Intelligent fire-extinguishing water taking bolt remote locking system and method thereof

Abstract

The invention relates to the technical field of intelligent fire-extinguishing water taking bolts, and discloses an intelligent fire-extinguishing water taking bolt remote locking system and method. The system comprises a state monitoring module, an abnormity identification module, a risk assessment module, a dynamic locking module and a hierarchical execution module. The state monitoring module collects real-time working state data of the water taking hydrant through a multi-source sensing device, the anomaly recognition module generates an anomaly probability distribution diagram based on a random forest algorithm, and the risk assessment module outputs a safety risk level and a locking priority sequence through a space-time correlation model. The dynamic locking module constructs a multi-constraint optimization model and adopts a chaos particle swarm algorithm for optimization, and the hierarchical execution module realizes distributed execution of locking instructions through a hierarchical communication architecture. According to the system, the state of the water taking hydrant can be monitored in real time, abnormity can be accurately recognized, risks can be scientifically evaluated, remote intelligent locking is achieved, the urban fire protection and water supply guarantee capacity is improved, and smart city construction is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fire - fighting water hydrants, and specifically to an intelligent fire - fighting water hydrant remote locking system and its method. Background Art

[0002] In the modern urban fire - fighting and water supply guarantee system, as a key infrastructure, the reliable operation state and safety of fire - fighting water hydrants are of vital importance. Traditional fire - fighting water hydrants usually lack effective real - time monitoring and remote control means, and there are many problems to be solved urgently.

[0003] In terms of status monitoring, previous water hydrants could only rely on manual regular inspections to understand their working conditions. This method is not only inefficient but also unable to detect faults or abnormalities that occur during the interval between two inspections in a timely manner. For example, an abnormal decrease in water pressure may be due to pipeline leakage, but manual inspections cannot detect it in real time. Once a fire breaks out, it will seriously affect the supply of fire - fighting water, delay the best time for extinguishing the fire, and cause immeasurable losses. At the same time, manual inspections are also easily affected by factors such as weather and human negligence, making it difficult to ensure the accuracy and comprehensiveness of detection.

[0004] In terms of abnormal handling, when problems occur with water hydrants, due to the lack of an automated abnormal recognition and risk assessment mechanism, it is often difficult for staff to accurately judge the severity of the problem and the emergency handling sequence in a timely manner. For some complex faults, it may take a lot of time to troubleshoot and analyze, resulting in a lag in problem handling. Moreover, existing water hydrants cannot be remotely locked after detecting abnormalities and can only rely on manual on - site operations to close the valves. This poses safety risks to staff in some emergency situations (such as abnormalities occurring in dangerous areas) and will further delay the handling of abnormal situations.

[0005] From the perspective of resource management, traditional water hydrant systems are unable to reasonably allocate resources when multiple water hydrants have abnormalities or different usage requirements simultaneously. For example, during peak water - usage periods, the over - use of some water hydrants may lead to insufficient local water pressure, affecting the normal water supply and fire - fighting emergency needs in other areas. However, due to the lack of an effective resource coordination mechanism, it is difficult to reasonably regulate the use of each water hydrant, resulting in resource waste and low utilization efficiency.

[0006] In addition, with the continuous expansion of urban scale and the demand for intelligent development, the traditional management mode of water hydrants has been difficult to meet the requirements of smart city construction. The digital transformation of cities requires the intelligent upgrade of various infrastructure facilities to achieve real - time data collection, transmission, and analysis for better urban management and services. The lag of traditional fire - fighting water hydrants in this regard restricts the improvement of the overall intelligent level of the city. Summary of the Invention

[0007] The object of the present invention is to provide an intelligent fire extinguishing and water intake hydrant remote locking system and its method to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: An intelligent fire extinguishing and water intake hydrant remote locking system, the system includes: Status monitoring module: used to collect real-time working state data of the hydrant through a multi-source sensing device, the real-time working state data includes water pressure, flow rate, valve opening degree and environmental humidity; Abnormality identification module: based on the random forest algorithm, perform multi-dimensional feature classification on the real-time working state data to generate an abnormality probability distribution map; Risk assessment module: input the abnormality probability distribution map into a pre-trained spatio-temporal association model, and output the safety risk level and locking priority sequence of the hydrant; Dynamic locking module: construct a multi-constraint optimization model according to the locking priority sequence, the multi-constraint optimization model takes minimizing the locking response time and balancing resource occupation as the optimization objectives, and uses the chaotic particle swarm algorithm to globally optimize the locking instruction parameters; Hierarchical execution module: based on the multi-constraint optimization model, output the optimal locking strategy, and realize the distributed execution of the locking instruction through a hierarchical communication architecture, the hierarchical communication architecture includes a central control layer, a regional coordination layer and a terminal execution layer, wherein the central control layer generates a global locking sequence based on the locking instruction, the regional coordination layer dynamically adapts the local locking parameters by using a sliding window optimization algorithm, and the terminal execution layer realizes the matching of the valve opening degree and the locking state based on the fuzzy logic control algorithm.

