water supply flow rate, and expected duration. Furthermore, before generating the final command, the command generation unit must call the system health status index provided by the self-diagnosis and health management module. ,when When the threshold is lower than the preset threshold, the instruction generation unit will automatically downgrade the response strategy or add device status alarm information to the instruction.

[0010] Preferably, the self-diagnosis and health management module includes a periodic self-test unit and a fault prediction unit. The periodic self-test unit is triggered according to a preset cycle to calibrate and diagnose the signal drift and zero-point error of each sensor in the sensing module, test the response delay and torque output of the valve motor in the execution module, and evaluate the link quality and data packet loss rate of the communication module. The fault prediction unit uses a long short-term memory network to build a time series prediction model based on historical self-test data and operation logs to predict the remaining service life and potential failure probability of each key component, and generates an early warning signal when the probability exceeds a set threshold.

[0011] A control method for an intelligent fire hydrant extinguishing system, applied to any of the aforementioned intelligent fire hydrant extinguishing systems, comprises the following specific steps: S1, through sensing modules deployed in the fire hydrant and surrounding environment, synchronously collects data on the valve opening, water flow rate, and internal water pressure of the fire hydrant, as well as ambient temperature, smoke concentration, infrared thermal imaging data, and water supply network inlet pressure data from multiple monitoring points. S2, the decision control module receives the multi-source heterogeneous data collected in step S1, and uses the data fusion unit to perform spatiotemporal alignment, noise filtering and feature extraction on the data to generate a fusion feature vector representing the current monitoring scene; S3, the risk assessment unit of the decision control module will integrate the feature vector input into the pre-trained fire risk assessment model to calculate the quantified fire risk level. ; S4, the instruction generation unit of the decision control module generates instructions based on the fire risk level. The system queries the tiered response strategy library to initially determine the target control parameters, and simultaneously obtains the real-time system health status index provided by the self-diagnosis and health management module. ; S5, the instruction generation unit is based on the system health status index. Adaptive adjustments are made to the initially determined target control parameters; like If the value exceeds the health threshold, the original strategy will be executed. like If the water flow rate is below the health threshold but above the safety threshold, the water supply time will be reduced proportionally or the continuous water supply time will be shortened. like If the value falls below the safety threshold, an equipment fault alarm will be generated and manual intervention will be requested, while the automatic start command will be blocked. S6, the final control command generated in step S5 is sent to the execution module through the communication module. The execution module drives the fire hydrant valve motor to adjust the valve to the target opening degree to achieve on-demand water supply. S7. During the firefighting operation, steps S1 to S6 are continuously executed to form a closed-loop control circuit. The control commands are dynamically adjusted based on the real-time feedback of the fire risk level and changes in pipeline pressure until the fire risk level drops below the safety threshold, at which point a smooth shutdown procedure is executed.

[0012] Preferably, in step S2, the data fusion unit uses an adaptive weighted fusion algorithm to assign dynamic weights to sensor data from different sources and of different types. For environmental fire sensor data, the weighting is based on the distance between the sensor and the suspected fire source. and the confidence level of the sensor itself Dynamic adjustment, weight The calculation method is as follows: in This is the distance attenuation coefficient; For fire hydrant status data, its weight Positively correlated with equipment health sub-score; For pipeline pressure data, its weight Automatically increase when pipeline pressure fluctuations exceed twice the historical standard deviation.

[0013] Preferably, in step S3, the fire risk assessment model is a hybrid model based on gradient boosting decision tree and logistic regression; the gradient boosting decision tree is used to mine high-order nonlinear interaction features from the fused feature vector, while the logistic regression model uses the interaction features and the original linear features to fit the final risk probability; during the training phase, the model uses historical fire case data and corresponding sensor data for supervised learning, and determines the optimal hyperparameters through cross-validation.

[0014] Preferably, the fire risk level The calculation process is represented by the following formula: in, , , Representing the first The normalized values ​​of temperature, smoke concentration, and thermal radiation intensity at each monitoring point. This is a nonlinear feature extraction function built upon a multilayer perceptron. For the spatial weighting coefficients of the corresponding monitoring points, and These are the real-time flow rate and valve opening of the current fire hydrant, respectively. A linear function characterizing the contribution of equipment condition to risk. and The balance coefficient for adjusting the weights of environmental risk and equipment condition risk is set to a value within the range of [value missing]. , And satisfy .

[0015] Preferably, in step S4, the hierarchical response strategy library contains at least three response levels; Level 1 response corresponds to The instruction is to open the valve to 100% of its maximum opening and supply water at the maximum design flow rate, while simultaneously activating the emergency water supply plan for nearby fire hydrants. Level 2 response corresponds to The instruction is to open the valve to 60%-80% of its maximum opening and supply water at a moderate flow rate; Level 3 response The instruction is to open the valve to less than 30% of its maximum opening for trial spraying or to maintain a warning standby state.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a sensing module integrating multi-source environmental perception, fire hydrant status monitoring, and pipeline pressure monitoring. Employing an adaptive weighted and model-driven data fusion method, it achieves comprehensive, accurate, and real-time situational awareness of fire scenarios, overcoming the shortcomings of single signal sources, such as false alarms and incomplete information. By designing a fire risk assessment mechanism and a tiered response strategy library based on a hybrid machine learning model, control decisions are upgraded from simple binary judgments to refined and differentiated control based on quantified risk levels, significantly improving the accuracy of fire response and resource utilization efficiency. Furthermore, this invention innovatively introduces a self-diagnosis and health management module throughout the decision-making process, enabling self-monitoring and fault prediction in non-fire alarm states. During fire alarm responses, it dynamically adjusts control strategies based on equipment health status, ensuring functional degradation safety and overall reliability even when some components experience performance degradation. Ultimately, a complete closed loop of perception, decision-making, execution, and feedback is formed, which can dynamically adjust the water supply strategy based on real-time feedback during the fire extinguishing process, adapt to changes in fire intensity and fluctuations in pipeline pressure. It is particularly suitable for complex urban scenarios such as high-density building areas and old pipeline areas, achieving a dynamic balance between fire extinguishing efficiency and infrastructure constraints. It solves the problems of current intelligent fire hydrant fire extinguishing systems, such as simple control logic, poor functional module coordination, weak self-diagnosis capabilities, and control methods that are not suitable for complex urban scenarios, resulting in insufficient reliability and limited adaptability. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall technical architecture of the intelligent fire hydrant fire extinguishing system proposed in this invention; Figure 2 This is a flowchart illustrating the core principle of risk assessment decision-making based on multi-source data fusion and hybrid models in this invention. Figure 3 This is a logical flowchart of the multi-source heterogeneous data acquisition and adaptive weighted fusion of the sensing module in this invention; Figure 4 This is a logical flowchart of the decision control module in this invention generating adaptive control instructions based on risk level and system health status. Figure 5 This is a flowchart illustrating the overall process framework of the closed-loop control method for sensing, decision-making, execution, and feedback in this invention. Detailed Implementation

