An abnormality monitoring method and system of an unmanned vehicle gas pipeline network simulation system

Abstract

The application relates to the technical field of computers and discloses an abnormality monitoring method and system for a gas pipeline network simulation system of an unmanned vehicle, which comprises the following steps: a high-fidelity digital twin simulation field is constructed by fusing pipeline network topology, fluid dynamics and real-time environmental parameters; the simulation model is dynamically corrected based on multi-modal sensing data of the unmanned vehicle; the confidence of a candidate leakage source is evaluated through space-time consistency comparison; and a refined detection instruction is generated to guide the closed-loop verification of an unmanned vehicle cluster. The system comprises six modules, namely, pipeline network basic data management, multi-source data fusion sensing, a digital twin simulation engine, leakage source hypothesis evaluation, intelligent task planning and unmanned vehicle cluster scheduling. Through the virtual-real linkage and closed-loop feedback mechanism, the application can significantly improve the leakage positioning accuracy, speed and robustness and reduce the inspection cost.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, specifically relating to an anomaly monitoring method and system for an unmanned vehicle gas pipeline network simulation system. Background Technology

[0002] With the continuous expansion of urban gas pipeline networks and the increasing complexity of their structures, higher demands are placed on the ability to respond quickly and accurately locate leaks. Traditional gas leak monitoring mainly relies on fixed sensor networks to collect concentration data and combines this with threshold alarm mechanisms for anomaly identification. However, in urban environments with high-density building clusters, fluctuating wind directions, and complex terrain, the gas diffusion process exhibits highly nonlinear and time-varying characteristics. Single sensor readings are easily affected by local turbulence, shielding effects, or cross-contamination, making it difficult to accurately reflect the true location of the leak source. Furthermore, the high cost of sensor deployment and numerous blind spots lead to delayed system response and ambiguous location during sudden leak events, severely restricting emergency response efficiency and public safety assurance levels.

[0003] Mobile gas monitoring technology based on unmanned vehicle platforms has attracted widespread attention in recent years. This technology uses autonomous inspection vehicles equipped with high-sensitivity gas sensors to dynamically collect continuous spatiotemporal concentration field data along pipeline networks, providing a new observational dimension for leak source inversion. Its core objective is to integrate mobile sensing data with physical diffusion models to achieve probabilistic inference and dynamic correction of potential leak points. However, existing methods often employ deterministic inversion strategies or simplified diffusion assumptions, failing to fully consider the impact of environmental disturbances, sensor noise, and the uncertainty of unmanned vehicle trajectories on positioning accuracy. This results in significant problems such as high false alarm rates, slow convergence speed, and poor robustness in complex urban scenarios.

[0004] The following problems exist in integrating probabilistic robotics frameworks with physical models of gas diffusion: There is a lack of effective modeling of the autonomous vehicle's own pose uncertainty, leading to biases in the spatial mapping of observational data; gas diffusion simulations typically employ steady-state or ideal boundary conditions, making it difficult to characterize transient transport behavior under urban microclimates. Furthermore, the mining of spatiotemporal concentration data largely remains at the statistical averaging level, without establishing a probabilistic correlation mechanism with leakage source parameters (such as release rate and location). Summary of the Invention

[0005] This invention provides an anomaly monitoring method and system for an unmanned vehicle-based gas pipeline network simulation system, aiming to solve the technical problem that relying solely on sensor readings is insufficient for accurately and quickly locating gas leak sources in complex urban environments. In existing technologies, gas pipeline network leak monitoring mainly relies on fixed sensor networks or manual inspections. The former is limited by deployment density and cost, failing to cover all pipeline sections, while the latter suffers from drawbacks such as response lag, strong subjectivity, and low efficiency. Even with the introduction of mobile unmanned vehicles equipped with gas sensors for dynamic inspections, the collected single-dimensional concentration data is still easily affected by environmental disturbances such as wind direction, temperature, humidity, and building obstructions, leading to distorted concentration gradients and making it difficult to infer the true leak location. Furthermore, traditional methods lack the ability to collaboratively model pipeline topology, fluid dynamics characteristics, and real-time environmental factors, failing to achieve high-precision, low-latency leak source location in dynamically changing urban environments.

[0006] To overcome the aforementioned technical bottlenecks, this invention proposes an anomaly monitoring method and system that integrates multi-source heterogeneous data, constructs a high-fidelity digital twin simulation model, and drives unmanned vehicle collaborative detection based on a closed-loop feedback mechanism. This method goes beyond simple threshold judgment or local gradient analysis of raw sensor readings. Instead, it constructs a virtual simulation environment synchronized in real-time with the physical pipeline network, reproducing the physical process of gas diffusion within it. The simulation model parameters are continuously corrected using measured data transmitted back by unmanned vehicles, thereby efficiently converging the most probable leak source location in the virtual space. Based on this, the optimal detection path is planned, guiding the unmanned vehicle cluster to conduct targeted verification, forming a closed-loop intelligent monitoring system encompassing perception, simulation, decision-making, and verification.