[0009] Preferably, the multi-source sensing device includes a pressure sensor, a flow meter, an angle encoder, a humidity sensor and an infrared detector; the collection of the real-time working state data of the hydrant through the multi-source sensing device includes: Perform spatio-temporal calibration on the pressure sensor data and the flow meter data to construct a three-dimensional water pressure distribution model; perform multi-scale wavelet packet decomposition on the angle encoder data and the humidity sensor data to generate a device action feature sequence; Construct a dual-path feature fusion network, the first path uses a sparse autoencoder to extract the spatial distribution features of the three-dimensional water pressure distribution model, and the second path uses a gated recurrent network to extract the temporal correlation features of the device action feature sequence; Fuse the spatial distribution features and the temporal correlation features through a cross-channel attention mechanism to generate a joint feature tensor; perform state alignment modeling on the joint feature tensor based on the dynamic time warping algorithm, and output a comprehensive state feature including water pressure fluctuation, valve action trajectory and environmental interference factors.

[0010] Preferably, the spatio-temporal correlation model adopts a multi-layer graph neural network structure to generate multiple targets for the locking priority sequence based on an adaptive weight allocation mechanism; the multi-layer graph neural network structure includes: Construct a water hydrant - pipe network interaction graph, where the nodes in the graph include water hydrant nodes, pipe network nodes, environmental nodes, and user nodes, and the node attributes include water pressure values, flow values, and locking request frequencies; Adopt a two-level attention mechanism. The first level calculates the interaction correlation degree between water hydrant nodes and adjacent nodes through a spatial graph attention layer, and the second level conducts key screening on historical locking records through a temporal graph attention layer; Iteratively optimize the node features based on a multi-head graph convolution module. Each convolution head fuses node attributes and external environmental variables; stabilize the training process through skip connections and layer normalization mechanisms, and finally output locking instructions including pipe network pressure constraints and user priority rules.

[0011] Preferably, the chaotic particle swarm optimization algorithm integrating dynamic inertia weight and constraint relaxation strategy includes: Model the locking parameter optimization problem as a multi-constraint mixed integer nonlinear programming problem, where the decision variables include discrete locking switch variables and continuous response time adjustment variables; Initialize the chaotic mapping sub-problem and calculate the non-dominated solution set, and adopt a dynamic inertia weight mechanism to update the objective function weight coefficient according to the real-time pipe network pressure; In the decomposition stage, divide the global problem into multiple single-constraint sub-problems based on the constraint space decomposition technology; in the coordination stage, introduce a relaxation factor to balance the conflicting constraints between sub-problems; Adopt a parallel differential evolution algorithm to iteratively optimize the sub-problems, update the local solution each time, and synchronize the global solution through a coordination factor.

[0012] Preferably, the sliding window optimization algorithm dynamically adapts the local locking parameters, including: Construct a time-varying prediction model, discretize the dynamic response equation of the water hydrant into a state transition model, and the state transition model includes water pressure prediction error, valve response delay, and locking trigger threshold; Design a sliding window optimization objective function, including a response time fluctuation suppression term, a resource occupancy balance term, and a user priority matching term.

[0013] Preferably, the terminal execution layer realizes the matching of valve opening and locking state based on a fuzzy logic control algorithm, including: Design a two-layer fuzzy rule base to map the valve opening error and locking state deviation to fuzzy activation degrees; construct an adaptive membership function adjustment mechanism to dynamically adjust the coverage interval of the fuzzy set according to the deviation change rate.

[0014] Preferably, the sparse autoencoder adopts a multi-scale pyramid structure to accelerate feature extraction, including: dividing the three-dimensional water pressure distribution model into multi-level grids, storing the water pressure gradient and flow rate change rate statistics in each grid cell, using dilated convolutional kernels to expand the receptive field in the encoding stage, and restoring the spatial resolution through transposed convolutional layers in the decoding stage.

[0015] Preferably, the spatial graph attention layer adopts a relative distance encoding mechanism, including: Defining the relative distance vector between the fire hydrant node and adjacent nodes, including the pipe network path length, water pressure transmission efficiency, and valve response time difference; Mapping the relative distance vector to the bias parameter of the graph attention kernel through a non-linear transformation layer; Superimposing the bias parameter onto the standard graph attention operation.

[0016] Preferably, the dynamic inertia weight mechanism is implemented based on an online learning strategy, including: collecting the objective function value and constraint violation degree during the historical optimization process as training samples, constructing a Gaussian process regression model to fit the mapping relationship between the inertia weight and the system state, and triggering the weight emergency reset operation when a sudden change in the pipe network pressure is detected.

