Gas safety emergency management dispatching command method and system

Abstract

The present application relates to the technical field of smart city safety management and Internet of Things control, in particular to a gas safety emergency management dispatching command method and system, comprising the following steps: S100, acquiring real-time sensing data of a gas pipe network and space-time data of an external environment, and performing space mapping of a pipe network topology and a city road network to construct a heterogeneous space-time graph network; wherein the real-time sensing data comprises pipe network pressure, instantaneous flow and node position, and the space-time data of the external environment comprises meteorological parameters; S200, when a mutation rate of the pipe network pressure or the instantaneous flow exceeds a preset threshold, establishing a leakage source node, and combining a preset fluid mechanics model, a physical pipe diameter and a current pipe network pressure to calculate an initial leakage rate of the leakage source node. The present application effectively solves the technical problems of physical prediction distortion of a traditional system, lack of safety constraints in path finding and lack of a closed loop in control, and improves safety and scientificity of gas accident disposal under complex working conditions.

Description

Technical Field

[0001] This invention relates to the fields of smart city safety management and Internet of Things control technology, specifically to a gas safety emergency management dispatch and command method and system. Background Technology

[0002] With the advancement of urbanization, the coverage and complexity of urban gas pipeline networks continue to grow. In the event of sudden accidents such as gas pipeline ruptures and leaks, rapid and scientific emergency dispatch and equipment control are crucial to ensuring urban public safety. Currently, most gas emergency dispatch systems rely on static Geographic Information Systems (GIS) and pre-set business rules for early warning and dispatching. However, when dealing with complex dynamic physical evolution and high-concurrency emergency resource dispatch, they suffer from the following core technical deficiencies: Existing technologies cannot achieve deep coupling between the evolution of the underlying fluid physical field and the upper-level heterogeneous spatial scheduling network. In terms of physical evolution prediction, traditional systems typically treat the leak point as a static parameter source, using a fixed leak rate and an empirical buffer radius based on a two-dimensional plane to delineate the danger zone. They fail to combine the real-time pipeline pressure drop gradient to perform inversion calculations of the leak orifice diameter and dynamic source terms, and also fail to consider the mechanical turbulence blocking effect of real three-dimensional urban building complexes as physical obstacles on airflow. This results in severe boundary distortion in the calculated spatial envelope of the explosion limit concentration.

[0003] During scheduling optimization and control execution, the lack of a cross-mapping mechanism between multiphysics and graph data structures prevents existing systems from converting the dynamic secondary risk zone coordinates output by the fluid dynamics model into modification instructions for the topology weights of the underlying traffic map network. This results in the generated emergency repair paths easily intersecting with real hazardous plume entities. Simultaneously, traditional valve remote control logic typically outputs instantaneous full-closing electrical signals, completely ignoring the sudden increase in transient flux caused by the abrupt reduction in cross-sectional area of ​​the fluid within high-pressure pipelines. This can easily induce gas hammer effects in the pipeline network and cause mechanical tearing of upstream pipe sections. Furthermore, the existing scheduling process is a unidirectional open-loop control, lacking edge-side visual perception and real-time data feedback for newly added physical obstacles and disaster characteristic parameters. The scheduling algorithm cannot perform secondary convergence and clock-rolling optimization of the objective function based on changes in physical boundary conditions, resulting in emergency scheduling schemes lacking physical safety constraints and dynamic adaptive control capabilities in complex and ever-changing accident scenes. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a gas safety emergency management and dispatching method, comprising the following steps: S100, acquiring real-time sensing data of the gas pipeline network and spatiotemporal data of the external environment, and spatially mapping the pipeline network topology with the urban road network to construct a heterogeneous spatiotemporal graph network; wherein, the real-time sensing data includes pipeline pressure, instantaneous flow rate and node location, and the spatiotemporal data of the external environment includes meteorological parameters; S200, when the rate of change of the pipeline pressure or instantaneous flow exceeds a preset threshold, a leakage source node is identified, and the initial leakage rate of the leakage source node is calculated by combining the preset fluid dynamics model, physical pipe diameter and current pipeline pressure. S300, taking the initial leakage rate as the source term, combined with the meteorological parameters and the convection-diffusion physical field equation, calculates the dynamic gas concentration distribution under the three-dimensional spatial grid, and extracts the spatial envelope of the concentration in the explosion limit range to generate a high-risk secondary risk zone. S400, on the heterogeneous spatiotemporal graph network, with the joint optimization objectives of minimizing valve cut-off time and minimizing emergency repair path time, an optimal valve cut-off sequence is generated, and an optimal emergency repair route that avoids the high-risk secondary risk area is generated. S500 converts the optimal valve cutoff sequence into equipment control commands and sends them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions. At the same time, the optimal emergency repair route is sent to the matching field terminal for closed-loop scheduling and command.

[0005] Furthermore, the spatiotemporal data of the external environment also includes mobile signal population density heat map data and real-time urban road congestion index data; The acquisition of real-time sensing data of the gas pipeline network specifically involves: acquiring the device's underlying readings at a preset sampling frequency through a group of IoT sensing nodes deployed on the gas pipeline network, and then synchronizing the data to the system control terminal after protocol parsing and timing marking by the edge gateway. The spatial mapping of the pipeline network topology to the urban road network to construct a heterogeneous spatiotemporal graph network is achieved in the following manner: Define the graph network model as follows: ,in For a set of nodes, Let be the set of edges. The time slice set is used; the pipeline intersections, equipment valves, and sensor coordinates of the gas pipeline network are extracted as pipeline entity nodes, and the intersections and emergency resource outposts of the urban road network are extracted as traffic entity nodes. The pipeline entity nodes and the traffic entity nodes are merged to form the node set. ; Pipeline topology edges are constructed based on the physical pipeline connections between the pipeline entity nodes, and traffic road network edges are constructed based on the passable road attributes between the traffic entity nodes, thus generating the edge set. ; The real-time sensing data and the spatiotemporal data of the external environment are timestamped according to a preset time slice granularity, and then assigned as multi-dimensional feature vectors to the corresponding node sets. With the set of edges middle; In the process of aligning the real-time sensing data with the spatiotemporal data of the external environment according to a preset time slice granularity, a spatiotemporal multi-source data fusion algorithm based on Kalman filtering is introduced to perform noise reduction and smooth frame interpolation processing on multi-source heterogeneous data with different sampling frequencies.