[0007] This invention provides an anomaly monitoring method for an unmanned vehicle gas pipeline network simulation system, comprising: acquiring static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network; real-time acquisition of gas concentration data, three-dimensional wind speed vector data, atmospheric temperature and humidity data, and the unmanned vehicle's own position and attitude data in the environment through a multimodal sensor array deployed on the unmanned vehicle; constructing an initial gas pipeline network fluid dynamics basic model based on the static topology data and historical operating condition data; inputting the real-time data acquired by the multimodal sensor array into the gas pipeline network fluid dynamics basic model to drive its dynamic evolution and generate a gas diffusion digital twin simulation field consistent with the current physical world state; setting multiple candidate leakage source hypothetical points in the gas diffusion digital twin simulation field, and for each Assuming a point, a simulated gas diffusion pattern is generated under current environmental conditions. The simulated gas diffusion pattern is compared with the spatial distribution data of gas concentration actually collected by the unmanned vehicle to calculate the confidence score of each candidate leak source assumption point. Based on the confidence score, the candidate leak source region with the highest confidence is selected, and a detection task instruction for the region is generated based on the spatial coordinate information of the region. The detection task instruction is issued to the unmanned vehicle to guide it into the candidate leak source region for high-density grid sampling to obtain higher resolution local gas concentration field data. The high resolution local gas concentration field data is fed back to the gas diffusion digital twin simulation field to further correct the position, leakage rate, and direction parameters of the candidate leak source until the positioning error converges to within a preset threshold.

[0008] In one embodiment of the present invention, the multimodal sensing array includes a high-precision catalytic combustion gas concentration sensor, an ultrasonic three-dimensional anemometer, a digital temperature and humidity sensor, and a combined navigation unit integrating a global positioning system module and an inertial measurement unit module. The high-precision catalytic combustion gas concentration sensor has a measurement range of 0 to 5% volume concentration and a resolution of 1×10⁻⁶. -5 The response time is less than 3 seconds; the wind speed measurement range of the ultrasonic three-dimensional anemometer is 0 to 60 m / s with an accuracy of ±0.5 m / s, and the wind direction measurement accuracy is ±3°; the position update frequency provided by the integrated navigation unit is not less than 20 Hz, and the horizontal positioning error is less than 1 m.

[0009] As one embodiment of the present invention, the construction of the initial gas pipeline network fluid dynamics basic model specifically includes: parsing the static topology data into a directed graph structure composed of nodes and pipe segments, wherein nodes represent valves, pressure regulating stations or user access points, and pipe segments represent pipelines connecting nodes; assigning corresponding pipe material attribute data to each pipe segment, including inner diameter, wall thickness, and roughness coefficient; determining the pressure and flow boundary conditions of each node under normal operating conditions based on historical operating condition data; discretizing the entire pipeline network using a one-dimensional transient compressible fluid control equation set, wherein the control equation set includes a continuity equation, a momentum equation, and an energy equation; and numerically solving the control equation set using the finite volume method to obtain the initial steady-state distribution of pressure, velocity, and temperature at various points within the pipeline network.

[0010] As one embodiment of the present invention, the process of driving the dynamic evolution to generate a digital twin simulation field of gas diffusion consistent with the current physical world state specifically includes: using real-time atmospheric temperature and humidity data transmitted back by the unmanned vehicle as environmental background field parameters and inputting them into the computational fluid dynamics simulation engine; constructing a dynamic wind field model over the entire simulation area using the three-dimensional wind speed vector data transmitted back by the unmanned vehicle through the Kriging interpolation algorithm; unidirectionally coupling the dynamic wind field model with the gas pipeline network fluid dynamics basic model, i.e., the pipeline network model provides the initial injection velocity and direction of the leak point, while the computational fluid dynamics simulation engine simulates the turbulent diffusion process of leaked gas under the influence of flow around complex urban building clusters; the computational fluid dynamics simulation engine adopts the large eddy simulation method, with a spatial grid resolution of not less than 0.5m in the near-ground region and non-uniform grid refinement in the vertical direction, and a time step set to 0.1s to ensure accurate capture of transient diffusion phenomena.

[0011] As one embodiment of the present invention, the calculation of the confidence score of each candidate leak source hypothesis point specifically includes: defining a spatiotemporal matching function, which calculates the weighted root mean square error between the predicted gas concentration values ​​and the actual measured values ​​in the simulation field at all sampling times and sampling locations of the unmanned vehicle under the candidate leak source hypothesis; the weighting factor of the weighted root mean square error is determined by the Euclidean distance between the unmanned vehicle and the hypothetical leak source, and the closer the distance, the higher the weight; the confidence score is inversely proportional to the weighted root mean square error, and the smaller the error, the higher the score; and all candidate leak source hypothesis points are sorted in descending order according to their confidence scores.