[0017] Preferably, the present invention further includes a method applied to the intelligent fire hydrant remote locking system described above, and the method includes the following steps: Step 1: Collecting the real-time working state data of the fire hydrant through the multi-source sensing device of the state monitoring module, where the real-time working state data includes water pressure, flow rate, valve opening, and environmental humidity; Step 2: Using the anomaly recognition module to perform multi-dimensional feature classification on the real-time working state data based on the random forest algorithm to generate an anomaly probability distribution map; Step 3: Inputting the anomaly probability distribution map into the pre-trained spatio-temporal correlation model of the risk assessment module to output the fire hydrant safety risk level and the locking priority sequence; Step 4: The dynamic locking module constructs a multi-constraint optimization model according to the locking priority sequence, where the multi-constraint optimization model takes the minimization of the locking response time and the balance of resource occupancy as the optimization objectives, and uses the chaotic particle swarm algorithm to globally optimize the locking instruction parameters; Step 5: The hierarchical execution module outputs the optimal locking strategy based on the multi-constraint optimization model, and realizes the distributed execution of the locking instruction through a hierarchical communication architecture, where the hierarchical communication architecture includes a central control layer, a regional coordination layer, and a terminal execution layer. Among them, the central control layer generates a global locking sequence based on the locking instruction, the regional coordination layer dynamically adapts the local locking parameters using a sliding window optimization algorithm, and the terminal execution layer realizes the matching of the valve opening and the locking state based on the fuzzy logic control algorithm.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: In terms of real-time monitoring and precise perception, the system can collect real-time working state data such as water pressure, flow rate, valve opening, and environmental humidity through multi-source sensing devices in the state monitoring module, including pressure sensors, flow meters, angle encoders, humidity sensors, and infrared detectors. Moreover, the data of the pressure sensor and the flow meter are calibrated in space and time to construct a three-dimensional water pressure distribution model; the data of the angle encoder and the humidity sensor are decomposed by multi-scale wavelet packet decomposition to generate an equipment action feature sequence. On this basis, through a dual-path feature fusion network and a cross-channel attention mechanism, a comprehensive state feature containing various key information is generated, realizing a comprehensive and precise perception of the working state of the water hydrant. This greatly improves the real-time performance and accuracy of monitoring compared with traditional manual inspections, can promptly detect any subtle abnormal changes, and provides reliable data support for subsequent abnormal handling and risk prevention and control.

[0019] In terms of the scientific nature of abnormal identification and risk assessment, the abnormal identification module classifies multi-dimensional features of the real-time working state data based on the random forest algorithm to generate an abnormal probability distribution map. This method can effectively mine potential patterns and features in the data and accurately identify various abnormal situations. The risk assessment module inputs the abnormal probability distribution map into a pre-trained spatio-temporal correlation model. This model adopts a multi-layer graph neural network structure and an adaptive weight allocation mechanism, considering the attributes and correlation relationships of various nodes in the water hydrant-pipeline network interaction graph, as well as factors such as historical locking records, and outputs the accurate safety risk level and locking priority sequence of the water hydrant. This enables the staff to clearly understand the risk status of each water hydrant, reasonably arrange the processing order according to the priority, avoids blindness in the processing process, and improves the efficiency and pertinence of abnormal handling.

[0020] In terms of remote locking and optimization control, the dynamic locking module constructs a multi-constraint optimization model according to the locking priority sequence, with the optimization goals of minimizing the locking response time and balancing resource occupancy, and uses the chaotic particle swarm algorithm to globally optimize the locking instruction parameters. This optimization strategy ensures that under different constraint conditions, the locking instruction parameters can be quickly and reasonably determined. The hierarchical execution module realizes the distributed execution of the locking instruction through a hierarchical communication architecture. The central control layer generates a global locking sequence, the regional coordination layer dynamically adapts the local locking parameters using a sliding window optimization algorithm, and the terminal execution layer realizes the matching of the valve opening and the locking state based on the fuzzy logic control algorithm. The whole process realizes the intelligence and precision of remote locking, not only greatly shortening the locking response time, but also enabling the balanced utilization of resources in complex pipeline network environments and multi-water hydrant collaborative working scenarios, improving the stability and reliability of the system.

[0021] From the perspective of resource management and overall coordination, the system can real-time monitor the working status of water hydrants and resource usage, and rationally allocate resources through a multi-constraint optimization model and dynamic adaptation algorithm. For example, during peak water usage periods, according to the actual demands and water pressure conditions of each region, the system can intelligently regulate the use of water hydrants to avoid situations such as insufficient local water pressure or over-occupation of resources, ensuring the stable operation of the entire water supply and fire protection system. At the same time, this system is easily integrated with other management systems of the smart city, providing strong support for the digital management of the city and promoting the process of urban intelligent development.