[0006] Furthermore, the step of identifying the leak source node when the rate of change in the pipeline pressure or instantaneous flow exceeds a preset threshold specifically includes: The instantaneous pressure drop rate is obtained by continuously calculating the time derivative of the pipeline pressure within two adjacent time slices. When the instantaneous pressure drop rate is detected to be greater than the preset pressure drop threshold, and the instantaneous flow surge of the upstream pipeline node is greater than the preset flow surge threshold within the same time window, a physical anomaly alarm is triggered. Based on the location of the sensing device that triggered the physical anomaly alarm, combined with the attenuation curve characteristics of the fluid pressure gradient along the pipeline and the pipeline network topology, the specific physical pipe segment that was damaged was located in reverse spatial positioning, and the coordinate point of the pipe segment mapped in the heterogeneous spatiotemporal graph network was marked as the leakage source node. The initial leakage rate of the leakage source node is calculated by combining a preset fluid dynamics model, physical pipe diameter, and current pipeline pressure. The solution is then obtained by solving the fluid dynamics equations established as follows: in, To solve for the obtained initial leakage rate, The gas emission coefficient is used to characterize the resistance at the edge of a pipeline rupture. The equivalent leakage orifice area at the point of physical pipe damage. For the fluid density used to transport the gas, The current pipeline pressure is obtained through the real-time sensing data. The ambient atmospheric pressure.

[0007] Furthermore, the equivalent leakage orifice area due to the initial stage of pipe rupture... For implicit physical quantities that cannot be directly measured, dynamic estimation is performed by introducing a pressure gradient inversion method: A feature mapping matrix representing the pressure relief gradient of different physical pipe diameters under different damage scales is pre-established in the cloud; The instantaneous pressure drop rate, acquired in real time, is input into the pressure relief gradient feature mapping matrix for nonlinear interpolation matching to estimate the equivalent leakage orifice area at the current moment. ; Using the estimated equivalent leakage orifice area By continuously substituting into the orifice-dissolving fluid dynamics equation, a rolling correction calculation for the initial leakage rate is achieved.

[0008] Furthermore, the calculation of the dynamic gas concentration distribution under a three-dimensional spatial grid includes the following preliminary steps: three-dimensional mesh generation and boundary condition setting: Using the three-dimensional geographic coordinates of the leak source node as the origin, a three-dimensional Euclidean spatial grid containing horizontal, vertical and vertical elevations is constructed by expanding outwards according to a preset spatial resolution. Read the real-time wind speed and direction data from the meteorological parameters and convert them into convective boundary conditions in the three-dimensional flow field; at the same time, retrieve the three-dimensional digital model of the city buildings around the pipeline network topology and extract the geometric contours of the buildings as physical obstacle boundary conditions that are impervious to fluid. The solution process for the convective-diffusion physical field equations is as follows: The initial leakage rate is treated as a continuous point source emission term, and a three-dimensional Gaussian plume diffusion operator considering terrain roughness is introduced for spatiotemporal evolution solution: in, Spatial coordinates ( )exist The volume percentage concentration of the gas at any given time. The initial leakage rate, This refers to the real-time downwind wind speed mapped according to the aforementioned convective boundary conditions. These are dynamically corrected based on atmospheric stability and the boundary conditions of the physical obstacles, respectively. Axial turbulent diffusion coefficient.

[0009] Furthermore, the extraction of a spatial envelope with a concentration within the explosive limit range to generate a high-risk secondary disaster zone is achieved through the following data processing method: Extract the lower explosive limit (LEL) and upper explosive limit (UEL) concentration thresholds from the physical properties of the gas. Traverse all grid nodes in the three-dimensional spatial grid and use a threshold slicing algorithm to extract those that satisfy the following conditions: The set of grid coordinate points for the decision criteria; Perform three-dimensional convex hull calculation and spatial smoothing reconstruction on the set of grid coordinate points to generate a three-dimensional closed hazardous plume entity; The three-dimensional closed dangerous plume entity is vertically projected onto the two-dimensional geographic plane where the heterogeneous spatiotemporal map network is located to generate a polygonal dangerous area, and this polygonal dangerous area is highlighted and topologically isolated as the high-risk secondary risk area.

[0010] Furthermore, the joint optimization objective of minimizing valve shut-off time and minimizing emergency repair path time is specifically manifested in constructing the following multi-objective cost function solution model in the heterogeneous spatiotemporal graph network: in, To provide a comprehensive emergency cost index, The total time for upstream valve communication and mechanical closure, obtained through spatiotemporal graph network topology search, is... The set of nodes covered by the high-risk secondary risk zone. For nodes The risk equivalent, For nodes Population density thermal value, This is the set of road segments on the generated emergency repair route. For road sections dynamically calculated based on real-time urban road congestion index Travel time These are the preset decision weight coefficients; The optimal emergency travel route that avoids the high-risk secondary risk area is generated, and the preceding graph network weight update strategy for path planning includes: Extract the spatial geometric boundary of the high-risk secondary risk area; Spatial intersection tests are performed in the heterogeneous spatiotemporal graph network to identify traffic entity nodes and traffic network edges that spatially intersect with the spatial geometric boundary. The identified traffic entity nodes are marked as unreachable, and the passage cost weight of the corresponding traffic network edge is dynamically modified to infinity, thereby logically severing the connectivity of crossing high-risk areas in the underlying graph data structure. Subsequently, a heuristic pathfinding algorithm is run on the heterogeneous spatiotemporal graph network with updated weights to generate the optimal rapid progress route; In the process of generating the optimal emergency travel route using the multi-objective cost function solution model, a multi-dimensional emergency resource constraint matrix is ​​also introduced to perform hard boundary filtering of candidate solutions: Extract the pipeline physical attributes of the leak source node to generate an accident feature vector containing the leaking pipe diameter and operating pressure level; Extract the attribute information of each candidate emergency repair resource and generate a resource feature vector that includes the specifications of onboard spare parts and the special operation qualifications of personnel; Calculate the inner product matching degree between the accident feature vector and the resource feature vector of each candidate resource, and directly remove the candidate emergency repair resources whose inner product matching degree does not reach the full rank state from the search space of the heterogeneous spatiotemporal graph network.

[0011] Furthermore, the process of converting the optimal valve cutoff sequence into equipment control commands and sending them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions is achieved in the following manner: The optimal valve cutoff sequence is mapped into a time sequence with control delay according to the order of action. The device control commands are sent asynchronously in encrypted message form to the remote terminal unit (RTU) or programmable logic controller (PLC) where the corresponding control valve is located via the industrial Internet of Things (IoT) communication protocol. The RTU or PLC analyzes the control delay and drives the electric actuator of the physical control valve to perform a closing action at a preset step-by-step speed.

