Intelligent gas pipe network dynamic inspection path planning method and system

Abstract

The invention provides an intelligent gas pipe network dynamic inspection path planning method, which is applied to a central control platform and comprises the following steps: acquiring pipeline fixed monitoring equipment data, external environment data, intelligent driving inspection vehicle state data and inspection data in real time in an inspection process according to an initial inspection path; if the inspection data does not include newly added hidden danger point information, dynamically updating an inspection path according to pipeline fixed monitoring equipment data, external environment data and state data of the intelligent driving inspection vehicle, and issuing the inspection path to the intelligent driving inspection vehicle; and if the inspection data comprises newly added hidden danger point information, generating an optimal reinspection path and a reinspection observation point position by taking the newly added hidden danger point as a center according to the external environment data and issuing the optimal reinspection path and the reinspection observation point position to the intelligent driving inspection vehicle, and after the intelligent driving inspection vehicle finishes reinspection of the hidden danger point, recovering the inspection path before reinspection.

Description

Technical Field

[0001] The present invention relates to the field of gas inspection technology, and in particular to a method and system for planning dynamic inspection paths for an intelligent gas pipeline network. Background Art

[0002] As municipal gas pipeline networks continue to expand, traditional manual inspection methods are increasingly limited in their ability to meet the demands of modern urban management and safe gas operations. Gas pipeline inspections not only require comprehensive coverage of the vast network, but also must address complex and volatile urban road conditions and emergencies such as road construction and traffic congestion. Furthermore, the diffusion and inverse location of gas leaks place higher demands on the quality and coverage of inspection observation points.

[0003] While existing intelligent driving inspection vehicle solutions have achieved improved automation levels, such as CN116506470B, which provides a safety inspection method based on smart gas GIS, their path planning often relies on fixed routes or preset rules, lacking dynamic adaptability and real-time adjustment capabilities. These solutions face the following major challenges: (1) Complexity of urban roads: Urban road environments are complex and ever-changing, including regulatory constraints such as one-way streets and no-traffic restrictions, as well as real-time dynamics such as road closures and traffic congestion. These factors significantly affect inspection efficiency. Existing intelligent driving inspection vehicle solutions typically rely on fixed routes and preset rules, lacking flexibility and unable to respond to these dynamic changes in real time. This makes it difficult to prioritize coverage of high-risk areas and adapt to emergencies (such as road closures and traffic changes).

[0004] (2) The particularity of gas pipeline networks: Gas pipeline networks are usually distributed underground, and hidden dangers are often concentrated in specific high-risk areas (such as old pipelines and intersections); the spread of leakage is affected by a variety of external factors (such as wind direction, terrain, climate, etc.), which increases the difficulty of data collection and analysis; at the same time, complex distribution structures such as double-sided pipeline networks and dead-end roads also put higher requirements on inspection route planning.

[0005] (3) Insufficient data collection and inversion accuracy: The existing solutions are difficult to support high-quality, multi-round inspection tasks, resulting in low sampling density of potential danger points and insufficient data coverage, which in turn affects the accuracy of gas leak inversion positioning.

[0006] (4) Lack of flexibility: Fixed routes are difficult to dynamically adapt to emergencies within the inspection area (such as the addition of new hidden danger points and changes in construction areas). The allocation of inspection resources is not optimized, which easily leads to waste of time and cost.

[0007] Therefore, urban gas pipeline inspections urgently need an intelligent and flexible path planning method that can be driven by risk assessment to achieve dynamically optimized inspection paths and task allocation, while supporting high-precision data collection and inversion analysis, and fully adapting to the gas pipeline inspection needs in complex urban environments.

[0008] In order to solve the above problems, people have been seeking an ideal technical solution. Summary of the Invention

[0009] Based on this, it is necessary to address the problems of insufficient flexibility, poor dynamic adaptability, untimely inspection response in high-risk areas, and low accuracy of data collection and leakage inversion in traditional gas pipeline inspection methods. An intelligent gas pipeline network dynamic inspection path planning method based on the fusion of municipal gas pipeline network GIS information, urban road rules, dynamic risk assessment and real-time data is proposed.

[0010] To achieve the above objectives, the present invention provides a first aspect of a municipal gas network inspection route planning method based on GIS, which is applied to a central control platform and includes the following steps: During the inspection process according to the initial inspection route, real-time data from pipeline fixed monitoring equipment, external environment data, status data of intelligent driving inspection vehicles, and inspection data are obtained; If the inspection data does not include information on newly added potential risk points, the risk level and inspection frequency of the pipeline network nodes and sections are dynamically obtained based on the data from the pipeline's fixed monitoring equipment, external environmental data, and the status data of the intelligent driving inspection vehicle. The path planning model is dynamically updated based on the external environmental data and risk level. The inspection path is dynamically updated and sent to the intelligent driving inspection vehicle based on the genetic algorithm, the updated path planning model, road rules, the status data of the intelligent driving inspection vehicle, and the risk level and inspection frequency obtained in the previous steps. If the inspection data includes information about newly added potential danger points, the optimal re-inspection path and re-inspection observation point locations are generated based on the external environment data, with the newly added potential danger point as the center, and sent to the intelligent driving inspection vehicle. After the intelligent driving inspection vehicle completes the re-inspection of the potential danger point, it resumes the inspection path before the re-inspection. Repeat the above steps until the inspection is completed.

[0011] Furthermore, after the optimal re-inspection path and re-inspection observation point location are sent to the intelligent driving inspection vehicle, the leakage source location and intensity returned by the intelligent driving inspection vehicle are also received, and the hidden danger points of abnormal alarms are eliminated based on the leakage source location and intensity; the leakage source location and intensity are obtained by the intelligent driving inspection vehicle based on the re-inspection data and real-time external environment data using the Gaussian diffusion model inversion.

[0012] Furthermore, the following formula is used to comprehensively evaluate the risk levels of pipeline network nodes and sections: Rv=α·Rhistory+β·Rsensor+γ·Rexternal+δ·Rcoverage Among them, Rv represents the comprehensive risk level of the node or pipe section; Rhistory represents the historical risk value; Rsensor represents the sensor risk value; Rexternal represents the external environment risk value; Rcoverage represents the inspection coverage value; α represents the weight of the historical risk factor; β represents the weight of the real-time sensor data; γ represents the weight of the external environment; δ represents the weight of the inspection coverage status.