[0022] In terms of safety guarantee and emergency handling capabilities, when an abnormality occurs in the water hydrant, the system can quickly conduct a risk assessment and remotely lock it according to the priority, avoiding the safety risks brought by on-site manual operations. Especially when an abnormality occurs in a dangerous area, this remote locking function is particularly important. In addition, the system's rapid response and precise control can ensure the stable supply of fire-fighting water in case of emergencies such as fires, improving the city's fire emergency handling capabilities and effectively guaranteeing the safety of people's lives and property. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is the working principle diagram of an intelligent fire-extinguishing water hydrant remote locking system according to the present invention; Figure 2 is the flowchart of data collection and processing of a multi-source sensing device; Figure 3 is the flowchart of optimization by the chaotic particle swarm algorithm; Figure 4 is the working principle diagram of the dynamic inertia weight mechanism. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] Please refer to Figures 1-4 , the present invention provides a technical solution: an intelligent fire-extinguishing water hydrant remote locking system, the system includes: Working process of the status monitoring module: The status monitoring module collects real-time working status data of the water hydrant through a multi-source sensing device. This data includes water pressure, flow rate, valve opening, and environmental humidity. The multi-source sensing device is a key component for the entire system to obtain basic information. Among them, the pressure sensor is responsible for accurately measuring the water pressure inside the water hydrant, the flow meter is used to monitor the flow rate of the water flow, the angle encoder can accurately obtain the valve opening information, and the humidity sensor senses the environmental humidity around the water hydrant. Each sensor works in coordination to ensure the comprehensiveness and accuracy of the collected data.

[0026] Working process of the anomaly identification module: After obtaining the real-time working status data, the anomaly identification module performs multi-dimensional feature classification on this data based on the random forest algorithm. The random forest algorithm constructs multiple decision trees and combines the classification results of these decision trees to generate an anomaly probability distribution map. This distribution map intuitively shows the possibility of anomalies occurring in the water hydrant under different states, providing an important basis for subsequent risk assessment.

[0027] Working process of the risk assessment module: The risk assessment module inputs the anomaly probability distribution map generated by the anomaly identification module into a pre-trained spatio-temporal correlation model. The spatio-temporal correlation model is trained with a large amount of historical data and can fully mine the spatio-temporal features and correlation relationships in the data. Through the operation of this model, the safety risk level of the water hydrant and the locked priority sequence are output, thereby clarifying the risk degree of each water hydrant and the order of locking operations that need to be carried out.

[0028] Working process of the dynamic locking module: The dynamic locking module constructs a multi-constraint optimization model based on the locked priority sequence output by the risk assessment module. This model aims to minimize the locking response time and balance resource occupancy, and uses the chaotic particle swarm optimization algorithm to globally optimize the locked instruction parameters. In actual application scenarios, different water hydrants may be in different environmental conditions and working states. The multi-constraint optimization model can comprehensively consider various factors, such as network transmission delay, device performance limitations, etc., and use the chaotic particle swarm optimization algorithm to find the optimal locked instruction parameters in the complex solution space, ensuring that the system can meet the locking requirements while maximizing the system's response speed and resource utilization efficiency.

[0029] Workflow of the hierarchical execution module: Based on the optimal locking strategy output by the multi-constraint optimization model, the hierarchical execution module realizes the distributed execution of locking instructions through the hierarchical communication architecture. The hierarchical communication architecture includes a central control layer, a regional coordination layer, and a terminal execution layer. The central control layer generates a global locking sequence based on the locking instructions, and conducts macroscopic planning and scheduling for the locking operations of the entire system; the regional coordination layer uses a sliding window optimization algorithm to dynamically adapt the local locking parameters, and flexibly adjusts the locking parameters according to the actual situations of different regions to improve the accuracy and adaptability of the locking operations; the terminal execution layer realizes the matching of the valve opening and the locking state based on the fuzzy logic control algorithm, ensuring that the valve can accurately respond during the execution of the locking operation and achieving the expected locking effect.

[0030] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: This embodiment focuses on elaborating the specific process of the multi-source sensing device collecting data and generating comprehensive state features. As the source for the system to obtain information, the accuracy and effectiveness of the data collected by the multi-source sensing device directly affect the working performance of subsequent modules. In actual application scenarios, such as the urban fire protection pipe network system, the hydrants are widely distributed and the surrounding environments are complex and diverse, so it is crucial to accurately obtain their working state data.

[0031] The multi-source sensing device includes a pressure sensor, a flowmeter, an angle encoder, a humidity sensor, and an infrared detector. When collecting data, first perform spatio-temporal calibration on the data of the pressure sensor and the flowmeter. This is because the installation positions and measurement times of different sensors may vary, and spatio-temporal calibration can eliminate these differences and construct a three-dimensional water pressure distribution model, making the water pressure data consistent and comparable in the spatial and temporal dimensions. Assume that the water pressure data collected by the pressure sensor is , where represents time; the flow data collected by the flowmeter is . Through spatio-temporal calibration, the and data at different times and positions are integrated, and using the spatial coordinate information, a three-dimensional water pressure distribution model is constructed, where respectively represent spatial coordinates, represents time, and this model can intuitively reflect the water pressure distribution of the hydrant at different spatial positions and time points.