[0019] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0020] In the description of the embodiments of the present invention, it should be understood that the terms "length", "vertical", "horizontal", "top", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0021] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0022] In this embodiment of the invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this embodiment of the invention according to the specific circumstances.

[0023] In embodiments of the present invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0024] The following disclosure provides many different implementations or examples for carrying out different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. Furthermore, reference numerals and / or reference letters may be repeated in different examples of the embodiments of the present invention; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various implementations and / or arrangements discussed.

[0025] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] like Figures 1 to 5 As shown, this embodiment discloses an intelligent fire hydrant extinguishing system. The fire hydrant connected to the water supply network also includes: The sensing module is used to collect fire hydrant status parameters, environmental fire parameters, and water supply network pressure parameters. The decision control module communicates with the perception module and is used to receive and integrate multi-source heterogeneous data collected by the perception module, and generate control commands based on a preset intelligent decision model. The execution module, which communicates with the decision control module, is used to receive control commands and execute the opening and closing control of fire hydrant valves and the regulation of water supply flow. The communication and data management module is connected to the sensing module, decision control module and execution module respectively, and is used to realize data interaction and command transmission between the modules within the system and data synchronization with the external fire command center. The self-diagnosis and health management module is integrated into the decision control module and is used to perform periodic status assessments and fault warnings for each component of the system in non-fire alarm states.

[0027] Preferably, the sensing module includes a fire hydrant status sensor group, an environmental fire sensor group, and a pipeline pressure sensor group; the fire hydrant status sensor group includes at least an opening sensor installed at the fire hydrant valve, a flow sensor installed at the fire hydrant outlet, and a water pressure sensor installed inside the fire hydrant; the environmental fire sensor group includes at least a temperature sensor, a smoke sensor, and an infrared thermal imaging sensor distributed at multiple monitoring points on the fire hydrant protective cover and in the adjacent area; the pipeline pressure sensor group is installed on the municipal water supply pipe connected to the fire hydrant to monitor the pressure stability of the water supply source, and the number of fire hydrants connected to the water supply network is at least one, but can also be multiple.

[0028] Furthermore, the decision control module includes a data fusion unit, a risk assessment unit, and an instruction generation unit. The data fusion unit receives the raw data stream from the sensing module and uses a Kalman filter-based adaptive weighted fusion algorithm to perform spatiotemporal alignment and noise reduction on the multi-source data, generating a fused feature vector with a unified timestamp. The risk assessment unit has a built-in fire risk assessment model, which takes the aforementioned fused feature vector as input and outputs a quantified fire risk level. .

[0029] Fire risk level The calculation process is represented by the following formula: in, , , Representing the first The normalized values ​​of temperature, smoke concentration, and thermal radiation intensity at each monitoring point. This is a nonlinear feature extraction function built upon a multilayer perceptron. For the spatial weighting coefficients of the corresponding monitoring points, and These are the real-time flow rate and valve opening of the current fire hydrant, respectively. A linear function characterizing the contribution of equipment condition to risk. and The balance coefficient for adjusting the weights of environmental risk and equipment condition risk is set to a value within the range of [value missing]. , And satisfy .

[0030] The core of the formula construction is to solve the shortcomings of existing systems that rely solely on a single signal for judgment and ignore the coordination between equipment status and environment. Through the design of core environmental fire parameters, auxiliary equipment status parameters, and dynamic weight balancing, the risk level can be accurately quantified to meet the fire extinguishing decision-making needs of complex urban scenarios.

[0031] Among them, environmental fire parameters , , The selection logic is as follows: The core physical characteristics of a fire are temperature rise, smoke generation, and thermal radiation diffusion. These three constitute a complete signal chain from the initial stage to the spread of a fire, and none of them can be omitted. Single parameters are prone to misjudgment, such as high temperature not indicating a fire, or smoke not indicating an open flame. Multi-parameter fusion can cover the entire stage of smoldering, open flame, and spread, solving the problem of false alarms or missed alarms caused by the one-sided signals of existing systems.

[0032] Normalization is necessary because the range and accuracy of sensors at different monitoring points vary. Normalization can eliminate the influence of dimensions and ensure that data from different sources can be compared and fused horizontally.

[0033] Spatial weight coefficient The selection logic is as follows: The reliability of signals varies depending on the distance and obstruction between distributed monitoring points and the fire source. For example, data from a sensor 5 meters away from a suspected fire source is more reliable than data from 50 meters away. A spatial weighting coefficient is introduced. It can assign higher weight to monitoring points near fire sources and those with unobstructed views, thus solving the problem of risk assessment bias caused by treating all monitoring point data equally.

[0034] Equipment status parameters Selection logic: The fire extinguishing capacity of a fire hydrant directly depends on its own operating status and flow rate. Determine the intensity of the fire extinguishing water supply and the valve opening. This reflects the effectiveness of equipment response. For example, if a valve is stuck, the opening degree cannot reach the set value, and the flow rate will also be limited. Existing systems ignore equipment status, which may lead to a contradiction where the fire situation is accurately assessed but the equipment cannot meet the firefighting requirements. Introducing both risk levels and firefighting feasibility allows the risk level to reflect not only the severity of the fire but also the feasibility of firefighting.