[0013] Furthermore, when dynamically updating the inspection path based on the updated path planning model, road rules, intelligent driving inspection vehicle status data, and the risk level and inspection frequency obtained in the above steps, an incremental update method is used to dynamically update the inspection path.

[0014] Furthermore, the path planning model includes a dynamic directed graph G = (V, E) that connects gas pipeline nodes and urban road nodes based on the gas pipeline GIS information, where V represents the node and E represents the connection path; each edge of the dynamic directed graph is set with a basic weight: We = α·Wd + β·Wr + γ·Wt + δ·Wtime + λ·Wcoverage; Among them, α, β, γ, and δ are adjustment coefficients; λ represents the adjustment coefficient of the weight factor, which is used to give priority to covering uninspected pipelines; Wr = 1 / Rv represents the risk weight; Wd represents the actual distance between representative nodes; Wr represents the risk weight, and high-risk points have lower weights; Wt represents the real-time traffic congestion weight, and the more severe the congestion, the higher the weight; Wtime represents the time weight, which is used to estimate the time required to complete the task; and Wcoverage indicates whether the edge has been covered.

[0015] Furthermore, during the inspection process, the real-time traffic data interface is called to filter out closed or prohibited road sections; for passable road sections, the traffic congestion weight Wt is dynamically updated based on the real-time traffic data; For road closures not near pipelines due to construction or traffic accidents, directly remove the corresponding edge E in the path planning model. If a dead-end road node is isolated based on real-time traffic data, a virtual edge is inserted between the node and the nearest inspected node in the path planning model, and its comprehensive weight Wvirtual is calculated, where Wvirtual=Wd+Wr+Wt; Determine the newly added high-risk points based on the risk levels before and after the update, and adjust the risk weight of the risk node in the path planning model.

[0016] Furthermore, the optimal re-inspection path is generated based on real-time external environment data, including: Taking the potential risk point as the center, a multi-hop search is performed on its neighboring nodes in the path planning model to construct a navigable subgraph. In the subgraph, all "ring" or "loop" paths that can pass through the potential risk point and surrounding key nodes are searched as candidate paths. Traverse all candidate paths, score and sort them based on the edge weight formula; If a closed-loop path with the highest score is found to cover the alarm area, the re-inspection path will be directly sent to the intelligent driving inspection vehicle; If a single loop is insufficient to cover all high-risk sub-areas, several candidate loops with higher scores will be spliced or executed in segments; During the re-inspection process, if new road closures, worsening congestion, or expanded potential hazards are detected again, the dynamic path planning algorithm will be invoked to update the weights of the affected sections and incrementally adjust the current circular path to ensure continued high-density coverage of the alert area. In the case of one-way streets or where U-turns are not possible, the vehicle can immediately return to the potential hazard via surrounding U-turn nodes after leaving, forming a closed loop to ensure multiple coverage of the same area. Generate recheck observation point locations based on external environment data, including: During re-inspection, when the distance to the hidden danger point in the inspection lane is greater than or equal to the set threshold, the inspection vehicle is controlled to travel at the first driving speed and the inspection data is sampled at the preset sampling frequency; When the distance between the inspection vehicle and the potential hazard point is less than a set threshold, the inspection vehicle is controlled to travel at a second driving speed, wherein the first driving speed is greater than or equal to the second driving speed; At the same time, wind direction and wind speed are obtained in real time, and the sampling frequency is increased according to the wind direction and wind speed.

[0017] Furthermore, the inspection frequency calculation formula is as follows: Fv=Fbase+θ·(1 / Rv)+μ·Sr Where: Fv represents the inspection frequency of node v; Fbase represents the basic inspection frequency; Rv represents the risk level; Sr represents the dynamic risk factor, which indicates the adjustment of the inspection frequency due to real-time changes; θ and μ represent the adjustment coefficients, which are used to balance the impact of risk level and dynamic factors on the inspection frequency.

[0018] To achieve the above objectives, the second aspect of the present invention provides a GIS-based municipal gas pipeline network inspection system, characterized by comprising a central control platform and an intelligent driving inspection vehicle, wherein the central control platform is used to execute the municipal gas pipeline network inspection path planning method described in the first aspect; The intelligent driving inspection vehicle is equipped with an AI pilot module, a communication module, a sensor system, a navigation system and an edge computing unit; The sensor system includes a methane telemeter, a pump-type methane and ethane detector, a vehicle-mounted positioning module, a lidar, a camera, and a mobile weather station. It is used to collect real-time data on gas leakage concentration and the surrounding environment of the pipe network, and generate accurate inspection point data based on the positioning information. The navigation system is used to realize the navigation of the intelligent driving inspection vehicle in a complex road environment based on the high-precision positioning module and the laser radar; The AI navigation module is used to dynamically calculate the pointing position of the methane telemeter based on the gas pipeline network GIS information, risk level distribution and real-time vehicle location, thereby achieving fully automatic scanning of the gas pipeline network; The communication module is used to perform two-way communication with the central control platform to receive task instructions and upload inspection data; The edge computing unit is used to process the sensor data collected by the sensor system in real time and implement preliminary local decision-making functions.

[0019] The beneficial effects of the present invention are: 1. Multi-objective optimization and risk-driven task optimization This invention combines a genetic algorithm with a dynamic path optimization algorithm to achieve a dynamic balance between full inspection coverage and prioritized coverage of high-risk areas. Through a dynamic adjustment mechanism, high-risk areas are prioritized based on real-time risk assessment, construction, weather, and alarm data. This ensures that task priorities are optimized when resources are limited, while achieving an efficient combination of full coverage and re-inspection tasks when resources are sufficient. Furthermore, for low- and medium-risk areas, the dynamic frequency reduction or postponement of inspections optimizes task resource allocation and improves overall inspection efficiency.