[0032] Perform multi-scale wavelet packet decomposition on the data of the angle encoder and the humidity sensor. The valve opening data measured by the angle encoder is denoted as , and the environmental humidity data collected by the humidity sensor is denoted as . Multi-scale wavelet packet decomposition is a signal processing technique that can analyze signals at different frequencies and time scales. By performing multi-scale wavelet packet decomposition on and , a device action feature sequence is generated, where represents the sequence index after decomposition. The decomposition process can be expressed by the following formula: .

[0033] where, is the weight coefficient, is the th wavelet packet decomposition function, and is the number of decomposition layers.

[0034] Construct a dual-path feature fusion network. The first path uses a sparse autoencoder to extract the spatial distribution features of the three-dimensional water pressure distribution model. A sparse autoencoder is a neural network structure that can reduce the dimensionality of data while retaining the main features of the data, extracting more representative spatial distribution features. Assume the input of the sparse autoencoder is , and the output is the spatial distribution feature , where represents the feature dimension. The second path uses a gated recurrent network to extract the temporal correlation features of the device action feature sequence. A gated recurrent network can effectively process time series data and capture the temporal information in the data. Let the input of the gated recurrent network be , and the output is the temporal correlation feature , where represents the feature dimension.

[0035] Fuse the spatial distribution features and the temporal correlation features through a cross-channel attention mechanism to generate a joint feature tensor , where and represent the two dimensions of the joint feature tensor respectively. The cross-channel attention mechanism can adaptively assign weights according to the importance of different features, enhancing the expression ability of key features. Based on the dynamic time warping algorithm, perform state alignment modeling on the joint feature tensor, and output a comprehensive state feature that includes water pressure fluctuations, valve action trajectories, and environmental interference factors. The dynamic time warping algorithm can find the best matching path between time series of different lengths, thereby achieving state alignment and accurately reflecting the comprehensive working state of the water hydrant.

[0036] Example 2: This example focuses on the multi-layer graph neural network structure and working principle of the spatio-temporal correlation model. The spatio-temporal correlation model plays a key role in risk assessment in the entire system, and its performance directly affects the accuracy of judging the safety risk level and locking priority sequence of the hydrants. In a large-scale fire protection pipe network system, there are complex interaction relationships between hydrants, pipe networks, the environment, and users. The spatio-temporal correlation model needs to effectively capture these relationships to provide a reliable basis for subsequent locking decisions.

[0037] The spatio-temporal correlation model adopts a multi-layer graph neural network structure. First, a hydrant-pipe network interaction graph is constructed. The nodes in the graph include hydrant nodes, pipe network nodes, environmental nodes, and user nodes. Each node has specific attributes. The attributes of the hydrant node include the water pressure value , flow value , the attributes of the pipe network node include the pipe network pressure , flow transmission capacity , the attributes of the environmental node include environmental temperature , humidity , etc., and the attributes of the user node include the locking request frequency . These attributes are important bases for the model to analyze and make decisions.

[0038] A two-level attention mechanism is adopted. The first level calculates the interaction correlation degree between the hydrant node and adjacent nodes through the spatial graph attention layer. In the spatial graph attention layer, a relative distance encoding mechanism is adopted. Define the relative distance vector between the hydrant node and adjacent nodes, which includes the pipe network path length , water pressure transmission efficiency , and valve response time difference . The relative distance vector is mapped to the bias parameter of the graph attention kernel through a non-linear transformation layer. The formula is: .

[0039] Among them, represents the non-linear transformation function. The bias parameter is superimposed on the standard graph attention operation, so as to enhance the feature expression of adjacent nodes closely related to the hydrant node and highlight the spatial interaction relationship between nodes.

[0040] The second level conducts key screening on historical locking records through the temporal graph attention layer. Historical locking records include locking time , locking reason , etc. The temporal graph attention layer can, based on these historical information, dig out key records that have an important impact on the current risk assessment, and improve the model's processing ability for time series data and the accuracy of risk judgment.

[0041] Iteratively optimize the node features based on the multi-head graph convolutional module. Each convolutional head fuses the node attributes and external environmental variables, such as the change rate of pipe network pressure , the changing trend of user demand , etc. Stabilize the training process through skip connections and layer normalization mechanisms to prevent problems such as gradient vanishing or gradient explosion in the model. Finally, the spatio-temporal correlation model outputs locking instructions that include pipe network pressure constraints and user priority rules, ensuring that the locking decision takes into account both the operational safety of the pipe network and the actual needs of users.