[0035] Balance coefficient , Selection logic: Fire risk assessment needs to take into account both the threat level of the fire itself and whether the equipment can cope with the threat. The dominant environmental fire situation (the core contradiction). Supplement equipment status (basic support), and Ensure weight normalization to avoid double counting. Value range. , This is because the fire threat is the primary consideration, and the equipment status is used for auxiliary correction, which is in line with the fire extinguishing decision-making logic of first assessing the fire situation and then adapting the equipment.

[0036] Feature extraction function , The selection logic is as follows: the interaction of environmental fire parameters is a complex nonlinear relationship. For example, the risks of high temperature, high smoke, and high heat radiation are not simply added together, but increase exponentially. Therefore, a multilayer sensing mechanism is used to construct... Mining high-order interaction features; the relationship between device status parameters and risk is linear: the higher the flow and the greater the opening, the stronger the device's ability to cope with risk, and the lower the risk contribution. Therefore, using... Simplify representation while balancing computational efficiency and accuracy.

[0037] The core advantages of formula construction: Multi-parameter fusion and spatial weighting: solve the problems of false alarms from single signals and long-distance data interference, allowing risk assessment to focus on the core area of ​​the actual fire; A dual-dimensional approach encompassing both environment and equipment: This addresses the problem of focusing solely on the fire situation without considering equipment, leading to a disconnect between decision-making and actual firefighting capabilities, and ensures that risk levels simultaneously reflect both the degree of threat and the feasibility of response. Combination of nonlinear and linear functions: Resolves the contradiction between the difficulty in characterizing complex features and the insufficient accuracy of simple functions, while taking into account both the complexity of fire situation interaction and the intuitiveness of equipment status; Weights are adjustable: via and The value can be adapted to different urban scenarios - newly built urban areas or old urban areas, solving the problems of rigidity and limited adaptability of existing system control methods.

[0038] In summary, the formula is constructed based on the core fire signal, modified by the equipment status, and adapted by dynamic weights. The selection and design of each parameter directly addresses the pain points of existing technologies, ultimately achieving accurate, comprehensive, and scenario-based quantification of fire risk levels, providing a scientific basis for graded fire suppression decisions.

[0039] In addition, the instruction generation unit connects the risk assessment unit and the self-diagnosis and health management module; the instruction generation unit has a pre-built hierarchical response strategy library, which includes fire risk levels. This is mapped to a specific set of control commands, which at least include the target valve opening, target water supply flow rate, and expected duration. Simultaneously, before generating the final command, the command generation unit must call the system health status index provided by the self-diagnosis and health management module. ,when When the threshold is lower than the preset threshold, the instruction generation unit will automatically downgrade the response strategy or add device status alarm information to the instruction.

[0040] Preferably, the self-diagnosis and health management module includes a periodic self-test unit and a fault prediction unit. The periodic self-test unit is triggered according to a preset cycle to calibrate and diagnose the signal drift and zero-point error of each sensor in the sensing module, test the response delay and torque output of the valve motor in the execution module, and evaluate the link quality and data packet loss rate of the communication module. The fault prediction unit uses a long short-term memory network to build a time series prediction model based on historical self-test data and operation logs to predict the remaining service life and potential failure probability of each key component, and generates an early warning signal when the probability exceeds a set threshold.

[0041] A control method for an intelligent fire hydrant extinguishing system, comprising the following steps: S1, through sensing modules deployed in the fire hydrant and surrounding environment, synchronously collects data on the valve opening, water flow rate, and internal water pressure of the fire hydrant, as well as ambient temperature, smoke concentration, infrared thermal imaging data, and water supply network inlet pressure data from multiple monitoring points. S2, the decision control module receives the multi-source heterogeneous data collected in step S1, and uses the data fusion unit to perform spatiotemporal alignment, noise filtering and feature extraction on the data to generate a fusion feature vector representing the current monitoring scene; S3, the risk assessment unit of the decision control module will integrate the feature vector input into the pre-trained fire risk assessment model to calculate the quantified fire risk level. The model comprehensively considers fire intensity, spread trend, and equipment availability; S4, the instruction generation unit of the decision control module generates instructions based on the fire risk level. The system queries the tiered response strategy library to initially determine the target control parameters, and simultaneously obtains the real-time system health status index provided by the self-diagnosis and health management module. ; S5, the instruction generation unit is based on the system health status index. Adaptive adjustments are made to the initially determined target control parameters; like If the value exceeds the health threshold, the original strategy will be executed. like If the water flow rate is below the health threshold but above the safety threshold, the water supply time will be reduced proportionally or the continuous water supply time will be shortened. like If the value falls below the safety threshold, an equipment fault alarm will be generated and manual intervention will be requested, while the automatic start command will be blocked. S6, the final control command generated in step S5 is sent to the execution module through the communication module. The execution module drives the fire hydrant valve motor to adjust the valve to the target opening degree to achieve on-demand water supply. S7. During the firefighting operation, steps S1 to S6 are continuously executed to form a closed-loop control circuit of perception, decision-making, execution and feedback. Based on the real-time feedback of fire risk level and pipeline pressure changes, the control commands are dynamically adjusted until the fire risk level drops below the safety threshold, and then a smooth shutdown procedure is executed.

[0042] Preferably, in step S2, the data fusion unit uses an adaptive weighted fusion algorithm to assign dynamic weights to sensor data from different sources and of different types. For environmental fire sensor data, the weighting is based on the distance between the sensor and the suspected fire source. and the confidence level of the sensor itself Dynamic adjustment, weight The calculation method is as follows: in This is the distance attenuation coefficient; For fire hydrant status data, its weight Positively correlated with equipment health sub-score; For pipeline pressure data, its weight Automatically increase when pipeline pressure fluctuations exceed twice the historical standard deviation.