[0012] As one embodiment of the present invention, the generation of detection task instructions for the region specifically includes: defining a circular detection area with a radius of 50m, centered on the geometric center of the candidate leakage source region with the highest confidence level; planning a spiral or serpentine high-density sampling path within the circular detection area, with the path spacing set to 5m; instructing the unmanned vehicle to travel at a constant speed of no more than 5 kilometers per hour along the high-density sampling path, and activating its multimodal sensor array to collect data at the highest sampling frequency.

[0013] This invention also provides an anomaly monitoring system for an unmanned vehicle gas pipeline network simulation system, comprising:

[0014] The pipeline network basic data management module is used to store and manage the static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network.

[0015] The multi-source data fusion sensing module is used to receive and synchronize gas concentration data, three-dimensional wind speed vector data, atmospheric temperature and humidity data, and position and attitude data collected from the multimodal sensor array of the unmanned vehicle cluster.

[0016] The digital twin simulation engine module is used to construct a basic model of gas pipeline fluid dynamics based on the data provided by the pipeline basic data management module, and to integrate the real-time data provided by the multi-source data fusion sensing module to dynamically generate a digital twin simulation field of gas diffusion.

[0017] The leak source hypothesis evaluation module is used to set multiple candidate leak source hypothesis points in the gas diffusion digital twin simulation field, simulate their diffusion patterns, compare them with measured data, and calculate the confidence score of each hypothesis point.

[0018] The intelligent task planning module is used to filter out high-confidence areas based on the confidence score output by the leakage source hypothesis evaluation module and generate detection task instructions.

[0019] The unmanned vehicle cluster scheduling module is used to send the detection task instructions to the designated unmanned vehicles and monitor their task execution status.

[0020] As one embodiment of the present invention, the digital twin simulation engine module integrates two sub-modules: a pipeline fluid solver and an atmospheric diffusion solver. The pipeline fluid solver uses an implicit scheme to solve the one-dimensional transient compressible fluid control equations with a time step of 1 second. The atmospheric diffusion solver uses a large eddy simulation solver based on the finite volume method. Its computational domain covers the entire urban pipeline service area, with a total number of spatial grids of no less than 10 million. It also supports dynamic grid adaptive refinement, automatically refining the local grid in the detected high-concentration gradient region.

[0021] As one embodiment of the present invention, the leakage source hypothesis evaluation module maintains a dynamically updated candidate leakage source hypothesis pool. The initial state of the hypothesis pool is initialized by the key nodes of the pipeline network and historical fault records. After receiving new unmanned vehicle perception data each time, the module performs a global evaluation iteration, eliminates hypothesis points with confidence scores below a preset threshold, and generates new and more refined hypothesis points around high-confidence hypothesis points to achieve gradual focusing on the location of the leakage source.

[0022] As one embodiment of the present invention, the unmanned vehicle cluster scheduling module has the ability to coordinate multiple unmanned vehicles. It can dynamically allocate detection tasks according to the current location, remaining battery power, task priority and road traffic conditions of each unmanned vehicle, so as to ensure that high confidence areas are covered in the most timely and sufficient manner.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] 1. This invention constructs a high-fidelity digital twin simulation field that is linked in real time with the physical pipeline network, and interprets isolated sensor readings in a complete physical context, effectively overcoming the positioning deviation caused by the susceptibility of single sensor data to environmental noise interference.

[0025] 2. This invention utilizes the powerful inversion and prediction capabilities of simulation models to efficiently screen and verify a large number of leakage source hypotheses in virtual space, greatly reducing blind exploration of the physical world and significantly improving the speed and accuracy of positioning.

[0026] 3. The closed-loop feedback mechanism adopted in this invention enables the system to self-correct and self-optimize as the detection data is continuously fed back, and the positioning results have self-consistency and robustness.

[0027] 4. Through intelligent task planning and unmanned vehicle cluster scheduling, the optimal allocation of detection resources is achieved. While ensuring positioning accuracy, the inspection cost and response time are significantly reduced, providing a brand-new and systematic technical solution for the safe operation and maintenance of urban gas pipeline networks. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0029] Figure 2 This is a schematic diagram of the core principle framework of the digital twin simulation field for gas diffusion in this invention;

[0030] Figure 3 This is a logical flow diagram of the multi-source heterogeneous data fusion and dynamic simulation driving in this invention;

[0031] Figure 4This is a flowchart illustrating the logical process of candidate leakage source hypothesis evaluation and confidence score generation in this invention.