[0042] Example 3: This example details the application of the chaotic particle swarm algorithm integrating dynamic inertia weight and constraint relaxation strategy in the dynamic locking module. The dynamic locking module is a key link to achieve precise locking operations. The chaotic particle swarm algorithm can find the optimal locking instruction parameters under complex multi-constraint conditions, improving the system's response speed and resource utilization efficiency. In the actual management of fire hydrants, different hydrants may face different working conditions and constraint limitations, such as network latency, equipment performance differences, etc. The chaotic particle swarm algorithm needs to adapt to these complex situations to achieve efficient locking control.

[0043] Model the locking parameter optimization problem as a multi-constraint mixed integer nonlinear programming problem. The decision variables include discrete locking switch variables ( , indicating unlocked, indicating locked) and continuous response time adjustment variables . The locking switch variable determines whether the hydrant needs to be locked, and the response time adjustment variable is used to adjust the execution time of the locking operation to meet the goal of minimizing the locking response time.

[0044] Initialize the chaotic mapping sub-problem and calculate the non-dominated solution set. The chaotic mapping is a mathematical model with randomness and ergodicity, which can help the algorithm search more widely in the solution space. Assume using the Logistic chaotic mapping, and the formula is: .

[0045] Where, is the value of the chaotic variable at the -th iteration, is the control parameter, usually taking . Generate initial solutions through chaotic mapping and calculate the non-dominated solution set of these solutions under multi-constraints. The non-dominated solution set contains a group of solutions that cannot be compared with each other in terms of superiority or inferiority between different objectives, providing a basis for subsequent optimization.

[0046] The dynamic inertia weight mechanism is adopted to update the weight coefficient of the objective function according to the real-time pipe network pressure. The dynamic inertia weight mechanism is implemented based on the online learning strategy. The objective function values and constraint violation degrees in the historical optimization process are collected as training samples to construct a Gaussian process regression model to fit the inertia weight and the mapping relationship with the system state (such as the pipe network pressure ). When a sudden change in the pipe network pressure is detected, a weight emergency reset operation is triggered to ensure that the algorithm can quickly adapt to the change in the pipe network pressure. Let the Gaussian process regression model be: .

[0047] Among them, is the Gaussian process regression function, is the noise term.

[0048] In the decomposition stage, the global problem is divided into multiple single-constraint sub-problems based on the constraint space decomposition technology. In this way, complex multi-constraint problems can be simplified into multiple relatively simple single-constraint problems for solution, reducing the computational complexity. In the coordination stage, a relaxation factor is introduced to balance the conflicting constraints between sub-problems. The value range of the relaxation factor is usually between , and by adjusting the value of , a balance can be found between different constraint conditions.

[0049] The parallel differential evolution algorithm is used to iteratively optimize the sub-problems. Each iteration updates the local solution and synchronizes the global solution through the coordination factor . The parallel differential evolution algorithm can utilize parallel computing of multiple processors to improve the optimization efficiency of the algorithm. The coordination factor is used to control the influence degree of the local solution on the global solution, ensuring that the algorithm maintains a balance between local search and global search, and finally finding the locked instruction parameters that meet multiple constraint conditions and have the optimal optimization objective.

[0050] Example 4: This example focuses on elaborating the process of the sliding window optimization algorithm for dynamically adapting local locked parameters at the regional coordination layer. As a key link in the hierarchical communication architecture, the regional coordination layer needs to flexibly adjust local locked parameters according to the actual situations of different regions to improve the accuracy and adaptability of the locking operation. In an actual fire protection pipe network system, the water intake hydrants in different regions may face different working environments and requirements. The sliding window optimization algorithm can perceive these differences in real time and make corresponding adjustments.

[0051] Build a time-varying prediction model, and discretize the dynamic response equation of the water intake hydrant into a state transition model. The dynamic response of the water intake hydrant is affected by various factors, such as water pressure changes, valve actions, etc. Assume that the water pressure prediction error of the water intake hydrant is , the valve response delay is , and the lock trigger threshold is . The state transition model can be expressed as: .

[0052] Among them, is the state vector of the system at moment, including information such as water pressure and valve opening; and are the state transition matrix and the control matrix; is the control input, such as the lock command; is the process noise. Through the discretization of the dynamic response equation of the hydrant, this state transition model is obtained, which can more accurately describe the state changes of the hydrant at different times.

[0053] Design a sliding window optimization objective function, which includes a response time fluctuation suppression term, a resource occupancy balance term, and a user priority matching term. The response time fluctuation suppression term aims to reduce the response time fluctuation of the locking operation and improve the stability of the system. Let the response time be , the response time fluctuation suppression term can be expressed as: .

[0054] Among them, is the response time of the th locking operation, is the average response time, is the number of locking operations. The resource occupancy balance term is used to balance the resource consumption of the system during the locking operation and ensure the reasonable allocation of resources. The user priority matching term adjusts the locking parameters according to the different needs and priorities of users to give priority to meeting the needs of high-priority users. By comprehensively considering these three terms, the sliding window optimization algorithm can find a balance between different objectives, generate the optimal local locking parameters, realize the dynamic adaptation of local locking parameters, and improve the working efficiency of the regional coordination layer and the accuracy of the locking operation.