[0043] Furthermore, in step S3, the fire risk assessment model is a hybrid model based on gradient boosting decision trees and logistic regression. The gradient boosting decision tree is used to mine high-order nonlinear interaction features from the fused feature vectors, while the logistic regression model uses the interaction features and the original linear features to fit the final risk probability. During the training phase, the model uses historical fire case data and corresponding sensor data for supervised learning, and determines the optimal hyperparameters through cross-validation.

[0044] In addition, in step S4, the hierarchical response strategy library contains at least three response levels; Level 1 response corresponds to The instruction is to open the valve to 100% of its maximum opening and supply water at the maximum design flow rate, while simultaneously activating the emergency water supply plan for nearby fire hydrants. Level 2 response corresponds to The instruction is to open the valve to 60%-80% of its maximum opening and supply water at a moderate flow rate; Level 3 response The instruction is to open the valve to less than 30% of its maximum opening for trial spraying or to maintain a warning standby state.

[0045] Example 1: In high-rise complexes located in core urban business districts, underground parking garages, equipment floors, and commercial podiums are densely populated with numerous electrical devices, posing a high risk of fire, and traditional fire response methods are often delayed. This invention's intelligent fire hydrant system is deployed at key nodes of the complex's fire protection ring network, aiming to achieve autonomous detection, accurate assessment, and rapid, adaptive fire suppression response to initial fires.

[0046] The intelligent fire hydrant extinguishing system of this invention includes a sensing module, a decision control module, an execution module, a communication and data management module, and a self-diagnosis and health management module integrated within the decision control module. Each module interacts with the others and transmits commands through a high-speed, reliable intranet built by the communication and data management module. Simultaneously, this module maintains synchronization with the data platform of the city's fire command center via a dedicated line or encrypted wireless network, enabling remote monitoring and collaborative command.

[0047] The sensing module is used to collect multi-source heterogeneous data related to the fire situation and the system's own operation around the clock and from all directions. The sensing module specifically includes a fire hydrant status sensor group, an environmental fire situation sensor group, and a pipeline pressure sensor group.

[0048] The fire hydrant status sensor group is directly installed on the fire hydrant. Among them, the opening sensor adopts a high-precision rotary encoder, which is rigidly connected to the valve stem of the fire hydrant itself. It measures the valve rotation angle in real time and converts it into an opening percentage value from 0% to 100%, with a measurement accuracy of 0.5%. The flow sensor uses an electromagnetic flow meter, which is installed between the fire hydrant outlet and the fire hose interface to continuously monitor the volumetric flow rate of water flowing through the fire hydrant. The data output frequency is 10 Hz, and the range covers 0 to 100 liters per second. The water pressure sensor uses a piezoresistive sensing element embedded in the internal flow channel of the fire hydrant valve body to directly sense the static and dynamic pressure of the water and output a standard current signal of 4 to 20 mA, corresponding to a pressure range of 0 to 1.6 MPa.

[0049] The environmental fire sensor array is distributed inside the fire hydrant protective cover and at multiple monitoring points within a radius of 15-30 meters centered on it, such as on the ceiling, corners, or pillars.

[0050] The temperature sensor uses a digital thermopile array, which can not only measure the ambient air temperature, but also sense changes in temperature gradient. The smoke sensor employs a dual redundancy design based on both photoelectric scattering and ion detection principles to reduce the false alarm rate. The infrared thermal imaging sensor is fixed on a gimbal that can rotate 360 ​​degrees horizontally and 90 degrees vertically. It scans the monitoring area at a rate of 5 frames per second to generate a thermal map of temperature distribution, and its thermal sensitivity is better than 0.05 degrees Celsius.

[0051] The pipeline pressure sensor group is installed at the tee of the municipal water supply main pipe connected to the branch pipe of the fire hydrant. It uses a flange-type pressure transmitter to continuously monitor the pressure stability of the water supply source, with a data sampling interval of 1 second.

[0052] All sensors have independent signal conditioning circuits and analog-to-digital conversion units, which convert the raw physical quantities into digital signals and then package and send them via fieldbus protocol.

[0053] The decision control module is deployed in a waterproof and dustproof control cabinet on site. This module includes a data fusion unit, a risk assessment unit, and an instruction generation unit.

[0054] The data fusion unit continuously receives raw data streams from the sensing module. These data have microsecond-level differences in timestamps, originate from different spatial locations, and vary in data type and dimension. The primary task of the data fusion unit is spatiotemporal alignment. Internally, it maintains a high-precision synchronous clock, assigning a uniform system timestamp to all input data packets. For data with different sampling frequencies, linear interpolation is used to resample them to a uniform 10 Hz time series. Noise filtering is then performed. For slowly varying signals such as temperature and pressure, moving average filtering combined with wavelet thresholding is used for noise reduction; for signals that may contain impulse noise, such as flow rate and vibration, median filtering is used.

[0055] After preprocessing, the core feature extraction and adaptive weighted fusion stage begins. For environmental fire data, the fusion weights are dynamically adjusted based on the Euclidean distance between the sensor and the system's preliminary assessment of suspected fire sources using thermal imaging, as well as the sensor's own historical confidence level.

[0056] Specifically, the dynamic weight of each environmental sensor is calculated as follows: the weight equals the sensor confidence level multiplied by the natural constant. negative distance attenuation coefficient With distance The product power. Among them, the sensor confidence score is calculated by statistically analyzing the consistency of recent data from the sensor with data from surrounding sensors and its self-diagnostic status, and its value ranges from 0 to 1; the distance attenuation coefficient... The value is preset according to the building structure type, with smaller values ​​in open areas and larger values ​​in areas with more partitions. It can be set based on historical data.

[0057] For fire hydrant status data, its fusion weight is positively correlated with the current health sub-score of the device, which is provided by the self-diagnosis and health management module and ranges from 0 to 100.

[0058] For pipeline pressure data, the basic weight is relatively low, but when the fluctuation of the real-time pressure value relative to its historical mean exceeds twice the historical standard deviation, its weight is automatically increased to three times the basic value to strengthen the impact of pipeline anomalies on decision-making.

[0059] Finally, all the weighted feature data are concatenated into a high-dimensional fusion feature vector, which comprehensively represents the fire situation and system status in the current monitoring scenario.