[0012] Furthermore, the optimal emergency repair route is sent to the matched field terminal for closed-loop scheduling and command, specifically including: The optimal emergency repair route is integrated with a pre-stored standardized emergency repair digital operation SOP package and pushed to the augmented reality (AR) smart terminal worn by on-site repair personnel for three-dimensional path navigation and visual step guidance. The AR smart terminal uses its image acquisition module to capture streaming video of the scene environment, and runs a target recognition algorithm on the edge computing chip of the AR smart terminal to extract the disaster situation feature parameters and the geometric coordinates of obstacles. The extracted on-site disaster characteristic parameters and the geometric coordinates of obstacles are fed back to the control terminal in real time via wireless network; The closed-loop dispatching and command also includes a rolling dynamic optimization step based on on-site feedback: The geometric coordinates of the obstacles fed back by the AR smart terminal are input into the heterogeneous spatiotemporal graph network to dynamically correct the local traffic weight matrix of the corresponding traffic entity nodes and traffic network edges. The on-site disaster characteristic parameters fed back are input into the convection-diffusion physical field equation to perform boundary condition re-convergence calculation and range reconstruction on the boundary of the high-risk secondary risk zone. The multi-objective cost function solution model is called again for secondary optimization until the physical control valve is completely closed and the real-time sensing data returns to the safety baseline.

[0013] On the other hand, a gas safety emergency management and dispatching system includes: The graph construction module is used to acquire real-time sensing data of the gas pipeline network and spatiotemporal data of the external environment, and to spatially map the pipeline network topology with the urban road network to construct a heterogeneous spatiotemporal graph network; wherein, the real-time sensing data includes pipeline pressure, instantaneous flow rate and node location, and the spatiotemporal data of the external environment includes meteorological parameters; The physical calculation module is used to identify the leakage source node when the sudden change rate of the pipeline pressure or instantaneous flow exceeds a preset threshold, and calculate the initial leakage rate of the leakage source node by combining the preset fluid dynamics model, physical pipe diameter and current pipeline pressure. The diffusion simulation module is used to take the initial leakage rate as the source term, combine the meteorological parameters and the convection-diffusion physical field equations, calculate the dynamic gas concentration distribution under a three-dimensional spatial grid, and extract the spatial envelope of the concentration in the explosion limit range to generate a high-risk secondary risk zone. The scheduling decision module is used to generate the optimal valve cut-off sequence and the optimal emergency repair route that avoids the high-risk secondary risk area on the heterogeneous spatiotemporal graph network, with the joint optimization objectives of minimizing valve cut-off time and minimizing emergency repair path time. The control execution module is used to convert the optimal valve cutoff sequence into equipment control commands and send them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions. At the same time, the optimal emergency repair route is sent to the matching field terminal for closed-loop scheduling and command.

[0014] Beneficial effects This invention utilizes pressure gradient inversion to calculate the dynamic initial leakage rate and, combined with three-dimensional building obstacle boundary conditions and a Gaussian diffusion operator, accurately extracts the explosion limit spatial envelope that conforms to real fluid mechanics. Simultaneously, it maps this physical risk boundary across dimensions to a heterogeneous spatiotemporal graph network. By modifying the local traffic weight matrix of the bottom-level nodes and traffic network edges, it logically severs the connectivity of the hazardous area from a data structure perspective. Combined with a multi-objective cost function, it generates the optimal emergency repair route and valve shut-off sequence. Finally, at the execution end, a step-by-step speed control strategy is introduced to eliminate the gas hammer effect in the pipeline caused by fluid shut-off. Edge computing terminals collect on-site coordinates in real time to feed back to the control end for reconvergence and rolling reconstruction optimization of local boundary conditions. This effectively solves the technical problems of distorted physical predictions, lack of safety constraints in pathfinding, and lack of closed-loop control in traditional systems, improving the safety and scientific rigor of gas accident handling under complex conditions. Attached Figure Description

[0015] Figure 1 This is a flowchart of the present invention; Figure 2 This is a system overall architecture block diagram of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] The present invention will now be described in further detail with reference to the accompanying drawings: Example: like Figures 1-2 As shown, a gas safety emergency management and dispatching method includes the following steps: S100 acquires real-time sensing data of the gas pipeline network and spatiotemporal data of the external environment, and spatially maps the pipeline network topology with the urban road network to construct a heterogeneous spatiotemporal graph network; among which, the real-time sensing data includes pipeline pressure, instantaneous flow rate and node location, and the spatiotemporal data of the external environment includes meteorological parameters. S200: When the rate of change of pipeline pressure or instantaneous flow exceeds a preset threshold, the leakage source node is identified, and the initial leakage rate of the leakage source node is calculated by combining the preset fluid dynamics model, physical pipe diameter and current pipeline pressure. S300 uses the initial leakage rate as the source term, combines meteorological parameters and convection-diffusion physical field equations to calculate the dynamic gas concentration distribution under a three-dimensional spatial grid, and extracts the spatial envelope of the concentration in the explosion limit range to generate a high-risk secondary risk zone. S400, on a heterogeneous spatiotemporal graph network, generates the optimal valve cut-off sequence and the optimal emergency repair route that avoids high-risk secondary risk areas by taking the joint optimization objective of minimizing valve cut-off time and minimizing emergency repair path time as the joint optimization objectives. The S500 converts the optimal valve shut-off sequence into equipment control commands and sends them to the edge nodes of the pipeline network, driving the physical control valves to perform automatic shut-off actions. At the same time, it sends the optimal emergency repair route to the matching field terminal for closed-loop scheduling and command.

[0019] Furthermore, the specific implementation process of step S100 is as follows: During system startup, the mapping module executes a multi-dimensional heterogeneous spatiotemporal data acquisition and pipeline digital twin mapping construction process. First, data acquisition begins, with the IoT sensing node group deployed at key control and monitoring points in the gas pipeline network commencing operation. This IoT sensing node group includes pressure sensors, instantaneous flow meters, and laser methane detectors. The pressure sensors continuously acquire scalar pressure data inside the pipeline at a fixed sampling frequency of 100Hz, while the instantaneous flow meters continuously acquire instantaneous volumetric flow rate data of the fluid within the pipe at a sampling frequency of 10Hz. The physical readings acquired by each sensor are transmitted to the field edge gateway via a hardware communication link based on an RS485 bus or industrial Ethernet.

[0020] The edge gateway is internally configured with a hardware microprocessor and protocol parsing firmware. This firmware calls the Modbus-TCP or MQTT protocol stack to perform low-level decoding of the received raw current signals or digital messages, converting them into pressure and flow values ​​in standard International Units (SI). After decoding, the edge gateway's clock chip writes a UNIX standard timestamp with millisecond-level precision to the header of the data packet, completing the time-series marking. The time-stamped structured sensing data is then synchronously sent in real-time to the central server at the system control end via a high-concurrency Kafka message queue.

[0021] In parallel with the acquisition of pipeline sensing data, the system control terminal synchronously acquires spatiotemporal data of the external environment from an external data platform via a network communication interface. Specifically, every 30 seconds, the system control terminal retrieves physical meteorological parameters output by the micro-weather station through a streaming media interface. These meteorological parameters explicitly include real-time wind speed vector values, wind direction angle values, atmospheric temperature scalar values, and relative humidity percentages. Simultaneously, the system control terminal connects with the base station signaling data server of the mobile communication operator to obtain mobile signal population density heat map data within a preset spatial grid. The population density heat map data is updated at a frequency of one frame every 5 minutes. The system control terminal also receives real-time urban road congestion index data at a 1-minute cycle through the traffic data bus of the smart city operation center. This congestion index data is correlated to specific urban road spatial links.