[0055] Example 5: This example details the specific method for the terminal execution layer to implement the matching of valve opening and locking state based on the fuzzy logic control algorithm. The terminal execution layer is the final execution link of the lock command, and its accuracy directly affects the effect of the locking operation. In practical applications, the matching of valve opening and locking state may be interfered by various factors, such as equipment aging, environmental changes, etc. The fuzzy logic control algorithm can effectively handle these uncertainties and achieve precise matching control.

[0056] Design a two-layer fuzzy rule base, and use the valve opening error and the locking state deviation Mapped to fuzzy activation degree. Valve opening error is the actual valve opening and the target valve opening The difference, that is . Locking state deviation is the actual locking state and the target locking state The difference, when indicates that the locking state is correct, indicates that there is a deviation. The fuzzy rule base contains a series of fuzzy rules. For example: If the valve opening error is positive and large, and the locking state deviation is negative and small, then the fuzzy activation degree is medium. These rules are formulated based on practical experience and system characteristics, and can convert precise input quantities into fuzzy control signals.

[0057] Build an adaptive membership function adjustment mechanism to dynamically adjust the coverage interval of the fuzzy set according to the deviation change rate. The deviation change rate is used to reflect the change trend of the valve opening error and the locking state deviation. Let the valve opening error change rate be , and the locking state deviation change rate be . When is large, it indicates that the valve opening changes rapidly. At this time, the coverage interval of the fuzzy set can be appropriately expanded to handle the change of the valve opening more flexibly; when is small, it indicates that the locking state is relatively stable, and the coverage interval of the fuzzy set can be reduced to improve the control accuracy. Through this adaptive adjustment mechanism, the fuzzy logic control algorithm can better adapt to different working conditions and change situations, achieve an accurate match between the valve opening and the locking state, ensure that the terminal execution layer accurately executes the locking instruction, and improve the reliability and stability of the entire system.

[0058] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0059] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent fire hydrant remote locking system, characterized in that: include: Status monitoring module: used to collect real-time working status data of the water tap through a multi-source sensor device, and the real-time working status data includes water pressure, flow, valve opening and ambient humidity; Abnormal identification module: Based on the random forest algorithm, multi-dimensional feature classification is performed on the real-time working status data to generate an abnormal probability distribution diagram; Risk assessment module: input the abnormal probability distribution map into the pre-trained spatiotemporal correlation model, and output the water tap safety risk level and locking priority sequence; Dynamic locking module: constructing a multi-constraint optimization model according to the locking priority sequence, wherein the multi-constraint optimization model takes minimization of locking response time and balancing of resource occupancy as optimization goals, and adopts a chaotic particle swarm algorithm to globally optimize the locking instruction parameters; Hierarchical execution module: outputs the optimal locking strategy based on the multi-constraint optimization model, and realizes the distributed execution of locking instructions through a hierarchical communication architecture. The hierarchical communication architecture includes a central control layer, a regional coordination layer and a terminal execution layer. The central control layer generates a global locking sequence based on the locking instruction, the regional coordination layer uses a sliding window optimization algorithm to dynamically adapt local locking parameters, and the terminal execution layer realizes the matching of valve opening and locking state based on a fuzzy logic control algorithm.

2. The intelligent fire hydrant remote locking system according to claim 1 is characterized in that: The multi-source sensing device includes a pressure sensor, a flow meter, an angle encoder, a humidity sensor and an infrared detector; The real-time working status data of the water tap is collected by the multi-source sensor device, including: Perform spatiotemporal calibration on the pressure sensor data and flow meter data to build a three-dimensional water pressure distribution model; perform multi-scale wavelet packet decomposition on the angle encoder data and humidity sensor data to generate a device action feature sequence; Constructing a dual-path feature fusion network, wherein the first path uses a sparse autoencoder to extract the spatial distribution features of the three-dimensional water pressure distribution model, and the second path uses a gated recurrent network to extract the temporal correlation features of the device action feature sequence; The spatial distribution features and temporal correlation features are fused through a cross-channel attention mechanism to generate a joint feature tensor. The joint feature tensor is state-aligned and modeled based on a dynamic time warping algorithm to output comprehensive state features including water pressure fluctuations, valve movement trajectories, and environmental interference factors.

