Large model-based urban lifeline gas pipeline flood emergency prevention and control system and method

Abstract

The application provides a kind of city lifeline gas pipeline flood emergency prevention and control system and method based on large model, applied to gas pipeline flood emergency technical field.System includes emergency supervision management platform, which is configured to: based on the pipeline pressure and gas flow rate of multiple pipeline nodes, construct gas pipe network atlas, and determine multiple suspected leakage points and leakage confidence;For each suspected leakage point, perform evaluation step, determine sampling radius according to leakage confidence, determine corrosion risk and deformation risk of suspected leakage point based on water body data and image data;Based on multiple corrosion risks, multiple deformation risks, determine the comprehensive risk value of the target pipeline node through the prediction model;In response to the comprehensive risk value meeting the drainage condition, automatically send control signal to the drainage robot to drive the drainage robot to drain water.The application can automatically evaluate risk and emergency disposal, and timely eliminate the safety hazards caused by flood disaster leading to gas pipeline damage.

Description

TECHNICAL FIELD

[0001] The present specification relates to the technical field of gas pipeline flood emergency, and in particular to a city lifeline gas pipeline flood emergency prevention and control system and method based on a large model. BACKGROUND

[0002] Gas pipelines are an important component of city lifeline projects. Long-term soaking and water flow scouring of floods not only accelerate the chemical corrosion of gas pipelines, but also may cause physical deformation of gas pipelines due to uneven settlement of the foundation, thereby significantly increasing the risk of gas leakage and posing a serious threat to public safety.

[0003] Traditional gas pipeline network risk assessment models cannot effectively integrate multiple environmental parameters during floods, such as water depth and water composition, and thus cannot perform real-time and accurate risk assessment.

[0004] Therefore, it is necessary to provide a city lifeline gas pipeline flood emergency prevention and control system and method based on a large model to improve the efficiency and quality of risk monitoring. SUMMARY

[0005] The summary includes a city lifeline gas pipeline flood emergency prevention and control method based on a large model. The method includes: constructing a gas pipeline network atlas based on pipeline pressure and gas flow rate of multiple pipeline nodes, and determining multiple suspected leakage points and leakage confidence; for each suspected leakage point, performing the following steps: determining a sampling radius according to the leakage confidence, and controlling a detection robot to travel to the suspected leakage point to collect water data, controlling a drone to move to the suspected leakage point to collect image data; based on the water data and the image data, determining the corrosion risk and deformation risk of the suspected leakage point; based on multiple corrosion risks and multiple deformation risks of multiple suspected leakage points, determining a comprehensive risk value of a target pipeline node through a prediction model; in response to the comprehensive risk value meeting drainage conditions, automatically sending a control signal to the drainage robot to drive the drainage robot to drain water.

[0006] The summary includes a city lifeline gas pipeline flood emergency prevention and control system based on a large model, which includes an emergency supervision and management platform configured to perform a city lifeline gas pipeline flood emergency prevention and control method based on a large model.

[0007] The beneficial effects brought by the above invention content include but are not limited to: (1) through the city lifeline gas pipeline flood emergency prevention and control system based on a large model, the emergency supervision of the gas pipeline under the flood disaster can automatically perform risk assessment and emergency disposal, and timely eliminate the safety hazards such as gas leakage and pipeline damage caused by flood disaster. (2) By dynamically associating the sampling strategy with the pipe material and soaking time, intelligent and fine management of the data collection process is realized, and the sampling efficiency is significantly improved. (3) By updating the corrosion risk and deformation risk evaluated on site to the gas pipe network atlas as dynamic characteristics, and using GNN model for learning and reasoning, the topological structure information of the pipe network and the spatial correlation of the risk data can be fully utilized to accurately judge the risk. (4) In the early stage of flood, when the flood causes the soil of buried pipeline to be loose or causes the pressure of open pipeline to increase, the early warning of outside pressure anomaly is triggered, and the pipe network pressure is reduced or the valve in the high-risk area is closed in advance, which can reduce the risk of pipeline rupture. BRIEF DESCRIPTION OF DRAWINGS

[0008] The present specification will be further illustrated in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same numbers represent the same structures, wherein:

[0009] Figure 1 is an exemplary system block diagram of a city lifeline gas pipeline flood emergency prevention and control system based on a large model according to some embodiments of the present specification;

[0010] Figure 2 is an exemplary flowchart of a city lifeline gas pipeline flood emergency prevention and control method based on a large model according to some embodiments of the present specification;

[0011] Figure 3 is an exemplary flowchart of a prediction model according to some embodiments of the present specification. DETAILED DESCRIPTION

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some examples or embodiments of the present specification, and for those skilled in the art, the present specification can be applied to other similar scenarios without creating creative labor. Unless it is obvious from the language environment or otherwise stated, the same reference numbers in the drawings represent the same structure or operation.