[0032] Figure 5 This is a logical flowchart of the refined exploration mission planning and unmanned vehicle cluster scheduling in this invention;

[0033] Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow between the unmanned vehicle and the digital twin simulation system in this invention. Detailed Implementation

[0034] Please refer to the attached document. Figure 1 To be continued Figure 6 This invention provides an anomaly monitoring method and system for an unmanned vehicle-based gas pipeline network simulation system, aiming to solve the technical problem that relying solely on sensor readings is insufficient for accurately and quickly locating gas leak sources in complex urban environments. In existing technologies, gas pipeline network leak monitoring mainly relies on fixed sensor networks or manual inspections. The former is limited by deployment density and cost, failing to cover all pipeline sections, while the latter suffers from drawbacks such as response lag, strong subjectivity, and low efficiency. Even with the introduction of mobile unmanned vehicles equipped with gas sensors for dynamic inspections, the single-dimensional concentration data collected is still easily affected by environmental disturbances such as wind direction, temperature, humidity, and building obstructions, leading to distorted concentration gradients and making it difficult to infer the true leak location. Furthermore, traditional methods lack the ability to collaboratively model pipeline topology, fluid dynamics characteristics, and real-time environmental factors, failing to achieve high-precision, low-latency leak source location in dynamically changing urban environments.

[0035] To overcome the aforementioned technical bottlenecks, this invention proposes an anomaly monitoring method and system that integrates multi-source heterogeneous data, constructs a high-fidelity digital twin simulation model, and drives unmanned vehicle collaborative detection based on a closed-loop feedback mechanism. This method is no longer limited to simple threshold judgment or local gradient analysis of raw sensor readings. Instead, it constructs a virtual simulation environment synchronized in real-time with the physical pipeline network, reproducing the physical process of gas diffusion within it. The simulation model parameters are continuously corrected using measured data transmitted back by unmanned vehicles, thereby efficiently converging the most probable leak source location in the virtual space. Based on this, the optimal detection path is planned, guiding the unmanned vehicle cluster to conduct targeted verification, forming a closed-loop intelligent monitoring system of perception-simulation-decision-verification.

[0036] The anomaly monitoring method of the unmanned vehicle gas pipeline network simulation system includes the following steps:

[0037] S1, acquire static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network;

[0038] S2 collects environmental data in real time, including gas concentration data, three-dimensional wind speed vector data, ambient temperature and humidity data, and the position and attitude data of the unmanned vehicle itself, through a multimodal sensor array.

[0039] S3. Based on the static topology data and historical operating condition data, construct an initial basic model of gas pipeline network fluid dynamics.

[0040] S4, input the environmental data into the basic model of gas pipeline fluid dynamics to generate a digital twin simulation field for gas diffusion;

[0041] S5. In the digital twin simulation field of gas diffusion, multiple candidate leakage source hypothetical points are set, and for each hypothetical point, the gas diffusion pattern generated under the current environmental conditions is simulated.

[0042] S6, compare the spatiotemporal consistency of the simulated gas diffusion pattern with the spatial distribution data of gas concentration actually collected by the unmanned vehicle, and calculate the confidence score of each candidate leak source hypothesis point;

[0043] S7. Based on the confidence score, the candidate leakage source area with the highest confidence is selected, and based on the spatial coordinate information of the area, a detection task instruction for the area is generated.

[0044] S8, the detection mission command is sent to the unmanned vehicle to guide it into the candidate leak source area to perform high-density grid sampling in order to obtain higher resolution local gas concentration field data;

[0045] S9, the high-resolution local gas concentration field data is fed back to the gas diffusion digital twin simulation field to further correct the location, leakage rate and direction parameters of the candidate leakage source until the positioning error converges to within the preset threshold.

[0046] In step S1, static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network are acquired. The static topology data includes the start and end node numbers, connection relationships, burial depth information, laying method, and branch structure of all pipeline segments, forming a complete directed graph representation. The pipe material attribute data includes the material type, inner diameter, outer diameter, wall thickness, roughness coefficient, allowable pressure rating, and service life of each pipeline segment. The valve status data records the installation location, model specifications, current opening status, and control logic of all control valves, shut-off valves, and pressure regulating valves. The historical operating condition data covers gas supply pressure curves, flow load curves, user gas usage patterns, and historical leakage event records for each typical day (weekday, holiday, and extreme weather day) within the past year. All of the above data is stored in the structured database of the pipeline network basic data management module and indexed by a unified resource identifier to ensure accurate retrieval during subsequent modeling processes.