3. The intelligent fire hydrant remote locking system according to claim 1 is characterized in that: The spatiotemporal correlation model adopts a multi-layer graph neural network structure and generates multiple targets for the locking priority sequence based on an adaptive weight allocation mechanism; The multi-layer graph neural network structure includes: Construct a water tap-pipeline network interaction graph, where the nodes include water tap nodes, pipeline network nodes, environment nodes, and user nodes. The node attributes include water pressure value, flow value, and lock request frequency. A two-level attention mechanism is adopted. The first level calculates the interactive correlation between the tap node and its adjacent nodes through the spatial graph attention layer, and the second level performs critical screening of historical locking records through the temporal graph attention layer. Based on the multi-head graph convolution module, the node features are iteratively optimized. Each convolution head fuses the node attributes and external environment variables. The training process is stabilized by the jump connection and layer normalization mechanism, and finally the locking instructions containing the pipeline pressure constraints and user priority rules are output.

4. The intelligent fire hydrant remote locking system according to claim 1 is characterized in that: The chaotic particle swarm algorithm specifically includes: The locking parameter optimization problem is modeled as a multi-constrained mixed integer nonlinear programming problem, and the decision variables include discrete locking switch variables and continuous response time adjustment variables. Initialize the chaotic mapping subproblem and calculate the non-dominated solution set, and use the dynamic inertia weight mechanism to update the objective function weight coefficient according to the real-time pipe network pressure; In the decomposition phase, the global problem is divided into multiple single-constraint sub-problems based on the constraint space decomposition technology; in the coordination phase, relaxation factors are introduced to balance the conflicting constraints between sub-problems; A parallel differential evolution algorithm is used to iteratively optimize the subproblems. In each iteration, the local solution is updated and the global solution is synchronized through the coordination factor.

5. The intelligent fire hydrant remote locking system according to claim 1 is characterized in that: The sliding window optimization algorithm dynamically adapts the local locking parameters, including: A time-varying prediction model is constructed to discretize the dynamic response equation of the water tap into a state transition model, wherein the state transition model includes water pressure prediction error, valve response delay, and lock trigger threshold; Design a sliding window optimization objective function, which includes response time fluctuation suppression items, resource occupancy balancing items, and user priority matching items.

6. The intelligent fire hydrant remote locking system according to claim 1 is characterized in that: The terminal execution layer realizes the matching of valve opening and locking state based on fuzzy logic control algorithm, including: A double-layer fuzzy rule base is designed to map the valve opening error and the locking state deviation into fuzzy activation degree; an adaptive membership function adjustment mechanism is constructed to dynamically adjust the coverage interval of the fuzzy set according to the deviation change rate.

7. The intelligent fire hydrant remote locking system according to claim 2 is characterized in that: The sparse autoencoder adopts a multi-scale pyramid structure to accelerate feature extraction, including: dividing the three-dimensional water pressure distribution model into multi-level grids, each grid unit stores water pressure gradient and flow rate change rate statistics, using dilated convolution kernels to expand the receptive field in the encoding stage, and restoring spatial resolution through transposed convolution layers in the decoding stage.

8. The intelligent fire hydrant remote locking system according to claim 3 is characterized in that: The spatial graph attention layer adopts a relative distance encoding mechanism, including: Define the relative distance vector between the water tap node and the adjacent nodes, including the pipe network path length, water pressure transmission efficiency and valve response time difference; The relative distance vector is mapped to the bias parameter of the graph attention kernel through a nonlinear transformation layer; The bias parameters are added to the standard graph attention operation.

9. The intelligent fire hydrant remote locking system according to claim 4 is characterized in that: The dynamic inertia weight mechanism is implemented based on an online learning strategy, including: collecting the objective function value and constraint violation degree in the historical optimization process as training samples, constructing a Gaussian process regression model to fit the mapping relationship between the inertia weight and the system state, and triggering an emergency weight reset operation when a sudden change in the pipeline network pressure is detected.

10. A method for applying the intelligent fire hydrant remote locking system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: Collecting real-time working status data of the water tap through the multi-source sensing device of the status monitoring module, wherein the real-time working status data includes water pressure, flow, valve opening and ambient humidity; Step 2: Using an abnormality recognition module to perform multi-dimensional feature classification on the real-time working status data based on a random forest algorithm to generate an abnormality probability distribution diagram; Step 3: Input the abnormal probability distribution map into the pre-trained spatiotemporal correlation model of the risk assessment module, and output the water tap safety risk level and locking priority sequence; Step 4: The dynamic locking module constructs a multi-constraint optimization model according to the locking priority sequence. The multi-constraint optimization model takes minimization of locking response time and balancing of resource occupancy as optimization goals, and adopts a chaotic particle swarm algorithm to globally optimize the locking instruction parameters; Step 5: The hierarchical execution module outputs the optimal locking strategy based on the multi-constraint optimization model, and realizes the distributed execution of the locking instruction through the hierarchical communication architecture. The hierarchical communication architecture includes a central control layer, a regional coordination layer and a terminal execution layer. The central control layer generates a global locking sequence based on the locking instruction, the regional coordination layer uses a sliding window optimization algorithm to dynamically adapt the local locking parameters, and the terminal execution layer realizes the matching of the valve opening and the locking state based on the fuzzy logic control algorithm.

Images