[0022] Because IoT sensing data and external environmental spatiotemporal data come from different hardware sources and have sampling frequencies that differ by orders of magnitude, the system control terminal, upon receiving the aforementioned multi-source heterogeneous data, invokes a built-in spatiotemporal multi-source data fusion algorithm based on Kalman filtering for alignment processing. The algorithm constructs a state transition matrix and an observation matrix in memory, using high-frequency sampled instantaneous pressure and flow rate of the pipeline network as core state variables. It utilizes state prediction equations and measurement update equations to dynamically filter and denoise the pressure and flow rate data, eliminating abnormal spikes caused by sensor hardware thermal noise and random jitter in the communication link. For population density heatmap data and real-time urban road congestion index data with lower update frequencies, the algorithm introduces a linear interpolation function and a smoothing frame interpolation operator in the time domain. Using the time axis of pipeline network pressure as a benchmark, it calculates external environmental feature estimates that are precisely aligned with the millisecond-level timestamps of the pipeline network pressure, generating a time-series feature matrix with completely unified timestamps, eliminating time-series feature misalignment caused by communication delays and equipment heterogeneity.

[0023] After data alignment is completed, the graph construction module performs spatial mapping and graph network construction steps in server memory. The system defines the heterogeneous spatiotemporal graph network as a mathematical representation structure. ,in Represents a set of nodes. Represents the set of edges. This represents a set of time slices divided into 5-second increments. The system control terminal reads the topology data of the gas pipeline network from a pre-stored gas 3D GIS database, extracting the 3D geographic information (including longitude, latitude, and elevation) of the physical intersections of pipelines, equipment valve installation locations, and sensor deployment coordinates, and instantiating them as pipeline entity nodes in memory. Simultaneously, the system control terminal reads the topology data of the urban road network from a city traffic network database, extracting the 3D geographic information of road intersections, junctions, and emergency repair team locations, and instantiating them as traffic entity nodes in memory. The system merges the pipeline entity nodes and traffic entity nodes into a unified Euclidean coordinate system, forming a set of nodes in a graph network. .

[0024] Next, the topology relationship establishment process begins. Based on the actual pipeline connection matrix of the gas pipeline network, the system establishes pipeline topology edges between two adjacent pipeline entity nodes that are physically connected by pipes, and writes the physical pipe diameter, pipe material roughness, and elevation difference into the edge's attribute topology table. Simultaneously, based on the passable topology matrix of urban roads, the system establishes traffic network edges between two adjacent traffic entity nodes, and writes the number of lanes, design maximum speed, and physical length of the road segment into the edge's attribute topology table. The pipeline topology edges and traffic network edges together form the edge set of the graph network. .

[0025] Finally, the system maps the aforementioned Kalman-filtered aligned temporal feature matrix into a multidimensional feature vector and binds it to the corresponding graph topology elements. For the node set... For each pipeline entity node, its feature vector is assigned a corresponding real-time pressure value and instantaneous flow rate value; for each traffic entity node, its feature vector is assigned a real-time population density value at the corresponding geographical coordinates; for the edge set... For each pipeline topology edge, its feature vector contains the real-time convection velocity of the fluid within the pipe. For each traffic road network edge, its feature vector contains the corresponding real-time congestion index scalar. These multi-dimensional feature vectors are sequentially filled into the corresponding time slice sets according to the time step order. In the end, a high-dimensional heterogeneous spatiotemporal graph network topology with spatiotemporal evolution characteristics is generated in memory, providing underlying graph data support for subsequent physical triggering and scheduling optimization calculations.

[0026] Furthermore, the specific implementation process of step S200 is as follows: After the graph construction module completes the construction of the heterogeneous spatiotemporal graph network, the physical calculation module cyclically reads the feature vectors of topological elements in each time slice and initiates the leak source location and physical calculation process. The physical calculation module retrieves the pressure data of the pipeline entity nodes in two adjacent time slices and calculates the time first derivative of the pipeline pressure using the finite difference method to obtain the current instantaneous pressure drop rate. The system control terminal compares the instantaneous pressure drop rate with the pre-stored pressure drop threshold in real time. When the instantaneous pressure drop rate of a certain pipeline entity node is detected to be greater than the pressure drop threshold, and within the same set spatiotemporal sliding window, the instantaneous flow surge of the adjacent node in the upstream direction of the pipeline entity node in the graph network is greater than the preset flow surge threshold, the system control terminal immediately determines that a structural physical damage has occurred in the pipeline and triggers a physical anomaly alarm.

[0027] Upon triggering a physical anomaly alarm, the system initiates a spatial positioning procedure. This procedure retrieves real-time pressure scalars from multiple monitoring nodes surrounding the rupture point and constructs a decay curve of the fluid pressure gradient along the pipe axis in the spatial geometric model. Based on the slope variation characteristics of the decay curve, and combined with the topological constraints of the pipeline connection matrix in the heterogeneous spatiotemporal graph network, the system traces the confluence of pressure drops along the pipeline's extension direction, thereby pinpointing the spatial coordinates of the damaged physical pipe segment. The topological nodes mapped to these physical pipe segments in the heterogeneous spatiotemporal graph network are then marked as the leak source nodes.

[0028] After identifying the leak source node, the physical calculation module retrieves a pre-defined fluid dynamics model from its local configuration memory. The core of this fluid dynamics model is based on orifice outflow theory and is used to quantitatively calculate the mass flow rate of gas injected from the high-pressure pipeline network to the atmospheric environment at normal pressure. Specifically, the system calls the following established orifice elimination fluid dynamics equations for solution calculation: In the orifice digestion fluid dynamics equations, the correspondence between various physical quantities and parameters is as follows: The initial leakage rate to be solved is expressed in kilograms per second. The dimensionless gas emission coefficient is used to characterize the degree of obstruction to fluid flow caused by frictional resistance and flow bundle contraction effect at the edge of the pipe rupture. In this embodiment, it is preset to be between 0.60 and 0.65 according to the roughness of the pipe material. The equivalent leakage orifice area at the point of physical pipe rupture, in square meters; The real-time fluid density of the transported gas medium is expressed in kilograms per cubic meter. Its value is corrected in real time based on the current pipeline temperature and component pressure using the equation of state. The current pipeline pressure is collected and aligned in real time by a group of IoT sensing nodes, and the unit is Pascal. The standard atmospheric pressure constant is 101325 Pascals.