[0020] 2. Data collection optimization and accurate re-inspection To address the difficulties in locating gas leak sources, this invention incorporates high-density re-inspection route planning and a multi-round data collection strategy to improve data quality at potential risk points. Combining weighted data fusion with a Gaussian inversion model effectively improves the reliability of leak source location, reduces false alarm rates, and provides precise support for emergency repair decisions. The re-inspection route achieves high-density coverage through low-speed driving and multiple sampling, ensuring data accuracy and meeting the precise requirements of gas inspections.

[0021] 3. Scenario adaptability and technical advantages This invention optimizes path planning and task allocation, and is particularly suitable for complex scenarios encountered during gas pipeline network inspections. Through a dynamic adjustment mechanism, combined with real-time data (such as leak alarm points and traffic conditions), it rapidly responds to high-risk points, resolving the difficulty traditional GIS path planning systems face in coping with real-time risk adjustments. Furthermore, in complex scenarios such as dead-end roads and two-sided pipeline networks, a virtual edge insertion and incremental update strategy are employed to ensure seamless path optimization and continuity. Compared to traditional technologies, this invention significantly improves inspection flexibility, real-time performance, and task efficiency, addressing the shortcomings of static inspection methods and meeting the needs of gas pipeline network inspections in dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 1 is a flow chart of the inspection method according to embodiment 1 of the present invention; Figure 2 It is a schematic diagram of the process of path planning and real-time data evaluation of the present invention; Figure 3 It is a schematic diagram of the process of triggering the re-inspection task and inverting the leakage source of the present invention; Figure 4 is a network diagram of the inspection system according to Example 2 of the present invention; DETAILED DESCRIPTION

[0023] The technical solution of the present invention is further described in detail below through specific implementation methods.

[0024] Example 1 This embodiment provides a municipal gas network inspection route planning method based on GIS, which is applied to the central control platform, such as Figure 1-3 As shown, the following steps are included: During the inspection process according to the initial inspection route, pipeline fixed monitoring equipment data, external environment data, intelligent driving inspection vehicle status data and inspection data are obtained in real time.

[0025] Specifically, the external environmental data includes at least traffic data, construction data, and meteorological data; the intelligent driving inspection vehicle status data includes vehicle location, inspection progress, remaining battery power, and other information. The inspection data includes gas concentration and potential danger point information. It is important to note that when the intelligent driving inspection vehicle discovers that the gas concentration exceeds the alarm point during inspection, it determines that the point is a potential danger point and reports this potential danger point information along with the gas concentration to the central control platform; otherwise, only the gas concentration is uploaded to the central control platform.

[0026] If the inspection data does not include information on new hidden danger points, the risk level and inspection frequency of pipeline network nodes and sections will be dynamically obtained based on the pipeline fixed monitoring equipment data, external environment data and intelligent driving inspection vehicle status data.

[0027] It is understandable that in the risk assessment step, it is first necessary to integrate all data sources, clean and pre-process the data to ensure its integrity and accuracy.

[0028] Specifically, the following formula is used to comprehensively evaluate the risk levels of pipeline network nodes and sections: Rv=α·Rhistory+β·Rsensor+γ·Rexternal+δ·Rcoverage Among them, Rv represents the comprehensive risk level of the node or pipe section, and the higher the value, the higher the risk; Rhistory represents the historical risk value; Rsensor represents the sensor risk value, which reflects the abnormal data detected during the inspection (such as gas concentration, wind speed, and wind direction); Rexternal represents the external environment risk value, which is used to characterize external influencing factors such as construction, weather, and traffic; Rcoverage represents the inspection coverage value, and the pipe sections that are not covered or need to be re-inspected are given a higher weight. In specific implementation, the pipe sections that need to be re-inspected have the highest weight; α represents the weight of the historical risk factor, which is suitable for scenarios with a large number of old pipelines or high fault points; β represents the weight of real-time sensor data, which is suitable for high-sensitivity inspection scenarios; γ represents the weight of the external environment, which is suitable for areas with a large impact of construction and bad weather; δ represents the weight of the inspection coverage status, which is suitable for the priority of areas where tasks are not completed.

[0029] Specifically, in actual use, the historical risk value Rhistory is primarily calculated based on a comprehensive analysis of the number of failures, maintenance records, and pipe age. Typically, the number of failures Nfault over the past year or multiple years is divided by the network's maximum number of failures Nmax. Weighted scores are then applied to factors such as pipe age and maintenance costs. The final result is mapped to a range of 0 to 1, where 0 indicates no historical failures and 1 indicates frequent failures or extremely old pipes.

[0030] The sensor risk value, Rsensor, is calculated based on the alarm concentrations of pump-type methane and ethane detectors, methane telemeters, and fixed monitoring equipment. If the methane concentration exceeds the threshold (CH4th), the risk is marked as high. The risk value can be further modified based on information such as the duration and number of alarms. If the concentration exceeds the threshold continuously or the alarm is triggered multiple times, the risk value approaches 1. If the concentration never exceeds the threshold and alarms are rare, the risk value approaches 0. Ultimately, all values are mapped to the range of 0 to 1.

[0031] The external environmental risk value (Rexternal) is adjusted based on external environmental factors such as construction, weather, traffic, terrain, and illegal occupation. When construction distance is detected to be less than 20 meters, or when extreme weather conditions such as heavy rain or strong winds occur, the risk factor for that section is increased. Corrections are also made based on factors such as hazard distribution and population density that could exacerbate the consequences of an accident. The result is ultimately normalized to a range of 0 to 1. If overall external environmental conditions are favorable, the risk value approaches 0; if conditions are poor, it approaches 1.

[0032] Inspection coverage value Rcoverage: If no inspection has been conducted, the risk value can be set to 1; if the inspection has been completed and no abnormalities have been found, it can be set to 0; for cases where there is only partial coverage or suspected defects have been found but not yet confirmed, an intermediate value between 0 and 1 can be assigned to dynamically reflect the impact of the inspection coverage on the risk.