[0060] The risk assessment unit incorporates a fire risk assessment model trained on extensive historical fire case data. This model is a hybrid of gradient boosting decision trees and logistic regression. The gradient boosting decision tree component takes the aforementioned fused feature vectors as input and constructs multiple decision trees to iteratively uncover the complex high-order nonlinear interactions between features such as temperature, smoke, thermal radiation, flow rate, and pressure, outputting a set of deep nonlinear features. These deep features, along with the original linear features, are input to the logistic regression component. The logistic regression component performs the final regression calculation, outputting a quantitative fire risk level value between 0 and 1. This value is calculated using the following formula: Fire risk level equals environmental risk weight coefficient. Multiply by the weighted sum of the normalized values ​​of temperature, smoke concentration, and thermal radiation intensity at each monitoring point after nonlinear feature extraction, and add the equipment condition risk weight coefficient. Multiply by a linear function value based on the current real-time flow rate of the fire hydrant and the valve opening.

[0061] Among them, the nonlinear feature extraction function is constructed by a multilayer perceptron; the spatial weight coefficient reflects the difference in importance of different monitoring point locations; the linear function quantifies the contribution of the current operating status of the equipment to the overall risk; and the environmental risk weight coefficient... The value ranges from 0.6 to 0.8, representing the equipment condition risk weighting coefficient. The value of is between 0.2 and 0.4, and the sum of the two is always 1. This model transforms the ambiguous fire threat into a precise numerical indicator by comprehensively considering the fire intensity, spread trend, and availability of fire hydrants.

[0062] The instruction generation unit connects the risk assessment unit and the self-diagnosis and health management module. It internally contains a tiered response strategy library. This library maps continuous fire risk level ranges to discrete response levels and associates them with specific sets of control parameters. The strategy library contains at least three response levels. Level 1 response corresponds to a fire risk level between 0.7 and 1.0, classified as a major fire. The instruction is as follows: the drive execution module opens the fire hydrant valve to 100% of its maximum opening, supplying water at the system's maximum designed flow rate. Simultaneously, the communication and data management module sends an alarm to the fire command center and activates the emergency plan, requesting the coordinated opening of at least two nearby fire hydrants for pressurized water supply. Level 2 response corresponds to a fire risk level between 0.4 and 0.7, classified as a moderate fire. The instruction is as follows: the valve is opened to a value between 60% and 80% of its maximum opening, supplying water at a moderate flow rate, with a continuous water supply time set to 300 seconds. Simultaneously, the system feedback cycle is shortened to 1 second, closely monitoring changes in the fire's intensity. A Level 3 response corresponds to a fire risk level in the range of 0.1 to 0.4, which is determined to be an initial fire or a suspected fire. The instructions are as follows: open the valve to less than 30% of its maximum opening and conduct a short-term, low-flow exploratory spray for 30 seconds, or simply keep the valve in standby mode and increase the frequency of environmental monitoring.

[0063] Before generating the final instruction, the instruction generation unit forcibly invokes the real-time system health status index provided by the self-diagnosis and health management module. This system health status index is a comprehensive score from 0 to 100, calculated by weighting the health status of each sub-module. The calculation rules are as follows: sensor group health sub-score, weight 40%; execution module health sub-score, weight 30%; communication module health sub-score, weight 20%; and power module health sub-score, weight 10%. Each sub-score is based on periodic self-test data and fault prediction probability. A sub-score of 100 points is awarded when there is no fault. If the fault prediction probability of a single component exceeds 30%, 50 points are deducted from the corresponding sub-score; if it exceeds 50%, 80 points are deducted from the corresponding sub-score.

[0064] The instruction generation unit adaptively modifies the preliminary control parameters retrieved from the strategy library based on the system's health status index. If the system health status index is higher than the health threshold of 90, the original instructions from the strategy library are executed completely. If the system health status index is between the safety threshold of 70 and 90, a slight performance degradation is considered, and the target water supply flow rate or continuous water supply time is reduced proportionally by multiplying the target flow rate by the system health status index divided by 100. If the system health status index is lower than 70, a critical hidden danger that may affect fire extinguishing safety is determined, and the instruction generation unit will not generate an automatic start instruction. Instead, it will generate equipment fault alarm information containing specific fault codes, which will be reported to the fire command center via the communication module to request manual intervention. The automatic start function of the unit will be logically blocked to prevent malfunction.

[0065] The execution module is mainly composed of a high-performance waterproof servo motor, a high-reduction-ratio planetary gearbox, a clutch, and a position feedback unit. This module receives the final control commands from the command generation unit of the decision control module. Commands are typically issued in the form of a target opening percentage and an expected duration. After parsing the command, the servo motor driver drives the motor to rotate, which in turn drives the fire hydrant valve stem to rotate precisely through the gearbox. The position feedback unit transmits the actual opening value back in real time, forming a closed-loop position control to ensure the valve remains stable at the target opening. Flow regulation is indirectly achieved by controlling the opening. The system internally stores calibration curves of the opening and flow rate of the fire hydrant under different inlet pressures for feedforward compensation. The execution module also has over-torque protection and stall detection functions; it automatically shuts down and reports a fault when the drive current abnormally increases.

[0066] The communication and data management module adopts an architecture combining industrial Ethernet and redundant wireless networks. This module defines a unified internal communication protocol, responsible for the uplink aggregation of data from the sensing modules, the downlink distribution of instructions from the decision control modules, and the exchange of status information between modules. All transmitted data is augmented with serial numbers and checksums to ensure integrity. The module establishes a data buffer for short-term storage of high-frequency acquired data and, according to preset rules, uploads key data packets—including fused feature vectors, fire risk levels, control instructions, and system health status indices—compressed and encrypted every 5 minutes to the external fire command center's data platform, achieving remote data synchronization and historical archiving. Simultaneously, this module also receives remote instructions or policy library update packages from the command center, which take effect after security authentication.

[0067] The self-diagnosis and health management module is integrated into the software layer of the decision control module, and includes a periodic self-test unit and a fault prediction unit. The periodic self-test unit automatically triggers the system-level self-test process every 6 hours according to a preset cycle.