[0029] Because the shape of sudden pipeline damage caused by third-party construction or material fatigue is random, the equivalent leakage orifice area... This is an implicit physical quantity that cannot be directly measured by sensors under actual operating conditions. To achieve closed-loop solution of the equations, the physics calculation module introduces a pressure gradient inversion mechanism for dynamic parameter estimation. A pressure relief gradient characteristic mapping matrix, constructed from historical accident simulation data and fluid dynamics simulations, is pre-stored in the cloud database. This matrix establishes a four-dimensional nonlinear mapping relationship between the physical pipe diameter, initial operating pressure of the pipeline, instantaneous pressure drop rate, and equivalent leakage orifice area.

[0030] The physical calculation module extracts the baseline operating pressure and physical pipe diameter of the leak source node within a slice prior to the accident, and combines this with the currently calculated instantaneous pressure drop rate, using these as a multi-dimensional input vector into the pressure relief gradient feature mapping matrix. The algorithm performs high-order bi-trispline interpolation calculations in the local space of the pressure relief gradient feature mapping matrix to inversely estimate the equivalent leak orifice area at the current moment. The value. Complete the equivalent leakage orifice area. After the inversion estimation, the physical calculation module will calculate the equivalent leakage orifice area. The numerical value, the real-time current pipeline pressure and temperature-corrected fluid density Substitute these values ​​into the orifice-dissipating fluid dynamics equations and run the algebraic solver to calculate the current initial leakage rate. In each subsequent time slice Internally, the system continuously triggers the inversion mechanism and algebraic solution process based on the newly acquired pressure difference to determine the initial leakage rate. The high-frequency rolling correction calculation is performed, and the calculated values ​​for each time period are... The numerical value is written as a real-time dynamic source item into the feature vector of the leak source node in the current time slice.

[0031] Furthermore, the specific implementation process of step S300 is as follows: After the physics calculation module outputs the dynamically updated initial leakage rate, the diffusion simulation module uses the initial leakage rate as a continuous point source emission term to start the multiphysics diffusion and risk evolution simulation process.

[0032] The first step is to perform three-dimensional spatial mesh generation and boundary condition setting. The diffusion simulation module uses the three-dimensional geographic coordinates (including longitude, latitude, and absolute elevation) of the leakage source node as the spatial origin, and expands it stepwise to the surrounding geographic space according to the preset spatial step resolution of 0.5 meters, constructing a three-dimensional Euclidean spatial mesh in the computing memory that includes the horizontal X-axis, the vertical Y-axis, and the vertical elevation Z-axis.

[0033] After mesh generation, the system control terminal reads the real-time wind speed vector value and wind direction angle value from the acquired meteorological parameters. These values ​​are then substituted into the fluid dynamics boundary condition function to transform them into the convective boundary conditions required for three-dimensional flow field calculations. Simultaneously, the system control terminal retrieves batch three-dimensional geometric data of buildings surrounding the leakage source node from a locally configured urban three-dimensional building digital model library. It extracts the external geometric contour coordinate set of the buildings, converts this set into physical obstacle boundary conditions that are impermeable to the fluid, and maps it onto the corresponding three-dimensional Euclidean space mesh nodes.

[0034] After completing the mesh construction and boundary condition configuration, the diffusion simulation module introduces a three-dimensional Gaussian plume diffusion operator that considers terrain roughness to solve for the spatiotemporal evolution of the combustion gas in three-dimensional space. The specific solution algorithm uses the following convection-diffusion physical field equations: In the above convection-diffusion physical field equations, Spatial coordinates ( )exist The volume percentage concentration of the gas at any given time. The initial leakage rate, This refers to the real-time downwind wind speed mapped based on convective boundary conditions. These are dynamically corrected based on atmospheric stability and physical obstacle boundary conditions. Axial turbulent diffusion coefficient. When inputting numerical values, the diffusion simulation module determines the basic diffusion coefficient based on the Pasquale atmospheric stability classification standard, and combines this with the aforementioned physical obstacle boundary conditions. Nonlinear dynamic corrections are performed to quantify the mechanical turbulence obstruction effect of physical buildings on gas diffusion airflow.

[0035] By running the above convection-diffusion physical field equations, the diffusion simulation module calculates the concentration value of each grid node in the three-dimensional Euclidean space grid at the current time slice.

[0036] Next, the system performs the hazardous area spatial envelope extraction step. The diffusion simulation module reads the lower explosive limit (LEL) and upper explosive limit (UEL) concentration thresholds of the currently transported gas medium from the material property database. Subsequently, the algorithm traverses all grid nodes in the three-dimensional Euclidean space grid, performs a threshold slicing algorithm, numerically filters the node concentrations, and extracts those that meet the judgment criteria. The set of all grid coordinate points.

[0037] After extracting the set of grid coordinate points that meet the explosion limit determination conditions, the diffusion simulation module calls the three-dimensional convex hull calculation function to perform outer geometric contour calculation on the discrete grid coordinate points, and uses the Laplacian smoothing operator to perform spatial smooth reconstruction on the generated grid surface, generating a three-dimensional hazardous plume entity with a smooth surface and closed internal volume in the three-dimensional coordinate system.

[0038] Finally, the diffusion simulation module uses an orthogonal projection algorithm to project the generated 3D hazardous plume entity downwards along the Z-axis perpendicular to the ground, mapping it onto the 2D geographic plane where the aforementioned heterogeneous spatiotemporal map network is located. This automatically generates a polygonal hazardous area composed of multiple 2D geographic coordinate points connected end-to-end. The system defines the generated polygonal hazardous area as a high-risk secondary risk zone, and directly writes the polygonal boundary coordinate array of the high-risk secondary risk zone into the underlying topology of the heterogeneous spatiotemporal map network for topological isolation marking, and simultaneously highlights it in red on the 2D map layer of the display interface.

[0039] Furthermore, the specific implementation process of step S400 is as follows: After the diffusion simulation module outputs the high-risk secondary risk zone, the scheduling decision module initiates a multi-objective scheduling decision process based on the aforementioned heterogeneous spatiotemporal graph network.

[0040] First, the system performs dynamic updates of the preceding graph network weights and local topology reconstruction for the risk avoidance path. The scheduling decision module extracts the polygonal spatial geometric boundary coordinate array of the high-risk secondary risk area and runs a spatial bounding box intersection test algorithm in the underlying data structure of the heterogeneous spatiotemporal graph network. The algorithm traverses all traffic entity nodes and traffic network edges, identifying traffic entity nodes and traffic network edges that have geometric spatial intersections with the spatial geometric boundary coordinate array. Once a spatial intersection is detected, the scheduling decision module marks the identified traffic entity node as unreachable in the status register of the memory structure and forcibly modifies the passage cost weight value of the corresponding traffic network edge in the topology routing table to infinity. By rewriting the weight value in the underlying memory address, the system logically severs the spatial connectivity of the graph network traversing the explosion hazard area, generating a heterogeneous spatiotemporal graph network with updated weights.