[0033] For example, insufficient coverage: the on-board sensor only detected about 50% of the key areas in the pipe section, and the other part could not be completed due to construction or road conditions; unstable alarm: although the vehicle passed once, the sensor data fluctuated too much and needed to be re-inspected for confirmation. For example, the concentration rose sharply and then fell back quickly in a short period of time, or the same location was repeatedly measured with high and low readings. Possible reasons for this situation include temporary exhaust interference, equipment noise, sudden change in wind direction causing air mass movement, etc. In order to avoid misjudging such transient anomalies as real leaks, the present invention supports secondary or multiple re-inspections.

[0034] In a practical system, each of these risk values should first be normalized to a value between 0 and 1. Then, Rhistory, Rsensor, Rexternal, and Rcoverage should be linearly weighted using the four weights α, β, γ, and δ to determine the overall risk, Rv. By flexibly adjusting these weights, we can account for multiple factors, including historical risk, sensor alarms, the external environment, and inspection coverage, leading to a more accurate assessment of the overall risk profile of the municipal gas pipeline network.

[0035] It is understandable that in some scenarios, not all parameters need to be involved in the risk level assessment. For example, when located inside a residential area or factory, it is not easily affected by construction or bad weather. In this case, Rexternal and γ can be 0 and do not participate in the risk level assessment.

[0036] Specifically, the risk levels include high-risk areas, medium-risk areas and low-risk areas.

[0037] Among them, high-risk areas include old pipelines, high-pressure pipeline sections, pipeline sections with frequent failures, and construction points or sensor alarm areas next to pipelines.

[0038] The medium-risk area refers to the pipeline section with high daily operating pressure but relatively stable historical status.

[0039] Low-risk areas are those that operate normally and have a good maintenance record.

[0040] Furthermore, differentiated inspection frequencies are set according to risk levels.

[0041] Specifically, for high-risk areas, the inspection frequency will be increased and each mission must be covered; if inspection resources are limited, high-risk areas will be given priority in completing tasks.

[0042] For medium-risk areas, maintain the standard inspection frequency; when new risk points or external influences (such as weather changes) emerge, dynamically increase the priority.

[0043] For low-risk areas, the priority is the lowest. When resources are insufficient, the coverage rate can be temporarily reduced and wait for the next round of inspections.

[0044] In specific implementation, the inspection frequency calculation formula is as follows: Fv=Fbase+θ·(1 / Rv)+μ·Sr Where: Fv represents the inspection frequency of node v; Fbase represents the basic inspection frequency; Rv represents the risk level; Sr represents the dynamic risk factor, which indicates the adjustment of the inspection frequency due to real-time changes; θ and μ represent the adjustment coefficients, which are used to balance the impact of risk level and dynamic factors on the inspection frequency.

[0045] In the scenario of insufficient resources, the inspection frequency of high-risk levels is adjusted by adjusting the weights of Fbase and θ; in the scenario of sufficient resources, the stability of Fbase is maintained, and the inspection frequency is adjusted through the dynamic factor Sr.

[0046] Specifically, the steps for obtaining the dynamic risk factor Sr in the inspection frequency calculation are as follows: In practice, Sr primarily reflects real-time changes that may occur during the inspection process, such as sudden construction, deteriorating weather conditions, traffic control, or the addition of new high-risk alarm points. These external events or sensor alarm information are converted into a numerical value between 0 and 1: 0 indicates no major sudden change; 1 indicates the most severe or highest priority dynamic risk scenario. Further intervals (such as 0.2, 0.5, 0.8, etc.) can also be set based on business needs to measure the impact of dynamic factors on inspection frequency. The system defines an initial value for each event type based on task requirements and historical experience, and updates it in real time during the inspection process.

[0047] Adjustment coefficients θ and μ: θ controls the impact of node risk level 1 / Rv on inspection frequency; μ determines the weight of the dynamic risk factor Sr in the overall inspection strategy. Both are typically obtained through the following methods: Offline parameter adjustment: Multiple simulations or replays are conducted based on historical inspection data, comparing the impact of different parameter combinations on inspection efficiency and false negative rate. Ultimately, a value is selected that balances high risk sensitivity with overall stability.

[0048] Periodic correction: After a period of actual inspection operation, a small adjustment is made to θ,μ based on the local pipeline network characteristics, alarm frequency, and other feedback to adapt to the new risk distribution or resource conditions.

[0049] Through the above methods, Sr, θ, and μ can be flexibly set in actual deployment and incrementally adjusted during the inspection process, thereby achieving a comprehensive balance between risk level and real-time dynamics, ensuring that the inspection frequency can respond to high-risk events in a timely manner without excessively consuming resources.

[0050] It should be noted that when the distance between the updated high-risk area and the current inspection task is close, a local path adjustment will be made and the adjusted inspection path will be executed to reach the high-risk point; when the distance between the updated high-risk area and the current inspection task is farther away, the current inspection path will still be followed.

[0051] Then, the path planning model is dynamically updated based on external environment data and risk levels.

[0052] Specifically, the path planning model is a dynamic directed graph G = (V, E) that connects gas pipeline nodes and urban road nodes based on the gas pipeline GIS information. V represents the node and E represents the connection path. Each edge of the dynamic directed graph is set with a basic weight, We = α·Wd + β·Wr + γ·Wt + δ·Wtime + λ·Wcoverage. Among them, α, β, γ, and δ are adjustment coefficients used to balance the impact of various factors on path planning; λ represents the adjustment coefficient of the weight factor, which is used to give priority to covering pipelines that have not been inspected; distance weight Wd: calculated based on the actual road distance, that is, the actual distance between nodes; risk weight Wr: set according to the risk level provided by the risk assessment module, Wr=1 / Rv, and high-risk points have lower weights; traffic weight Wt: dynamically updates the congestion weight based on real-time traffic data, and the more severe the congestion, the higher the weight; time weight Wtime: estimates the time required to complete the task; coverage weight Wcoverage: used to indicate whether the pipeline section has been covered, and the uncovered section has a higher weight.

[0053] In actual use, the present invention connects with a third-party traffic platform (such as a city traffic management bureau or an internet map service) to obtain the current speed or congestion index of the target road (such as "green for smooth traffic," "yellow for slow traffic," and "red for congestion") in real time. It can also combine information such as driving speed and queue length uploaded by the unmanned inspection vehicle itself to quantify the degree of congestion. If the current actual speed of the road is recorded as vcurrent, and the theoretical speed of the road in an uncongested state is vfree, a congestion factor CI = 1-vcurrent / vfree can be defined. The larger the CI value, the more congested the road. To facilitate unified use in route planning, the CI can be mapped to [0,1] or other custom intervals and assigned to Wt or multiplied by an adjustment coefficient on this basis, so that "more congested roads" have a higher value in the comprehensive weight (which is not conducive to vehicle selection).