[0047] In step S2, a multimodal sensor array deployed on the unmanned vehicle collects real-time data on the gas concentration, three-dimensional wind speed vector data, atmospheric temperature and humidity data, and the unmanned vehicle's own position and attitude data. The multimodal sensor array consists of a high-precision catalytic combustion gas concentration sensor, an ultrasonic three-dimensional anemometer, a digital temperature and humidity sensor, and a combined navigation unit integrating a GPS module and an inertial measurement unit module. The high-precision catalytic combustion gas concentration sensor has a measurement range of 0 to 5% volume concentration and a resolution of 1×10⁻⁶. -5 The response time is less than 3 seconds, and the sampling frequency is set to 20Hz. The ultrasonic three-dimensional anemometer has a wind speed measurement range of 0 to 60 m / s with an accuracy of ±0.5 m / s and a wind direction measurement accuracy of ±3°, with a sampling frequency of 10Hz. The digital temperature and humidity sensor has a temperature measurement range of -40℃ to +80℃ with an accuracy of ±0.5℃ and a humidity measurement range of 0 to 100% relative humidity with an accuracy of ±3%, with a sampling frequency of 5Hz. The integrated navigation unit provides a position update frequency of no less than 20Hz, a horizontal positioning error of less than 1m, a vertical positioning error of less than 2m, and an attitude angle (roll, pitch, yaw) measurement error of less than 0.5°. All sensor data is timestamped and preliminarily filtered locally on the autonomous vehicle before being uploaded to the central server via a wireless communication link at a frequency of no less than once per second.

[0048] In step S3, an initial basic model of the gas pipeline network's fluid dynamics is constructed based on the static topology data and historical operating condition data. This process first parses the static topology data into a directed graph structure composed of nodes and pipe segments, where nodes represent valves, pressure regulating stations, or user access points, and pipe segments represent pipelines connecting the nodes. Then, each pipe segment is assigned corresponding pipe material attribute data, including inner diameter, wall thickness, and roughness coefficient. Next, based on historical operating condition data, the pressure and flow boundary conditions for each node under normal operating conditions are determined. On this basis, a one-dimensional transient compressible fluid control equation set is used to discretize and model the entire pipeline network. This control equation set includes a continuity equation, a momentum equation, and an energy equation. The continuity equation is expressed as:

[0049] ;

[0050] The momentum equation is expressed as:

[0051] ;

[0052] The energy equation is expressed as:

[0053] ;

[0054] in, For gas density, The cross-sectional area of ​​the pipe. For flow rate, For pressure, The coordinates are along the axial direction of the pipeline. For time, It is the acceleration due to gravity. For the pipe inclination angle, Darcy friction factor The inner diameter of the pipe. Total energy per unit mass The heat flux density per unit area is given. The above governing equations are numerically solved using the finite volume method. The spatial discretization adopts a second-order upwind scheme, and the time progression adopts an implicit Euler scheme with a time step of 1 s. The initial steady-state distribution of pressure, velocity, and temperature at various points in the pipe network is obtained, which serves as the initial field for subsequent dynamic simulations.

[0055] In step S4, the data collected in real time by the multimodal sensor array is input into the gas pipeline network fluid dynamics basic model, driving it to dynamically evolve and generate a digital twin simulation field of gas diffusion consistent with the current physical world state. This process first uses the real-time atmospheric temperature and humidity data transmitted back by the unmanned vehicle as environmental background field parameters, inputting them into the computational fluid dynamics simulation engine. Then, the three-dimensional wind speed vector data transmitted back by the unmanned vehicle is used to construct a dynamic wind field model over the entire simulation area using the Kriging interpolation algorithm. Kriging interpolation is based on a semi-variogram model, using a spherical model for fitting, with a search radius set to 100m and a maximum number of nearest neighbors of 12 to ensure the smoothness and physical rationality of the wind field reconstruction. Next, the dynamic wind field model is unidirectionally coupled with the gas pipeline network fluid dynamics basic model; that is, the pipeline model provides the initial injection velocity and direction at the leak point, while the computational fluid dynamics simulation engine simulates the turbulent diffusion process of leaked gas under the influence of flow around complex urban buildings. The computational fluid dynamics simulation engine employs the large eddy simulation method, with a spatial grid resolution of at least 0.5m in the near-ground region and a non-uniform grid refinement in the vertical direction. The bottom layer of the grid has a height of 0.2m, increasing progressively upwards, with a time step of 0.1s to ensure accurate capture of transient diffusion phenomena. The simulation domain covers the entire urban pipeline service area, with a horizontal range of at least 10km² and a vertical height of at least 50m.

[0056] In step S5, multiple candidate leak source hypothetical points are set in the gas diffusion digital twin simulation field, and for each hypothetical point, the gas diffusion pattern generated under the current environmental conditions is simulated. The initial set of candidate leak source hypothetical points is initialized by key nodes of the pipeline network and historical fault records. Key nodes include all pressure regulating station outlets, main pipeline branch points, junctions of old pipeline sections, and access points in densely populated areas, with a total number not exceeding 500. Each hypothetical point is assigned an initial leakage rate range (0.01m). 3 / h to 10m 3 / h) and direction (omnidirectional or along the pipe axis). For each hypothetical point, the simulation engine runs a large eddy simulation under the current dynamic wind field and temperature and humidity background to generate a three-dimensional concentration field evolution sequence for the next 30 minutes, with a time resolution of 5s and a spatial resolution of 1m.