[0041] Next, the scheduling decision module introduces a multi-dimensional emergency resource constraint matrix to filter the optimization search space using hard boundary conditions. The scheduling decision module reads the pipeline physical attribute table corresponding to the leak source node, extracts the leaking pipe diameter parameters and pipeline operating pressure level parameters, and concatenates the aforementioned parameters according to a preset dimensional order to generate an accident feature vector.

[0042] Simultaneously, the scheduling decision module traverses the data ledger of available emergency repair resources in the cloud-based emergency repair resource database, extracting the specifications of spare parts and special operation qualification codes of personnel for each group of candidate repair resources, and generating corresponding resource feature vectors. Subsequently, the system calls the linear algebra calculation library to calculate the inner product matching degree between the accident feature vector and the resource feature vectors of each group of candidate emergency repair resources. If the calculated inner product matrix does not reach full rank, the system determines that the candidate emergency repair resource does not have the hardware physical conditions to handle the current leak condition, and directly removes the traffic entity node corresponding to the candidate emergency repair resource whose inner product matching degree does not reach full rank from the pathfinding search space of the heterogeneous spatiotemporal graph network.

[0043] After completing the graph network topology reconstruction and candidate resource search space elimination, the scheduling decision module constructs a multi-objective cost function solution model in the heterogeneous spatiotemporal graph network. The system uses minimizing the valve cutoff time and minimizing the repair path time as joint optimization objectives, and the specific mathematical function expression is as follows: In the above multi-objective cost function solution model: The comprehensive emergency cost index to be solved; The upstream valve communication delay and total mechanical shut-off time are obtained through spatiotemporal graph network topology search; The set of pipeline network physical nodes covered by high-risk secondary risk areas; For nodes The risk equivalent scalar value; For nodes The linked real-time population density thermal value; This is a collection of traffic segments on the emergency repair route to be generated. For road sections dynamically calculated based on real-time urban road congestion index Travel time; These are the decision weight coefficients that are pre-written into the system configuration file.

[0044] The scheduling decision module invokes a heuristic pathfinding algorithm and a mixed-integer nonlinear programming solver to perform global iterative optimization of the multi-objective cost function solution model on a heterogeneous spatiotemporal graph network with updated weights, until the objective function converges. The solver output is the comprehensive emergency cost index. The system analyzes the solution set to obtain the minimum value, extracts the valve control instruction set to generate the optimal valve cutoff sequence, and extracts the traversal path of traffic entity nodes in the solution set to generate the optimal emergency travel route that avoids high-risk secondary risk areas. The results are then sent to the control execution module.

[0045] Furthermore, the specific implementation process of step S500 is as follows: After the scheduling decision module outputs the optimal valve cutoff sequence and the optimal emergency repair route, the control execution module starts the cloud-edge-device collaborative physical closed-loop control and dynamic feedback program.

[0046] First, the underlying industrial control actions are executed. The control execution module obtains the optimal valve cutoff sequence and maps it to a time sequence with millisecond-level control delays according to the valve topology hierarchy and the order of mechanical actions, generating device control instructions containing the device hardware address. The system control terminal calls the industrial IoT communication protocol stack to encapsulate the device control instructions into asynchronous encrypted messages, which are then sent to the remote terminal unit (RTU) or programmable logic controller (PLC) physically connected to the target physical control valve via a dedicated network channel. The microprocessor of the RTU or PLC receives and decrypts the asynchronous encrypted messages, extracting the control delay parameters and action step size parameters contained in the message protocol payload. Subsequently, the RTU or PLC outputs a pulse width modulation (PWM) control signal to the electric actuator configured on the target physical control valve. After receiving the PWM control signal, the electric actuator drives its internal servo motor to perform a mechanical closing action according to a preset step-by-step speed. The specific control logic is as follows: the servo motor rotates the valve stem in stages according to the preset angular velocity curve, gradually reducing the flow cross-sectional area of ​​the fluid inside the pipeline, thereby limiting the flow rate of the high-pressure gas fluid in a deceleration phase, avoiding the instantaneous change caused by the instantaneous blockage of the fluid inside the pipeline from being converted into an extremely high static pressure peak, and completely eliminating the mechanical tearing damage caused by the gas hammer effect of the pipeline network to the upstream undamaged pipe section.

[0047] In parallel with industrial control actions, the control execution module initiates visual scheduling and edge computing programs for on-site terminals. The system control terminal retrieves standardized emergency repair digital operation SOP packages from the local repository. The control execution module fuses and packages the optimal emergency repair route coordinates with the digital guidance drawings in the SOP package, and distributes this data via wireless broadband to the augmented reality (AR) smart terminals worn by the repair personnel arriving on site. The AR smart terminal's optical display module and built-in 3D rendering engine receive the fused data package, rendering and overlaying a 3D path navigation model and a visual mechanical operation guidance model in real time above the repair personnel's physical field of vision. During the repair personnel's movement and operation phases, the AR smart terminal's image acquisition module continuously captures streaming video sequences of the on-site physical environment. The AR smart terminal is equipped with an independent edge computing chip, and the system calls the convolutional neural network target recognition algorithm deployed on the edge computing chip to perform frame-by-frame pixel-level semantic segmentation and feature extraction on the input streaming video sequence, automatically identifying on-site disaster characteristic parameters that deviate from the normal state and the absolute geometric coordinates of newly added physical obstacles not included in the original topology of the urban road network. AR smart terminals transmit disaster characteristic parameters and absolute geometric coordinates of obstacles extracted by edge computing power back to the system control terminal in the cloud in real time via the uplink communication link of the wireless network.

[0048] Upon receiving the field perception data transmitted from the edge, the control execution module triggers a dynamic closed-loop feedback control mechanism based on clock-based rolling optimization. The control execution module resolves the absolute geometric coordinates of the obstacles transmitted back into spatial vectors, inputs them into the underlying data structure of the aforementioned heterogeneous spatiotemporal graph network, performs a matrix overwrite operation, dynamically modifies the traffic entity nodes corresponding to the spatial coordinates of the obstacles and the local traffic weight matrices of the traffic network edges, resetting the corresponding element values ​​in the weight matrices to infinity. Simultaneously, the control execution module inputs the disaster characteristic parameters transmitted back into the convection-diffusion physical field equations of the diffusion simulation module, performs partial differential parameter correction on the original three-dimensional flow field Gaussian diffusion operator based on real disaster data, performs re-convergence calculations of the hydrodynamic boundary conditions of the high-risk secondary risk zone, and reconstructs the polygonal geometric boundary of the danger zone in a three-dimensional Euclidean space grid. After completing the parameter reconstruction of the underlying graph data structure and the physical diffusion model, the system control end re-calls the aforementioned multi-objective cost function solution model for secondary global extremum optimization, outputting a dynamically updated supplementary valve cutoff sequence and a corrected emergency travel route. The system control unit continuously maintains the aforementioned iterative cycle of hardware closed-loop execution, edge vision feedback, and cloud-based rolling reconstruction until the pipeline pressure and instantaneous flow data fed back by the IoT sensing nodes deployed in the physical pipeline network return to the preset safe baseline value range, and all mechanical limit switches of the physical control valves that issue control commands feed back hard-wired closed status electrical signals to the system control unit. After all the aforementioned termination conditions are met, the system determines that the physical space isolation and emergency repair work are fully completed, and the control execution module ends the closed-loop scheduling and command execution program.