[0054] It is also important to note that in actual applications, this congestion factor can be automatically updated every few seconds (such as 3 to 10 seconds), or refreshed immediately in the event of a sudden traffic incident, and combined with incremental algorithms such as dynamic path planning to locally recalculate the weights of relevant road sections without the need for global path reconstruction.

[0055] Inspection coverage weight (Wcoverage) is typically set to 0 for "covered" and 1 for "uncovered." This gives uninspected edges a higher value in the overall weight calculation, guiding the algorithm to prioritize coverage. This allows for a more flexible balance between "inspection adequacy" and "route optimization" in edge scenarios. Once an edge (section) is marked as "valid inspection completed," the system reduces its coverage weight to 0, preventing it from being re-routed in subsequent dynamic route planning, thus reducing redundant travel.

[0056] By combining the above-mentioned congestion weight and coverage weight, the present invention can consider both real-time traffic conditions and inspection coverage requirements in the comprehensive weight calculation formula, thereby ensuring efficient completion of gas pipeline network inspection tasks.

[0057] It should be noted that in order to achieve deep integration of gas network GIS and urban road regulations, this system has designed a refined rule modeling method for different types of road restrictions in the path planning model, including but not limited to the following rules: (1) One-way street modeling: In a dynamic road network G = (V, E), define directional restrictions for one-way streets. For example, if the road Edge (i, j) is one-way, only traffic from node i to node j is allowed, and Edge (j, i) is removed.

[0058] (2) Modeling of prohibited road sections: After receiving real-time traffic data, if a road is marked as prohibited due to road closure construction or traffic accidents, the corresponding edge Edge(i,j) is directly removed from the path planning model; at the same time, the upstream and downstream nodes of the affected area (such as nodes i and j) are marked to trigger local path optimization.

[0059] (3) Treatment of dead-end roads and special areas: If a dead-end node is isolated based on real-time traffic data, a virtual edge is inserted between the node and the nearest inspected node in the path planning model, and its comprehensive weight Wvirtual is calculated, where Wvirtual=Wd+Wr+Wt: Wvirtual=Wd+Wr+Wt Recalculate the path based on the heuristic function h(n): h(n)=d(n,g)+∑Wvirtual.

[0060] It should be noted that the traffic congestion weight Wt is updated dynamically. Specifically, the real-time traffic data interface is called to filter out closed or prohibited sections of road. For passable sections of road, the traffic congestion weight Wt is dynamically updated according to the real-time traffic data.

[0061] It is also important to note that if the area where the dead-end road node is located is a high-risk area, it should be given a higher priority to ensure coverage of the inspection task.

[0062] (4) Path planning logic for two-sided gas pipeline networks: If the road width is narrow and the sensor coverage is large enough, the inspection vehicle can complete the inspection of the pipeline network on both sides at one time while driving.

[0063] When the road is wide, the route is further refined based on the road type: bidirectional edges require two passes to cover the pipe network on both sides. The route planning follows the bidirectional logic, covering one side first and then returning to cover the other side.

[0064] One-way roads can only be covered on one side according to the direction of travel.

[0065] In addition, when the risk level is updated, the newly added high-risk points are determined based on the risk levels before and after the update, and the risk weight of the risk node is adjusted in the path planning model.

[0066] Through the above steps, the path planning model can be updated in real time.

[0067] Afterwards, the inspection route is dynamically updated based on the updated path planning model, road rules, intelligent driving inspection vehicle status data, and the risk level and inspection frequency obtained in the previous steps.

[0068] It should be noted that in order to ensure seamless optimization and continuity of the path, the inspection path is dynamically updated using an incremental update method based on the updated path planning model, road rules, intelligent driving inspection vehicle status data, and the risk level and inspection frequency obtained in the previous steps.

[0069] If the inspection data includes information on newly added hidden danger points, the optimal re-inspection path and re-inspection observation point location will be generated based on the external environment data with the newly added hidden danger point as the center and sent to the intelligent driving inspection vehicle. After the intelligent driving inspection vehicle completes the re-inspection of the hidden danger point, it will resume the inspection path before the re-inspection.

[0070] Repeat the above steps until the inspection is completed.

[0071] Specifically, generating the optimal re-inspection path and re-inspection observation point location according to the real-time external environment data includes: generating the optimal re-inspection path according to the real-time external environment data, and generating the re-inspection observation point location according to the real-time external environment data.

[0072] The optimal re-inspection path is generated based on real-time external environment data, including: Taking the potential risk point as the center, a multi-hop search is performed on its neighboring nodes in the path planning model to construct a navigable subgraph. In the subgraph, all "ring" or "loop" paths that can pass through the potential risk point and surrounding key nodes are searched as candidate paths. All candidate paths are traversed and scored using the edge weight formula, path length L, traffic congestion weight Wt, risk weight Wr, and coverage C. All candidate paths are scored and ranked, and the highest-scoring candidate paths are selected. During this process, constraints such as one-way streets and no-U-turn regulations are met by removing illegal edges or restricting the search direction.

[0073] Single-path scenario: If a closed-loop path with the highest score is found to cover the alarm area, the re-inspection path is directly sent to the intelligent driving inspection vehicle; Multi-segment loop combination scenario: If a single loop is insufficient to cover all high-risk sub-areas, several candidate loops with the highest scores will be spliced or executed in segments.

[0074] It should be noted that during the re-inspection process, if new road closures, worsening congestion or expansion of potential hazards are detected again, the dynamic path planning algorithm will be called to update the weights of the affected sections of the road locally, and incremental adjustments will be made to the current circular path to ensure continued high-density coverage of the alarm area; in cases where there are one-way streets or U-turns are not possible, after leaving, the vehicle can immediately return to the vicinity of the potential hazard via the surrounding U-turn nodes to form a closed loop, so as to ensure multiple coverage of the same area as much as possible.