[0057] In step S6, the simulated gas diffusion pattern is compared with the spatial distribution data of gas concentration actually collected by the unmanned vehicle to determine spatiotemporal consistency, and the confidence score of each candidate leak source hypothesis point is calculated. This process defines a spatiotemporal matching function, which calculates the weighted root mean square error between the predicted gas concentration values ​​and the actual measured values ​​at all sampling times and locations of the unmanned vehicle under the candidate leak source hypothesis. Let the... An autonomous vehicle at any time Located in spatial position The measured concentration was Corresponding to the assumed source of the leak The simulated predicted concentration is The weighted root mean square error $E_k$ is then expressed as:

[0058] ;

[0059] Among them, weighting factors By unmanned vehicles and hypothetical leak sources Euclidean distance between Decision, defined as , The characteristic attenuation length is set to 50m. Confidence score. It is inversely proportional to the weighted root mean square error, and is defined as follows: All candidate leakage source hypotheses are sorted in descending order of confidence score, and the top 10% of high-scoring hypotheses are retained for the next iteration.

[0060] In step S7, based on the confidence score, the candidate leakage source region with the highest confidence is selected, and a detection task instruction for the region is generated based on the spatial coordinate information of the region. Specifically, a circular detection area with a radius of 50m is defined with the geometric center of the candidate leakage source region with the highest confidence as the center. Within the circular detection area, a spiral or serpentine high-density sampling path is planned, with a path spacing of 5m. The unmanned vehicle is instructed to travel along the high-density sampling path at a constant speed not exceeding 5 kilometers per hour and to activate its multimodal sensor array to collect data at the highest sampling frequency. The task instruction includes the path coordinate sequence, target speed, sampling frequency, communication reporting cycle, and safety obstacle avoidance strategy.

[0061] In step S8, the detection task command is issued to the unmanned vehicle (UAV), guiding it into the candidate leak source area to perform high-density grid sampling to obtain higher-resolution local gas concentration field data. Upon receiving the command, the UAV autonomously plans a local obstacle avoidance path, travels along a preset trajectory, and simultaneously collects gas concentration, wind speed, temperature, humidity, and location data at a frequency of 20Hz, transmitting this data back to the central server in real time. The sampling duration is 15 minutes, covering the entire detection area.

[0062] In step S9, the high-resolution local gas concentration field data is fed back to the gas diffusion digital twin simulation field to further correct the location, leakage rate, and direction parameters of the candidate leak source until the location error converges to within a preset threshold. This process generates a new, finer grid of hypothetical points around the original high-resolution hypothetical points, with a grid spacing of 5m and a leakage rate step size of 0.5m. 3 / h. Repeat steps S5 to S6 to calculate the confidence score of the new hypothesis point. If the position change of the highest confidence hypothesis point is less than 2m and the increase in confidence score is less than 5%, the localization is considered converged, and the final leak source coordinates, leak rate estimate, and confidence interval are output. Otherwise, return to step S7 to generate a new round of detection commands for a smaller range and continue iterating.

[0063] The anomaly monitoring system of the unmanned vehicle gas pipeline network simulation system includes: a pipeline network basic data management module, a multi-source data fusion perception module, a digital twin simulation engine module, a leak source hypothesis evaluation module, an intelligent task planning module, and an unmanned vehicle cluster scheduling module.

[0064] The pipeline network basic data management module is used to store and manage static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network. This module adopts a hybrid architecture of relational and graph databases. The static topology is stored in a graph structure, supporting fast traversal and subgraph queries; attribute and operating condition data are stored in relational tables, supporting efficient indexing and batch reading. The data interface conforms to the Open Geospatial Consortium standard and supports integration with other municipal information systems.

[0065] The multi-source data fusion sensing module receives and synchronizes gas concentration data, 3D wind speed vector data, atmospheric temperature and humidity data, and position and attitude data collected from the multimodal sensor array of the unmanned vehicle cluster. This module has a built-in time synchronization engine that uses a network time protocol to align the local clocks of each unmanned vehicle, with an error controlled within 10ms. Data reception employs a message queue mechanism to ensure no data loss under high concurrency. After Kalman filtering and outlier removal, the raw data is stored in a time-series database for real-time access by the simulation engine.

[0066] The digital twin simulation engine module is used to construct a basic fluid dynamics model of the gas pipeline network based on the data provided by the pipeline network basic data management module, and to integrate real-time data provided by the multi-source data fusion sensing module to dynamically generate a digital twin simulation field for gas diffusion. This module integrates two sub-modules: a pipeline network fluid solver and an atmospheric diffusion solver. The pipeline network fluid solver uses an implicit scheme to solve the one-dimensional transient compressible fluid control equations with a time step of 1 second. The atmospheric diffusion solver uses a large eddy simulation solver based on the finite volume method. Its computational domain covers the entire urban pipeline network service area, with a total spatial grid of no less than 10 million, and supports dynamic adaptive grid refinement. When a high concentration gradient region is detected, it automatically refines the local grid in that region, with a minimum grid size of 0.2m.