[0049] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A gas safety emergency management and dispatching method, characterized in that, Includes the following steps: S100: Acquire real-time sensing data of the gas pipeline network and spatiotemporal data of the external environment, and spatially map the pipeline network topology with the urban road network to construct a heterogeneous spatiotemporal graph network; wherein, the real-time sensing data includes pipeline pressure, instantaneous flow rate and node location, and the spatiotemporal data of the external environment includes meteorological parameters; S200, when the rate of change of the pipeline pressure or instantaneous flow exceeds a preset threshold, a leakage source node is identified, and the initial leakage rate of the leakage source node is calculated by combining the preset fluid dynamics model, physical pipe diameter and current pipeline pressure. S300, taking the initial leakage rate as the source term, combined with the meteorological parameters and the convection-diffusion physical field equation, calculates the dynamic gas concentration distribution under the three-dimensional spatial grid, and extracts the spatial envelope of the concentration in the explosion limit range to generate a high-risk secondary risk zone. S400, on the heterogeneous spatiotemporal graph network, with the joint optimization objectives of minimizing valve cut-off time and minimizing emergency repair path time, an optimal valve cut-off sequence is generated, and an optimal emergency repair route that avoids the high-risk secondary risk area is generated. S500 converts the optimal valve cutoff sequence into equipment control commands and sends them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions. At the same time, the optimal emergency repair route is sent to the matching field terminal for closed-loop scheduling and command.

2. The gas safety emergency management and dispatching method according to claim 1, characterized in that, The spatiotemporal data of the external environment also includes mobile signal population density heat map data and real-time urban road congestion index data; The acquisition of real-time sensing data of the gas pipeline network specifically involves: acquiring the device's underlying readings at a preset sampling frequency through a group of IoT sensing nodes deployed on the gas pipeline network, and then synchronizing the data to the system control terminal after protocol parsing and timing marking by the edge gateway. The spatial mapping of the pipeline network topology to the urban road network to construct a heterogeneous spatiotemporal graph network is achieved in the following manner: Define the graph network model as follows, where For a set of nodes, Let be the set of edges. The time slice set is used; the pipeline intersections, equipment valves, and sensor coordinates of the gas pipeline network are extracted as pipeline entity nodes, and the intersections and emergency resource outposts of the urban road network are extracted as traffic entity nodes. The pipeline entity nodes and the traffic entity nodes are merged to form the node set. ; Pipeline topology edges are constructed based on the physical pipeline connections between the pipeline entity nodes, and traffic road network edges are constructed based on the passable road attributes between the traffic entity nodes, thus generating the edge set. ; The real-time sensing data and the spatiotemporal data of the external environment are timestamped according to a preset time slice granularity, and then assigned as multi-dimensional feature vectors to the corresponding node sets. With the set of edges middle; In the process of aligning the real-time sensing data with the spatiotemporal data of the external environment according to a preset time slice granularity, a spatiotemporal multi-source data fusion algorithm based on Kalman filtering is introduced to perform noise reduction and smooth frame interpolation processing on multi-source heterogeneous data with different sampling frequencies.

3. The gas safety emergency management and dispatching method according to claim 2, characterized in that, When the rate of change in pipeline pressure or instantaneous flow exceeds a preset threshold, the leak source node is identified, specifically including: The instantaneous pressure drop rate is obtained by continuously calculating the time derivative of the pipeline pressure within two adjacent time slices. When the instantaneous pressure drop rate is detected to be greater than the preset pressure drop threshold, and the instantaneous flow surge of the upstream pipeline node is greater than the preset flow surge threshold within the same time window, a physical anomaly alarm is triggered. Based on the location of the sensing device that triggered the physical anomaly alarm, combined with the attenuation curve characteristics of the fluid pressure gradient along the pipeline and the pipeline network topology, the specific physical pipe segment that was damaged was located in reverse spatial positioning, and the coordinate point of the pipe segment mapped in the heterogeneous spatiotemporal graph network was marked as the leakage source node. The initial leakage rate of the leakage source node is calculated by combining a preset fluid dynamics model, physical pipe diameter, and current pipeline pressure. The solution is then obtained by solving the fluid dynamics equations established as follows: in, The initial leakage rate is obtained by solving for the gas emission coefficient, which characterizes the resistance at the pipeline rupture edge. The equivalent leakage orifice area at the point of physical pipe damage. For the fluid density used to transport the gas, The current pipeline pressure is obtained through the real-time sensing data. The ambient atmospheric pressure.

4. The gas safety emergency management and dispatching method according to claim 3, characterized in that, Due to the equivalent leakage orifice area in the initial stage of pipeline damage. For implicit physical quantities that cannot be directly measured, dynamic estimation is performed by introducing a pressure gradient inversion method: A feature mapping matrix representing the pressure relief gradient of different physical pipe diameters under different damage scales is pre-established in the cloud; The instantaneous pressure drop rate, acquired in real time, is input into the pressure relief gradient feature mapping matrix for nonlinear interpolation matching to estimate the equivalent leakage orifice area at the current moment. ; Using the estimated equivalent leakage orifice area By continuously substituting into the orifice-dissolving fluid dynamics equation, a rolling correction calculation for the initial leakage rate is achieved.

5. A gas safety emergency management and dispatching method according to claim 4, characterized in that, The calculation of dynamic gas concentration distribution in a three-dimensional spatial grid includes the following preliminary steps: three-dimensional grid generation and boundary condition setting. Using the three-dimensional geographic coordinates of the leak source node as the origin, a three-dimensional Euclidean spatial grid containing horizontal, vertical and vertical elevations is constructed by expanding outwards according to a preset spatial resolution. Read the real-time wind speed and direction data from the meteorological parameters and convert them into convective boundary conditions in the three-dimensional flow field; at the same time, retrieve the three-dimensional digital model of the city buildings around the pipeline network topology and extract the geometric contours of the buildings as physical obstacle boundary conditions that are impervious to fluid. The solution process for the convective-diffusion physical field equations is as follows: The initial leakage rate is treated as a continuous point source emission term, and a three-dimensional Gaussian plume diffusion operator considering terrain roughness is introduced for spatiotemporal evolution solution: in, Spatial coordinates ( )exist The volume percentage concentration of the gas at any given time. The initial leakage rate, This refers to the real-time downwind wind speed mapped according to the aforementioned convective boundary conditions. These are dynamically corrected based on atmospheric stability and the boundary conditions of the physical obstacles, respectively. Axial turbulent diffusion coefficient.