[0075] Generate recheck observation point locations based on external environment data, including: During re-inspection, when the distance to the hidden danger point in the inspection lane is greater than or equal to the set threshold, the inspection vehicle is controlled to travel at the first driving speed v1 (unit: m / s) and the inspection data is sampled according to the preset sampling frequency (unit: Hz = 1 / s). Theoretically, the single sampling interval is approximately Δd=v / f.

[0076] When the distance between the inspection vehicle and the potential hazard point is less than the set threshold (e.g., within 50 meters from the potential hazard point), the inspection vehicle is controlled to travel at the second driving speed v2, where the first driving speed 1 is greater than or equal to the second driving speed v2, for example, v1 is 15 km / h and v2 is 5 km / h. The above adjustment can shorten the sampling interval Δd and achieve intensive sampling.

[0077] When the speed of the inspection vehicle at the potential hazard point exceeds the set threshold again, it can return to the normal re-inspection speed (such as 15 km / h), balancing sampling accuracy and time cost.

[0078] At the same time, wind direction and speed are acquired in real time, and the sampling frequency is increased accordingly. For example, if the primary diffusion direction of the gas changes with wind direction, the original sampling interval along the corresponding route segment will be further shortened (for example, by another 30%) to obtain more intensive concentration data, providing richer input for leak source inversion in the Gaussian diffusion model. If wind direction or wind intensity changes dynamically, the system can synchronously adjust the scope and sampling frequency of these intensified sampling segments during subsequent inspections to ensure that the concentration distribution in key areas is always captured.

[0079] It is understandable that through high-density data collection, multiple rounds of coverage, and dynamic path planning, leak points are re-inspected after the initial inspection to ensure high-quality observation point data, thereby improving the inversion accuracy of the leak source location and intensity, reducing redundant work, and improving overall inspection efficiency. In addition, it is important to note that after triggering the re-inspection task, the intelligent driving inspection vehicle should maintain a low speed.

[0080] It's important to note that during the re-inspection process, if new road restrictions or changes in risk are encountered, the re-inspection route is updated in real time using a dynamic route optimization algorithm. For complex road environments (such as one-way streets and closed areas), the system will flexibly generate a circular route or adjust the route to ensure comprehensive coverage.

[0081] Understandably, in practice, the principle of multiple rounds of coverage is employed: at least two circular or round-trip passes are performed over the alarm area to ensure sufficient sampling data is obtained at different times and under different wind conditions, reducing false alarms or missed alarms caused by "instantaneous interference." If concentration anomalies or significant fluctuations in peak concentrations persist after the initial re-inspection sampling, the patrol vehicle is re-directed to perform another round of re-inspection. If the results of the second round of sampling differ significantly from the previous, a third round of encrypted sampling is triggered for further confirmation.

[0082] Among them, the inspection car uses weighted average or Kalman filtering method to fuse the concentration values obtained from multiple rounds of sampling with the spatial coordinates to obtain the concentration data C i (x i ,y i ) ; The concentration data C i (x i ,y i ) and real-time wind direction and speed are input into the Gaussian diffusion inversion model, and the possible location (x0, y0) and intensity Q of the leakage source are obtained through least squares fitting and returned to the central control platform.

[0083] Specifically, the basic formula of the Gaussian inversion model is as follows: C(x,y): observed concentration at point (x,y); Q: Leakage source intensity; (x0,y0): leak source location; σx,σy: diffusion parameters, related to wind speed, wind direction, and weather data.

[0084] It can be understood that multiple sets of data enable the Gaussian inversion model to be solved by the least squares method or other optimization algorithms, significantly improving the robustness and accuracy of the model.

[0085] If, after multiple rounds of data fusion, the concentration at the potential risk point approaches normal, the central control platform can automatically determine whether it is a false alarm or a transient disturbance, remove the potential risk point with the abnormal alarm, and release the re-inspection alarm. If the concentration still exceeds the standard and the Gaussian inversion results show a clear distribution of the leak source, the next step of the on-site investigation and remediation process will be entered.

[0086] It can be seen that dense sampling and multiple rounds of acquisition significantly improve the number and stability of the constraint equations in the inversion equation, and effectively reduce the positioning error caused by data sparsity or single wind direction.

[0087] In addition, by reducing vehicle speed during inspections and covering the alarm area multiple times, we ensure that enough observation points are obtained to eliminate abnormal sampling points that may be affected by short-term interference (such as car exhaust or other background pollution); and adjust the parameters σx and σy of the diffusion model in combination with real-time meteorological data to match the actual diffusion trend.

[0088] After obtaining the specific location and intensity of the leakage source, the system further outputs the re-inspection path, observation point distribution map, and leakage source positioning results, intuitively displays the coverage of the sampling area, and provides an integrated report of the observation data, including the GPS coordinates, concentration values, wind speed and direction information of each collection point, and instructions for eliminating abnormal points.

[0089] It can be understood that when the inspection data received by the central control platform includes information on potential danger points, it first determines whether it meets the natural gas leakage correlation based on the gas concentration. If so, the point is determined to be a potential danger point. If not, it is determined to be a false alarm point and the re-inspection task is not triggered. After receiving the leakage source location and intensity inverted by the intelligent driving inspection vehicle, it is judged again whether the point is a hidden danger point based on the leakage source location and intensity. If so, it is recorded for emergency investigation.

[0090] As you can understand, before the inspection begins, historical data, gas pipeline network GIS data, pipeline fixed monitoring equipment data, external environmental data, and intelligent driving inspection vehicle status data are collected. Based on this data, the risk level and basic inspection frequency of pipeline nodes and sections are comprehensively assessed. A dynamic path planning model is constructed based on the gas pipeline network GIS data, urban road rules, real-time traffic data, and risk levels. Based on this dynamic path planning model, road rules, intelligent driving inspection vehicle status data, and the risk level, basic inspection frequency, and intelligent driving inspection vehicle status data obtained in the previous steps, a genetic algorithm is used to generate an initial inspection route using the multi-objective optimization criteria of full coverage, high-risk priority, and shortest path. This route is then distributed to the intelligent driving inspection vehicle.