[0067] The leak source hypothesis evaluation module is used to set multiple candidate leak source hypothesis points in the gas diffusion digital twin simulation field, simulate their diffusion patterns, compare them with measured data, and calculate the confidence score of each hypothesis point. This module maintains a dynamically updated pool of candidate leak source hypotheses, the initial state of which is jointly initialized by key nodes of the pipeline network and historical fault records. Each time new unmanned vehicle perception data is received, the module performs a global evaluation iteration, eliminating hypothesis points with confidence scores below a preset threshold, and generating new, more refined hypothesis points around high-confidence hypothesis points to gradually focus on the leak source location. The evaluation calculation uses a parallel computing framework, which can complete the full simulation and scoring of 1000 hypothesis points within 5 minutes.

[0068] The intelligent task planning module is used to filter out high-confidence areas based on the confidence score output by the leakage source hypothesis evaluation module and generate detection task instructions. This module has a built-in path planning algorithm library, supporting multiple path generation strategies such as A-satellite, RRT, and optimized splines. It can generate safe and efficient detection trajectories based on urban road networks, traffic control information, and unmanned vehicle performance parameters. The task instructions are encapsulated in a structured message format, including the target area, path point sequence, speed constraints, sampling configuration, and emergency response plan.

[0069] The unmanned vehicle cluster scheduling module is used to issue the detection task instructions to the designated unmanned vehicles and monitor their task execution status. This module has the capability for multi-vehicle collaborative operation, dynamically allocating detection tasks based on each unmanned vehicle's current location, remaining battery power, task priority, and road conditions to ensure the most timely and comprehensive coverage of high-confidence areas. The scheduling strategy employs an auction-based task allocation algorithm, comprehensively considering task urgency, unmanned vehicle endurance, and path conflict risks to achieve optimal global resource allocation. During task execution, the module continuously receives unmanned vehicle status heartbeats; if it detects abnormalities such as insufficient battery power, communication interruption, or deviation from the path, it immediately triggers a task reassignment or emergency recall process.

[0070] Through the coordinated operation of the above methods and systems, this invention achieves high-precision, low-latency, and low-cost intelligent monitoring of urban gas pipeline network leakage events, effectively solving the core technical problem of difficulty in locating leakage sources in complex environments.

Claims

1. An anomaly monitoring method for an unmanned vehicle gas pipeline network simulation system, characterized in that, include: Acquire static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network; The system collects environmental data in real time, including gas concentration data, three-dimensional wind speed vector data, ambient temperature and humidity data, and the position and attitude data of the unmanned vehicle itself, through a multimodal sensor array. Based on the static topology data and historical operating condition data, an initial basic model of gas pipeline network fluid dynamics is constructed. The environmental data is input into the basic model of gas pipeline fluid dynamics to generate a digital twin simulation field for gas diffusion; In the digital twin simulation field of gas diffusion, multiple candidate leakage source hypotheses are set, and for each hypothetical point, the gas diffusion pattern generated under the current environmental conditions is simulated. The simulated gas diffusion pattern is compared with the spatial distribution data of gas concentration actually collected by the unmanned vehicle to calculate the confidence score of each candidate leak source hypothesis point. Based on the confidence score, the candidate leakage source area with the highest confidence is selected, and a detection mission instruction is generated; The detection mission command is issued to the unmanned vehicle, which is then guided to enter the candidate leak source area to obtain local gas concentration field data. The local gas concentration field data is fed back to the gas diffusion digital twin simulation field to correct the location, leakage rate and direction parameters of the candidate leakage source until the positioning error converges to within the preset threshold. The initial basic model of gas pipeline network fluid dynamics specifically includes: The static topology data is parsed into a directed graph structure consisting of nodes and pipe segments, where nodes represent valves, pressure regulating stations or user access points, and pipe segments represent pipelines connecting the nodes. Assign corresponding pipe material attribute data to each pipe section, including inner diameter, wall thickness, and roughness coefficient; Based on historical operating data, the pressure and flow boundary conditions of each node under normal operating conditions are determined. The entire pipeline network is discretized and modeled using a one-dimensional transient compressible fluid control equation set, which includes a continuity equation, a momentum equation, and an energy equation. The finite volume method was used to numerically solve the governing equations to obtain the initial steady-state distribution of pressure, flow velocity and temperature at various points in the pipeline network. The generation of a digital twin simulation field for gas diffusion specifically includes: The real-time ambient temperature and humidity data transmitted back by the unmanned vehicle are used as environmental background field parameters and input into the computational fluid dynamics simulation engine. The three-dimensional wind speed vector data transmitted back by the unmanned vehicle is used to construct a dynamic wind field model over the entire simulation area through the Kriging interpolation algorithm. The dynamic wind field model is unidirectionally coupled with the gas pipeline network fluid dynamics basic model. That is, the pipeline network model provides the initial injection velocity and direction of the leak point, while the computational fluid dynamics simulation engine simulates the turbulent diffusion process of the leaked gas under the influence of the flow around the complex urban building complex. The computational fluid dynamics simulation engine adopts the large eddy simulation method, with a spatial grid resolution of no less than 0.5m in the near-ground region and non-uniform grid refinement in the vertical direction, and a time step set to 0.1s. The calculation of the confidence score for each candidate leakage source hypothesis point specifically includes: Define a spatiotemporal matching degree function, which calculates the weighted root mean square error between the predicted gas concentration values ​​and the actual measured values ​​in the simulation field at all sampling times and sampling locations of the unmanned vehicle under the candidate leakage source assumption. The weighting factor of the weighted root mean square error is determined by the Euclidean distance between the unmanned vehicle and the hypothetical leakage source; the closer the distance, the higher the weight. The confidence score is inversely proportional to the weighted root mean square error; the smaller the error, the higher the score. All candidate leak source hypotheses are sorted in descending order of confidence score.