[0068] The self-test process includes several sub-items: For the sensing module, standard test signals are sent sequentially to each sensor to verify whether their output is within the allowable error range and to check for signal drift and zero-point error; for the execution module, the valve is controlled to perform small-amplitude opening and closing actions to test the motor response delay time, measure the steady-state operating torque, and compare it with the benchmark value; for the communication module, an end-to-end data packet loopback test is initiated to statistically analyze link delay and data packet loss rate. All self-test results are quantified, scored, and recorded in the operation log. The fault prediction unit, based on accumulated historical self-test data and operation logs, uses a long short-term memory network to build a time series prediction model. This model uses sensor reading deviation, motor response delay growth trend, and communication bit error rate as features to learn their evolution over time, thereby predicting the remaining service life of key components such as decreased temperature sensor sensitivity, valve motor brush wear, and network module component aging, as well as the probability of failure within the next 24 hours. When the predicted probability exceeds a set threshold, such as 30%, the fault prediction unit generates an early warning signal to notify maintenance personnel in advance, achieving predictive maintenance.

[0069] The control method of this invention forms a complete closed loop of perception, decision-making, execution, and feedback. The method begins in step S1, where sensing modules deployed in and around the fire hydrant simultaneously collect data on the fire hydrant's valve opening, water flow rate, and internal water pressure; ambient temperature, smoke concentration, and infrared thermal imaging data from multiple monitoring points; and water supply network inlet pressure data, with each monitoring point corresponding to a set of independent sensor data. All data acquisition is strictly synchronized via hardware triggering to ensure spatiotemporal consistency.

[0070] In step S2, the data fusion unit of the decision control module receives the multi-source heterogeneous data streams collected in step S1. This unit first invokes the time synchronization service to add a precise timestamp based on the GPS clock to all data packets. Subsequently, the data enters different preprocessing pipelines according to their type. For temperature sequences, a first-order hysteresis compensation filter is applied to eliminate delays caused by sensor thermal inertia; for thermal imaging data streams, non-uniformity correction and bad pixel replacement are performed first, followed by extraction of the coordinates of the highest temperature area and its average temperature value. After preprocessing, an adaptive weighted fusion algorithm is executed. The algorithm calculates the dynamic weight of each data source in real time: the weight of environmental fire data is dynamically adjusted based on its distance from the hotspot identified by thermal imaging and the confidence level calculated from the sensor's recent false alarm rate; the weight of fire hydrant status data is linearly correlated with its health sub-score; and the weight of pipeline pressure data is dynamically scaled according to the deviation of its instantaneous value from the average value over the past hour. The weighted data from each dimension are combined into a fusion feature vector of length 50, which is updated 10 times per second.

[0071] In step S3, the risk assessment unit inputs the fused feature vector, updated every second, into the pre-trained fire risk assessment model. The fire risk assessment model is first processed by a gradient boosting decision tree component, which contains 100 decision trees of depth 6, effectively capturing complex patterns such as a sudden increase in smoke concentration accompanied by an expansion of the temperature gradient in a specific area. The high-order features output by the gradient boosting decision tree, along with the linear components of the original features, such as the absolute value of pipeline pressure and the current valve opening, are input into the logistic regression layer. The logistic regression layer outputs a fire risk level value between 0 and 1. During the training phase, the fire risk assessment model uses at least 5000 labeled historical fire and simulated fire scenario data points and determines the optimal hyperparameter combination through five-fold cross-validation, achieving an area under the receiver operating characteristic (AUC) of 0.95 on the test set.

[0072] In step S4, the instruction generation unit queries the graded response strategy library based on the fire risk level value calculated in step S3 to initially determine the target control parameters, such as a target opening degree of 80% and an expected flow rate of 65 liters per second. Simultaneously, the instruction generation unit obtains the system health status index provided by the self-diagnosis and health management module in real time through an internal interface. This system health status index is calculated by weighting the latest score from the periodic self-inspection unit and the health trend score from the fault prediction unit, and is updated every 10 seconds. The update rules are as follows: if no new fault warnings are added during the period, the score of the previous period is maintained; if a minor fault is added to a single submodule with a predicted probability ≤30%, 20 points are deducted from the corresponding sub-score and the total score is recalculated; if a fault warning is added to a critical component with a predicted probability >30%, an emergency scoring mechanism is immediately triggered, deducting 50-80 points from the corresponding sub-score, updating the total score, and simultaneously issuing an alarm.

[0073] Step S5 is a crucial step in the adaptive correction of instructions. The instruction generation unit incorporates a three-level correction logic. First, it compares the system health status index with a preset health threshold of 90. If the index is higher than 90, the system is considered fully healthy, and the initial parameters from the strategy library are directly adopted. Second, if the index is between the safety baseline of 70 and the health threshold of 90, it enters the degradation logic. The correction algorithm calculates a decay factor based on the index value; for example, the decay factor is 0.9 when the index is 80. The initially determined target water supply flow rate is multiplied by this decay factor, or the continuous water supply time is shortened proportionally. Finally, if the index is lower than 70, the safety blockade logic is triggered. The instruction generation unit will not output any valve opening instructions, but instead generate a high-priority fault alarm data packet containing the specific module fault code that caused the index to be too low, and immediately report it through the emergency channel of the communication and data management module. At the same time, a flag to prevent automatic start is set in the local logic. This flag can only be reset after the system health status index recovers to above 85 after maintenance and is confirmed remotely.

[0074] Step S6: The final control command generated after the correction in step S5 is sent to the execution module through the real-time control channel of the communication and data management module. The command message adopts a fixed structure, including fields such as command type, target opening degree, expected flow rate, and action duration. After parsing the message, the controller of the execution module drives the servo motor to move. The motor control algorithm adopts a three-loop control of position loop, speed loop, and current loop to ensure that the valve can quickly and smoothly reach the target opening degree, and the overshoot during the dynamic adjustment process is strictly controlled within 2%. After the valve reaches the target position, the system enters the steady-state maintenance phase.