[0091] This embodiment solves key problems in traditional and existing gas inspection methods involving intelligent driving inspection vehicles, such as insufficient task coverage, redundant paths, and insufficient data collection density, through multi-source data fusion, dynamic path adjustment, real-time risk assessment, and multiple rounds of high-precision data collection. It provides an intelligent and efficient solution for modern urban gas safety management.

[0092] Specifically, (1) this embodiment uses a genetic algorithm for global optimization and combines it with a dynamic path optimization algorithm to achieve dynamic adjustment, which can dynamically generate accurate and efficient inspection paths in complex urban road and gas pipeline network environments. By comprehensively covering all pipeline nodes and pipe sections, while avoiding duplicate paths and resource waste, the inspection efficiency and path planning accuracy are greatly improved. By dynamically updating the weights of pipe sections and pipeline nodes, the system can achieve priority coverage of high-risk areas and supplement tasks in uninspected areas, while reducing frequent inspections in low-risk areas.

[0093] (2) This embodiment dynamically updates the weights of pipe sections and network nodes to prioritize coverage of high-risk areas and supplement tasks in uninspected areas, while reducing the frequency of inspections in low-risk areas. Combined with the dynamic risk levels provided by the risk assessment module, the system can achieve scientific optimization of task allocation, significantly improve inspection efficiency, and ensure that inspection resources are accurately allocated to key risk areas.

[0094] (3) This embodiment can quickly identify emergencies (such as road closures and newly added potential hazards) by combining real-time traffic data, construction site information, sensor alarms, and other dynamic factors. Through a dynamic path optimization algorithm, the system can complete path adjustments within 1-2 seconds, ensuring the real-time adaptability and efficient completion of inspection tasks in complex urban environments.

[0095] (4) This embodiment supports multiple rounds of re-inspection and data collection optimization: in the area around the potential hazard point, the re-inspection path is dynamically planned, combining a low-speed inspection strategy with a high-density sampling layout to collect higher-quality sensor data points. With the assistance of real-time wind direction and speed data, the distribution and coverage of observation points are optimized, improving the spatial positioning accuracy of the leak source inversion and the reliability of the diffusion intensity analysis.

[0096] (5) This embodiment uses a Gaussian inversion model and collected high-density concentration data to deduce the spatial location and diffusion intensity of the leak source. By weighted fusion and dynamic adjustment of multiple rounds of collected data, the accuracy of the inversion results is optimized, the false alarm rate caused by short-term interference (such as exhaust pollution) is reduced, and precise positioning guidance is provided for emergency repair tasks, significantly improving emergency repair timeliness and gas safety assurance capabilities.

[0097] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0098] Example 2 Based on the same inventive concept, the embodiment of the present application also provides a municipal gas network inspection system based on GIS, such as Figure 4 As shown, it includes a central control platform and an intelligent driving inspection vehicle, and the central control platform is used to execute the municipal gas pipeline network inspection path planning method described in the first aspect; The intelligent driving inspection vehicle is equipped with an AI pilot module, a communication module, a sensor system, a navigation system and an edge computing unit; The sensor system includes a methane telemeter, a pump-type methane and ethane detector, a vehicle-mounted positioning module, a lidar, a camera, and a mobile weather station. It is used to collect real-time data on gas leakage concentration and the surrounding environment of the pipe network, and generate accurate inspection point data based on the positioning information. The navigation system is used to realize the navigation of the intelligent driving inspection vehicle in a complex road environment based on the high-precision positioning module and the laser radar; The AI navigation module is used to dynamically calculate the pointing position of the methane telemeter based on the gas pipeline network GIS information, risk level distribution and real-time vehicle location, thereby achieving fully automatic scanning of the gas pipeline network; The communication module is used to perform two-way communication with the central control platform to receive task instructions and upload inspection data; The edge computing unit is used to process the sensor data collected by the sensor system in real time and implement preliminary local decision-making functions.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solutions of the present invention. They should all be included in the scope of the technical solutions claimed for protection by the present invention.

Claims

1. A GIS-based municipal gas pipeline inspection route planning method, applied to a central control platform, characterized in that: The following steps are involved: During the inspection process according to the initial inspection route, real-time data from pipeline fixed monitoring equipment, external environment data, status data of intelligent driving inspection vehicles, and inspection data are obtained; If the inspection data does not include information on newly added potential risk points, the risk level and inspection frequency of the pipeline network nodes and sections are dynamically obtained based on the data from the pipeline's fixed monitoring equipment, external environmental data, and the status data of the intelligent driving inspection vehicle. The path planning model is dynamically updated based on the external environmental data and risk level. The inspection path is dynamically updated and sent to the intelligent driving inspection vehicle based on the genetic algorithm, the updated path planning model, road rules, the status data of the intelligent driving inspection vehicle, and the risk level and inspection frequency obtained in the previous steps. If the inspection data includes information about newly added potential danger points, the optimal re-inspection path and re-inspection observation point locations are generated based on the external environment data, with the newly added potential danger point as the center, and sent to the intelligent driving inspection vehicle. After the intelligent driving inspection vehicle completes the re-inspection of the potential danger point, it resumes the inspection path before the re-inspection. Repeat the above steps until the inspection is completed.

2. A municipal gas network inspection route planning method based on GIS according to claim 1, characterized in that: The following formula is used to comprehensively evaluate the risk level of pipeline network nodes and sections: Rv=α·Rhistory+β·Rsensor+γ·Rexternal+δ·Rcoverage Among them, Rv represents the comprehensive risk level of the node or pipe section; Rhistory represents the historical risk value; Rsensor represents the sensor risk value; Rexternal represents the external environment risk value; Rcoverage represents the inspection coverage value; α represents the weight of the historical risk factor; β represents the weight of the real-time sensor data; γ represents the weight of the external environment; δ represents the weight of the inspection coverage status.

3. A municipal gas network inspection route planning method based on GIS according to claim 1 or 2, characterized in that: When dynamically updating the inspection path, an incremental update method is used based on the updated path planning model, road rules, intelligent driving inspection vehicle status data, and the risk level and inspection frequency obtained in the previous steps.