2. The anomaly monitoring method for the unmanned vehicle gas pipeline network simulation system according to claim 1, characterized in that, The multimodal sensing array includes a high-precision catalytic combustion gas concentration sensor, an ultrasonic three-dimensional anemometer, a digital temperature and humidity sensor, and a combined navigation unit integrating a global positioning system module and an inertial measurement unit module; the high-precision catalytic combustion gas concentration sensor has a measurement range of 0 to 5% volume concentration and a resolution of 1×10⁻⁶. -5 The response time is less than 3 seconds; the wind speed measurement range of the ultrasonic three-dimensional anemometer is 0 to 60 m / s with an accuracy of ±0.5 m / s, and the wind direction measurement accuracy is ±3°; the position update frequency provided by the integrated navigation unit is not less than 20 Hz, and the horizontal positioning error is less than 1 m.

3. The anomaly monitoring method for the unmanned vehicle gas pipeline network simulation system according to claim 2, characterized in that, The specific steps for generating exploration mission instructions for this area include: A circular detection zone with a radius of 50m is delineated with the geometric center of the candidate leakage source region with the highest confidence level as the center. Within the circular detection area, a spiral or serpentine high-density sampling path is planned, with the path spacing set at 5m. The unmanned vehicle is instructed to travel at a constant speed of no more than 5 kilometers per hour along the high-density sampling path and to activate its multimodal sensor array to collect data at the highest sampling frequency.

4. The anomaly monitoring method for the unmanned vehicle gas pipeline network simulation system according to claim 3, characterized in that, The initial set of candidate leakage source hypotheses is initialized by the key nodes of the pipeline network and historical fault records; the key nodes include the outlet of the pressure regulating station, the branch point of the main pipeline, the junction of old pipeline sections and the access point in densely populated areas.

5. The anomaly monitoring method for the unmanned vehicle gas pipeline network simulation system according to claim 4, characterized in that, The digital twin simulation field for gas diffusion supports dynamic mesh adaptive refinement. When a high concentration gradient region is detected, the local mesh is automatically refined in that region.

6. An anomaly monitoring system for an unmanned vehicle gas pipeline network simulation system, characterized in that, An anomaly monitoring method for implementing the unmanned vehicle gas pipeline network simulation system according to any one of claims 1 to 5 includes: The pipeline network basic data management module is used to store and manage the static topology data, pipe material attribute data, valve status data, and historical operating condition data of the urban gas pipeline network. The multi-source data fusion sensing module is used to receive and synchronize gas concentration data, three-dimensional wind speed vector data, atmospheric temperature and humidity data, and position and attitude data collected from the multimodal sensor array of the unmanned vehicle cluster. The digital twin simulation engine module is used to construct a basic model of gas pipeline fluid dynamics based on the data provided by the pipeline basic data management module, and to integrate the real-time data provided by the multi-source data fusion sensing module to dynamically generate a digital twin simulation field of gas diffusion. The leak source hypothesis evaluation module is used to set multiple candidate leak source hypothesis points in the gas diffusion digital twin simulation field, simulate their diffusion patterns, compare them with measured data, and calculate the confidence score of each hypothesis point. The intelligent task planning module is used to filter out high-confidence areas based on the confidence score output by the leakage source hypothesis evaluation module and generate detection task instructions. The unmanned vehicle cluster scheduling module is used to send the detection task instructions to the designated unmanned vehicles and monitor their task execution status.

7. The anomaly monitoring system of the unmanned vehicle gas pipeline network simulation system according to claim 6, characterized in that, The digital twin simulation engine module integrates two sub-modules: a pipeline fluid solver and an atmospheric diffusion solver. The pipeline fluid solver uses an implicit scheme to solve the one-dimensional transient compressible fluid control equations. The atmospheric diffusion solver uses a large eddy simulation solver based on the finite volume method, and its computational domain covers the entire urban pipeline service area.

Images