[0075] In step S7, after the fire extinguishing operation is initiated, the system does not stop working but continues to execute steps S1 to S6 in a high-speed loop, forming a dynamic closed loop. Within this loop, the system calculates an updated fire risk level based on the latest real-time feedback of sensing data and monitors changes in pipeline pressure. For example, when multiple fire hydrants are activated simultaneously, causing a drop in pipeline pressure, the automatic increase in the weight of the pressure sensor data will allow the system to perceive this constraint. The risk assessment model may therefore output a slightly lower risk level or the same level but with a pressure constraint flag. The command generation unit will then fine-tune accordingly, appropriately reducing the flow rate setpoint while ensuring fire extinguishing effectiveness to prevent pipeline collapse. The closed loop continues to run until the fire risk level value remains below the safety threshold of 0.1 for 30 seconds, at which point the system determines that the fire is under control or extinguished. At this point, the command generation unit no longer generates an opening command but instead generates a smooth closing command, controlling the execution module to slowly close the valve at a rate of 5% opening per second to avoid water hammer impact on the pipeline. After closing, the system automatically resets and returns to 24 / 7 monitoring mode.

[0076] Example 2: In complex scenarios such as significant pressure fluctuations in water supply networks in old urban areas and potential obstruction of fire lanes, the system and method of this invention demonstrate unique advantages by enhancing adaptability and coordination. This embodiment takes a row of buildings in a historical preservation area as an example, where the municipal pipeline diameter is small, the pressure is unstable, and the alleyways are narrow.

[0077] In this embodiment, the system deployment is similar to that in Embodiment 1, but the strategy library of the decision control module and the collaborative function of the communication module are specially configured. In the graded response strategy library, a pressure compensation sub-strategy has been added to address the large fluctuations in pipeline pressure. When the real-time pipeline pressure value obtained by the instruction generation unit from the data fusion unit is lower than 0.25 MPa, even if the fire risk level reaches Level 2 response, the system will automatically raise the response level by half a level, i.e., adopt parameters closer to Level 1 response, but simultaneously activate the low-pressure mode. In this mode, the system sends a collaborative request to other smart fire hydrants deployed in the same fire pipeline zone and closest to the fire source point through the communication and data management module. Upon receiving the request, the decision control module of the fire hydrant, provided its self-diagnostic status is good, will assess the necessity of collaboration based on the fire source location information shared by the main fire hydrant and its own sensing data, and automatically activate it to assist water supply with a lower opening degree, jointly increasing the pressure at the ignition point, rather than simply opening all hydrants and causing a momentary pressure drop in the pipeline.

[0078] Furthermore, the fault prediction unit of the self-diagnosis and health management module is more active in this scenario. Due to the humid environment in older areas, sensors are more susceptible to damage. The long short-term memory network model of the fault prediction unit is specifically enhanced for predicting sensor signal drift and intermittent failures. When the probability of failure of a critical environmental fire sensor exceeds 40% within the next few hours is predicted, the system not only issues an early warning but also proactively adjusts the data fusion strategy, reducing the weight of that sensor and attempting to compensate for functional deficiencies by relying on other types of sensors in the same area. For example, when a smoke sensor fails, the fusion weight of infrared thermal imaging and temperature sensors is increased to maintain the system's sensing redundancy.

[0079] In this embodiment, the system health status index provided by the self-diagnosis and health management module is adapted to the environmental characteristics of old urban areas. Its customized rules are as follows: the sensor group health sub-score adds an additional humidity adaptability deduction item. If the environmental humidity exceeds 85% for more than 24 hours, the corresponding sub-score will be deducted by 10 points; the execution module health sub-score strengthens the weight of valve corrosion detection. When the valve torque fluctuation exceeds the benchmark value by 20%, 30 points will be deducted; the communication module health sub-score adds a lane obstruction compensation coefficient. When the link packet loss rate exceeds 5% but is ≤10%, no points will be deducted after correction by the signal enhancement algorithm. When the packet loss rate exceeds 10%, 20 points will be deducted; the health threshold of the system health status index is lowered to 85, and the safety bottom line is lowered to 65 to adapt to the actual situation that the health status of old equipment is generally slightly lower.

[0080] When faced with potential local network obstruction, the communication and data management module enables multi-hop self-organizing network relay functionality. When the quality of the direct communication link between a fire hydrant node and the backbone network falls below a threshold, its communication module automatically searches for other nearby smart fire hydrant nodes and relays data via device-to-device communication, ensuring reliable transmission of alarm information and control commands in complex environments.

[0081] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0082] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. It should be noted that 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. An intelligent hydrant fire extinguishing system comprising a hydrant tapping into a water supply network, characterized in that Also comprising: a perception module for collecting fire hydrant state parameters, environmental fire parameters, and water supply network pressure parameters; a decision control module in communication with the perception module, for receiving and fusing the multi-source heterogeneous data collected by the perception module, and generating control instructions based on a pre-set intelligent decision model; an execution module in communication with the decision control module, for receiving control instructions and executing opening and closing control of the fire hydrant valve and water supply flow regulation; a communication and data management module connected with the perception module, the decision control module and the execution module, for realizing data interaction and instruction transmission among the modules and data synchronization with the external fire command center; a self-diagnosis and health management module integrated in the decision control module, for periodic state evaluation and fault warning of each component of the system under non-fire alarm state.

2. An intelligent hydrant fire extinguishing system according to claim 1, characterized in that The perception module includes a fire hydrant state sensor group, an environmental fire sensor group, and a pipe network pressure sensor group; the fire hydrant state sensor group at least includes an opening degree sensor installed at the fire hydrant valve, a flow sensor installed at the fire hydrant outlet, and a water pressure sensor installed inside the fire hydrant; the environmental fire sensor group at least includes temperature sensors, smoke sensors, and infrared thermal imaging sensors distributed in the fire hydrant protective cover and multiple monitoring points in the adjacent area; the pipe network pressure sensor group is installed on the municipal water supply main pipe connected with the fire hydrant, for monitoring the pressure stability of the water supply source.