6. The gas safety emergency management and dispatching method according to claim 5, characterized in that, The extraction of spatial envelopes with concentrations within the explosive limit range to generate high-risk secondary zones is achieved through the following data processing method: Extract the lower explosive limit (LEL) and upper explosive limit (UEL) concentration thresholds from the physical properties of the gas. Traverse all grid nodes in the three-dimensional spatial grid and use a threshold slicing algorithm to extract those that satisfy the following conditions: The set of grid coordinate points for the decision criteria; Perform three-dimensional convex hull calculation and spatial smoothing reconstruction on the set of grid coordinate points to generate a three-dimensional closed hazardous plume entity; The three-dimensional closed dangerous plume entity is vertically projected onto the two-dimensional geographic plane where the heterogeneous spatiotemporal map network is located to generate a polygonal dangerous area, and this polygonal dangerous area is highlighted and topologically isolated as the high-risk secondary risk area.

7. A gas safety emergency management and dispatching method according to claim 6, characterized in that, The joint optimization objective of minimizing valve shut-off time and minimizing emergency repair path time is specifically manifested in constructing the following multi-objective cost function solution model in the heterogeneous spatiotemporal graph network: in, To provide a comprehensive emergency cost index, The total time for upstream valve communication and mechanical closure, obtained through spatiotemporal graph network topology search, is... The set of nodes covered by the high-risk secondary risk zone. For nodes The risk equivalent, For nodes The population density thermal value represents the set of road segments along the generated emergency travel route. For road sections dynamically calculated based on real-time urban road congestion index Travel time These are the preset decision weight coefficients; The optimal emergency travel route that avoids the high-risk secondary risk area is generated, and the preceding graph network weight update strategy for path planning includes: Extract the spatial geometric boundary of the high-risk secondary risk area; Spatial intersection tests are performed in the heterogeneous spatiotemporal graph network to identify traffic entity nodes and traffic network edges that spatially intersect with the spatial geometric boundary. The identified traffic entity nodes are marked as unreachable, and the passage cost weight of the corresponding traffic network edge is dynamically modified to infinity, thereby logically severing the connectivity of crossing high-risk areas in the underlying graph data structure. Subsequently, a heuristic pathfinding algorithm is run on the heterogeneous spatiotemporal graph network with updated weights to generate the optimal rapid progress route; In the process of generating the optimal emergency travel route using the multi-objective cost function solution model, a multi-dimensional emergency resource constraint matrix is ​​also introduced to perform hard boundary filtering of candidate solutions: Extract the pipeline physical attributes of the leak source node to generate an accident feature vector containing the leaking pipe diameter and operating pressure level; Extract the attribute information of each candidate emergency repair resource and generate a resource feature vector that includes the specifications of onboard spare parts and the special operation qualifications of personnel; Calculate the inner product matching degree between the accident feature vector and the resource feature vector of each candidate resource, and directly remove the candidate emergency repair resources whose inner product matching degree does not reach the full rank state from the search space of the heterogeneous spatiotemporal graph network.

8. A gas safety emergency management and dispatching method according to claim 7, characterized in that, The process of converting the optimal valve cutoff sequence into equipment control commands and sending them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions is achieved in the following manner: The optimal valve cutoff sequence is mapped into a time sequence with control delay according to the order of action. The device control commands are sent asynchronously in encrypted message form to the remote terminal unit (RTU) or programmable logic controller (PLC) where the corresponding control valve is located via the industrial Internet of Things (IoT) communication protocol. The RTU or PLC analyzes the control delay and drives the electric actuator of the physical control valve to perform a closing action at a preset step-by-step speed.

9. A gas safety emergency management and dispatching method according to claim 8, characterized in that, The optimal emergency repair route is sent to the matching field terminal for closed-loop scheduling and command, specifically including: The optimal emergency repair route is integrated with a pre-stored standardized emergency repair digital operation SOP package and pushed to the augmented reality (AR) smart terminal worn by on-site repair personnel for three-dimensional path navigation and visual step guidance. The AR smart terminal uses its image acquisition module to capture streaming video of the scene environment, and runs a target recognition algorithm on the edge computing chip of the AR smart terminal to extract the disaster situation feature parameters and the geometric coordinates of obstacles. The extracted on-site disaster characteristic parameters and the geometric coordinates of obstacles are fed back to the control terminal in real time via wireless network; The closed-loop dispatching and command also includes a rolling dynamic optimization step based on on-site feedback: The geometric coordinates of the obstacles fed back by the AR smart terminal are input into the heterogeneous spatiotemporal graph network to dynamically correct the local traffic weight matrix of the corresponding traffic entity nodes and traffic network edges. The on-site disaster characteristic parameters fed back are input into the convection-diffusion physical field equation to perform boundary condition re-convergence calculation and range reconstruction on the boundary of the high-risk secondary risk zone. The multi-objective cost function solution model is called again for secondary optimization until the physical control valve is completely closed and the real-time sensing data returns to the safety baseline.

10. A gas safety emergency management and dispatching system, applied to the gas safety emergency management and dispatching method described in any one of claims 1-9, characterized in that, The system includes: The graph construction module is used to acquire real-time sensing data of the gas pipeline network and spatiotemporal data of the external environment, and to spatially map the pipeline network topology with the urban road network to construct a heterogeneous spatiotemporal graph network; wherein, the real-time sensing data includes pipeline pressure, instantaneous flow rate and node location, and the spatiotemporal data of the external environment includes meteorological parameters; The physical calculation module is used to identify the leakage source node when the sudden change rate of the pipeline pressure or instantaneous flow exceeds a preset threshold, and calculate the initial leakage rate of the leakage source node by combining the preset fluid dynamics model, physical pipe diameter and current pipeline pressure. The diffusion simulation module is used to take the initial leakage rate as the source term, combine the meteorological parameters and the convection-diffusion physical field equations, calculate the dynamic gas concentration distribution under a three-dimensional spatial grid, and extract the spatial envelope of the concentration in the explosion limit range to generate a high-risk secondary risk zone. The scheduling decision module is used to generate the optimal valve cut-off sequence and the optimal emergency repair route that avoids the high-risk secondary risk area on the heterogeneous spatiotemporal graph network, with the joint optimization objectives of minimizing valve cut-off time and minimizing emergency repair path time. The control execution module is used to convert the optimal valve cutoff sequence into equipment control commands and send them to the edge nodes of the pipeline network to drive the physical control valves to perform automatic cutoff actions. At the same time, the optimal emergency repair route is sent to the matching field terminal for closed-loop scheduling and command.

Images