4. A municipal gas network inspection route planning method based on GIS according to claim 3, characterized in that: The path planning model includes a dynamic directed graph G = (V, E) that connects gas pipeline nodes and urban road nodes based on the gas pipeline GIS information, where V represents the node and E represents the connection path. Each edge of the dynamic directed graph is set with a weight: We = α·Wd + β·Wr + γ·Wt + δ·Wtime + λ·Wcoverage. Among them, α, β, γ, and δ are adjustment coefficients; λ represents the adjustment coefficient of the weight factor, which is used to give priority to covering uninspected pipelines; Wr = 1 / Rv represents the risk weight; Wd represents the actual distance between the representative nodes; Wr represents the risk weight, and the weight of high-risk points is lower; Wt represents the real-time traffic congestion weight, and the more severe the congestion, the higher the weight; Wtime represents the time weight, which is used to estimate the time required to complete the task; Wcoverage indicates whether the edge has been covered, with covered being assigned a value of 0 and uncovered being assigned a value of 1.

5. A municipal gas network inspection route planning method based on GIS according to claim 4, characterized in that: During the inspection process, the real-time traffic data interface is called to filter out closed or prohibited sections of roads. For passable sections, the traffic congestion weight Wt is dynamically updated based on the real-time traffic data. For road closures not near pipelines due to construction or traffic accidents, directly remove the corresponding edge E in the path planning model. If a dead-end road node is isolated based on real-time traffic data, a virtual edge is inserted between the node and the nearest inspected node in the path planning model, and its comprehensive weight Wvirtual is calculated, where Wvirtual=Wd+Wr+Wt; Determine the newly added high-risk points based on the risk levels before and after the update, and adjust the risk weight of the risk node in the path planning model.

6. A GIS-based municipal gas network inspection route planning method according to claim 4 or 5, characterized in that: Generate the optimal re-inspection path based on real-time external environment data, including: Taking the potential risk point as the center, a multi-hop search is performed on its neighboring nodes in the path planning model to construct a navigable subgraph. In the subgraph, all "ring" or "loop" paths that can pass through the potential risk point and surrounding key nodes are searched as candidate paths. Traverse all candidate paths, score and sort all candidate paths by combining edge weight formula, path length, traffic congestion weight Wt, risk weight Wr and coverage rate; If a closed-loop path with the highest score is found to cover the alarm area, the re-inspection path will be directly sent to the intelligent driving inspection vehicle; If a single loop is insufficient to cover all high-risk sub-areas, several candidate loops with higher scores will be spliced or executed in segments; During the re-inspection process, if new road closures, worsening congestion, or expanded potential hazards are detected again, the dynamic path planning algorithm will be invoked to update the weights of the affected sections and incrementally adjust the current circular path to ensure continued high-density coverage of the alert area. In the case of one-way streets or where U-turns are not possible, the vehicle can immediately return to the potential hazard via surrounding U-turn nodes after leaving, forming a closed loop to ensure multiple coverage of the same area. Generate recheck observation point locations based on external environment data, including: During re-inspection, when the distance to the hidden danger point in the inspection lane is greater than or equal to the set threshold, the inspection vehicle is controlled to travel at the first driving speed and the inspection data is sampled at the preset sampling frequency; When the distance between the inspection vehicle and the potential hazard point is less than a set threshold, the inspection vehicle is controlled to travel at a second driving speed, wherein the first driving speed is greater than or equal to the second driving speed; At the same time, wind direction and wind speed are obtained in real time, and the sampling frequency is increased according to the wind direction and wind speed.

7. A municipal gas network inspection route planning method based on GIS according to claim 6, characterized in that: After sending the optimal re-inspection path and re-inspection observation point location to the intelligent driving inspection vehicle, the intelligent driving inspection vehicle also receives the leakage source location and intensity returned by the intelligent driving inspection vehicle, and eliminates the potential danger points of abnormal alarms based on the leakage source location and intensity; the leakage source location and intensity are obtained by the intelligent driving inspection vehicle based on the re-inspection data and real-time external environment data using the Gaussian diffusion model inversion.

8. A municipal gas network inspection route planning method based on GIS according to claim 7, characterized in that: The inspection frequency calculation formula is as follows: Fv=Fbase+θ·(1 / Rv)+μ·Sr Where: Fv represents the inspection frequency of node v; Fbase represents the basic inspection frequency; Rv represents the risk level; Sr represents the dynamic risk factor, which indicates the adjustment of the inspection frequency due to real-time changes; θ and μ represent the adjustment coefficients, which are used to balance the impact of risk level and dynamic factors on the inspection frequency.

9. A municipal gas network inspection route planning method based on GIS according to claim 8, characterized in that: In the case of insufficient resources, the inspection frequency of high-risk levels can be adjusted by adjusting the weights of Fbase and θ; In a scenario with sufficient resources, the stability of Fbase is maintained and the inspection frequency is adjusted through the dynamic factor Sr.

10. A GIS-based municipal gas network inspection system, characterized in that: It includes a central control platform and an intelligent driving inspection vehicle, wherein the central control platform is used to execute the municipal gas pipeline network inspection path planning method described in any one of embodiments 1-9; The intelligent driving inspection vehicle is equipped with an AI pilot module, a communication module, a sensor system, a navigation system and an edge computing unit; The sensor system includes a methane telemeter, a pump-type methane and ethane detector, a vehicle-mounted positioning module, a lidar, a camera, and a mobile weather station. It is used to collect real-time data on gas leakage concentration and the surrounding environment of the pipe network, and generate accurate inspection point data based on the positioning information. The navigation system is used to realize the navigation of the intelligent driving inspection vehicle in a complex road environment based on the high-precision positioning module and the laser radar; The AI navigation module is used to dynamically calculate the pointing position of the methane telemeter based on the gas pipeline network GIS information, risk level distribution and real-time vehicle location, thereby achieving fully automatic scanning of the gas pipeline network; The communication module is used to perform two-way communication with the central control platform to receive task instructions and upload inspection data; The edge computing unit is used to process the sensor data collected by the sensor system in real time and implement preliminary local decision-making functions.

Images