3. An intelligent hydrant fire extinguishing system according to claim 2, wherein The decision control module includes a data fusion unit, a risk assessment unit, and an instruction generation unit; the data fusion unit receives the original data stream from the perception module, and uses a Kalman filter and adaptive weighted fusion algorithm to perform time-space alignment and noise reduction processing on the multi-source data, generating a fusion feature vector under a unified timestamp; The risk assessment unit is built-in with a fire risk assessment model, which takes the aforementioned fusion feature vector as input and outputs a quantified fire risk level .

4. An intelligent hydrant fire extinguishing system according to claim 3, wherein, The instruction generating unit is connected with the risk assessment unit and the self-diagnosis and health management module. A hierarchical response strategy library is pre-stored in the instruction generating unit. The strategy library maps the fire risk level to a specific control instruction set, which at least includes a valve target opening degree, a target water supply flow and an expected duration. Meanwhile, the instruction generating unit needs to call the system health state index provided by the self-diagnosis and health management module before generating the final instruction. When the system health state index is lower than a preset threshold, the instruction generating unit will automatically downgrade the response strategy or attach device state alarm information in the instruction. When the system health state index is lower than a preset threshold, the instruction generating unit will automatically downgrade the response strategy or attach device state alarm information in the instruction.​ 5. An intelligent hydrant fire extinguishing system according to claim 4, wherein, The self-diagnosis and health management module includes a periodic self-check unit and a fault prediction unit; the periodic self-check unit triggers at a pre-set period, calibrates and diagnoses the signal drift and zero error of each sensor in the perception module, tests the response delay and torque output of the valve motor in the execution module, and evaluates the link quality and data packet loss rate of the communication module; the fault prediction unit uses a long short-term memory network to build a time series prediction model based on historical self-check data and operation logs, predicts the remaining useful life and potential fault occurrence probability of each key component, and generates a warning signal when the probability exceeds a set threshold.

6. A control method of an intelligent fire hydrant fire extinguishing system, characterized by, An intelligent fire hydrant fire extinguishing system applied to any one of the above claims 1-5, the specific steps are as follows: S1, through the perception module deployed in the fire hydrant and the surrounding environment, synchronously collecting the valve opening, water flow, and internal water pressure data of the fire hydrant, collecting the environmental temperature, smoke concentration, and infrared thermal imaging data of multiple monitoring points, and collecting the inlet pressure data of the water supply pipe network; S2, the decision control module receives the multi-source heterogeneous data collected in step S1, uses the data fusion unit to perform time-space alignment, noise filtering, and feature extraction on the data, and generates a fusion feature vector representing the current monitoring scene; S3, the risk assessment unit of the decision control module inputs the fused feature vector into the pre-trained fire risk assessment model to calculate a quantified fire risk level ; S4, the instruction generating unit of the decision control module generates instructions according to the fire risk level The hierarchical response strategy library is queried to preliminarily determine the target control parameter, and the real-time system health state index provided by the self-diagnosis and health management module is acquired ; S5, the instruction generating unit determines the target control parameter based on the system health index performing adaptive correction on the initially determined target control parameter; If above the health threshold, then the original policy is executed; If If the health threshold is exceeded, the water supply is increased proportionally or the duration of the water supply is extended. If If the safety bottom line is lower than the safety threshold, a device failure alarm is generated and manual intervention is requested, and the automatic start command is blocked. S6, the final control instruction generated in step S5 is issued to the execution module through the communication module, the execution module drives the hydrant valve motor to adjust the valve to the target opening, and the water supply is realized according to the demand; S7, during the fire extinguishing operation, steps S1 to S6 are continuously executed to form a closed loop control circuit, and the control instruction is dynamically adjusted according to the real-time feedback of the fire risk level and the change of the pipe network pressure, until the fire risk level is reduced to below the safety threshold, and then the smooth closing program is executed.

7. A control method for an intelligent hydrant fire extinguishing system according to claim 6, characterized in that, In step S2, the data fusion unit adopts an adaptive weighted fusion algorithm to assign dynamic weights to sensor data of different sources and different types; For environmental fire sensor data, its weight is determined by the distance between the sensor and the suspected fire point and the sensor's own confidence Dynamic adjustment of the weight The calculation method is: wherein is the distance attenuation coefficient; For hydrant status data, its weight positively correlates with the device health subscore; For pipe network pressure data, its weight Automatically promoted when pipe network pressure fluctuation exceeds 2 times the historical standard deviation.

8. A control method for a smart hydrant fire extinguishing system according to claim 7, characterized in that, In step S3, the fire risk assessment model is a hybrid model based on gradient boosting decision tree and logistic regression; the gradient boosting decision tree is used to mine high-order nonlinear interaction features from the fused feature vector, and the logistic regression model is used to fit the final risk probability using the interaction features and the original linear features; In the training stage, the model uses historical fire case data and corresponding sensor data for supervised learning, and determines the optimal hyperparameters through cross-validation.

9. A control method for an intelligent hydrant fire extinguishing system according to claim 8, characterized in that, Fire risk rating The calculation process is characterized by the following formula: wherein, , , respectively represent the normalized values of the temperature, smoke concentration and thermal radiation intensity of the first monitoring point, is a nonlinear feature extraction function based on multilayer perception mechanism, is the spatial weight coefficient of the corresponding monitoring point, and are the real-time flow and valve opening of the current fire hydrant, is a linear function representing the contribution of equipment state to risk, and are the balance coefficients for adjusting the weights of environmental risk and equipment state risk, and the value range is , , and satisfies .

10. A control method for an intelligent hydrant fire extinguishing system according to claim 9, characterized in that, In step S4, the hierarchical response strategy library contains at least three response levels; First level response corresponds The command is to open the valve to 100% of its maximum opening and supply water at the design maximum flow rate, while activating the adjacent hydrant coordinated supply plan. Second level response corresponds The command is to open the valve to 60-80% of maximum opening and supply water at medium flow. Third level response corresponds The command is to open the valve to less than 30% of the maximum opening for a trial spray or to remain in a pre-alarm standby state.

Images