[0013] It should be understood that the use of “system,” “apparatus,” “unit,” and / or “module” herein is a method of distinguishing different components, elements, parts, sections, or assemblies from one another. However, if other expressions are used, the same meaning can be replaced by other expressions.

[0014] As indicated in the specification and claims, unless the context clearly indicates otherwise, the words “a,” “an,” “one,” and / or “the” are not limited to the singular, but rather include plural, unless the context clearly indicates otherwise. Generally, the terms “comprises,” “comprising,” “includes,” “including,” and the like, specify the presence of stated steps and elements, but do not preclude the presence or addition of one or more other steps or elements.

[0015] Flowcharts are used in the specification to illustrate the operations performed by systems in accordance with embodiments of the present specification. It should be understood that the operations in the front or back are not necessarily performed in the order shown. Instead, various steps can be processed in reverse order, or at the same time. Meanwhile, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0016] Figure 1 is an exemplary system block diagram of a large model-based urban lifeline gas pipeline flood emergency prevention and control system according to some embodiments of the present specification.

[0017] In some embodiments, the large model-based urban lifeline gas pipeline flood emergency prevention and control system 100 includes an emergency supervision management platform 110, an emergency supervision sensor network platform 120, and an emergency supervision object platform 130.

[0018] The emergency supervision management platform 110 is a comprehensive management platform that manages the overall planning and coordination of the contact and cooperation between multiple platforms. The emergency supervision management platform (hereinafter referred to as the management platform) can be configured as a server or a processor. In some embodiments, the emergency supervision management platform 110 communicates with the emergency supervision object platform 130 through the emergency supervision sensor network platform 120.

[0019] In some embodiments, the management platform is configured to construct a gas pipeline network atlas based on the pipeline pressure and gas flow rate of a plurality of pipeline nodes, and determine a plurality of suspected leakage points and a leakage confidence. For each suspected leakage point, an evaluation step is performed, including: determining a sampling radius according to the leakage confidence, and controlling a detection robot and a drone to go to collect water and image data; based on the collected water data and image data, determining the corrosion risk and deformation risk of the suspected leakage point; based on the corrosion and deformation risks of all suspected leakage points, determining a comprehensive risk value of the target pipeline node through a prediction model; in response to the comprehensive risk value satisfying the drainage condition, automatically sending a control signal to the drainage robot to perform the drainage operation.

[0020] For more information about the management platform, please refer to Figures 2-3 and its related description.

[0021] The emergency supervision sensor network platform 120 is used for comprehensive management of sensing information and is a communication transmission platform for bidirectional data interaction between the emergency supervision management platform 110 and the emergency supervision object platform 130. In some embodiments, the emergency supervision sensor network platform 120 can be configured as a communication network or a gateway device. The emergency supervision sensor network platform 120 is responsible for uploading real-time sensing data collected by the emergency supervision object platform to the emergency supervision management platform 110 and issuing control instructions generated by the emergency supervision management platform 110 to the corresponding emergency supervision object platform 130 for execution.

[0022] The emergency supervision object platform 130 is a platform for generating supervision information and executing control information. In some embodiments, the emergency supervision object platform 130 can include various sensors (such as pressure sensors, flow meters) deployed at key nodes of the gas pipeline network, detection robots, drones, drainage robots, valves, and protection electrodes, etc. The emergency supervision object platform 130 is responsible for collecting real-time data of the gas pipeline and the surrounding environment and uploading them through the emergency supervision sensor network platform 120; at the same time, it also responds to the control instructions issued from the emergency supervision management platform 110 to execute specific emergency operations, such as automatically closing valves, starting detection robots, or dispatching drainage robots for drainage.

[0023] The large model refers to a model architecture in the Internet of Things (IoT) field, also known as an Internet of Things model architecture. The Internet of Things model architecture is configured to realize efficient flow and processing of massive data in the system. The Internet of Things model architecture also includes AI models (such as language large models such as ChatGPT), which can be applied to the Internet of Things model architecture to assist in data sensing and processing.

[0024] In some embodiments of the present specification, through the large model-based urban lifeline gas pipeline flood emergency prevention and control system, the emergency supervision of the gas pipeline under flood disasters can automatically perform risk assessment and emergency disposal, and timely eliminate safety hazards such as gas leakage and pipeline damage caused by flood disasters.

[0025] Figure 2 is an exemplary flowchart of the large model-based urban lifeline gas pipeline flood emergency prevention and control method according to some embodiments of the present specification. As Figure 2 shown, the flow 200 includes the following steps 210-240. In some embodiments, the flow 200 can be executed by the management platform every preset period. The preset period can be set by the technician according to experience. For example, the preset period can be 1 min, 5 min, etc.

[0026] At step 210, a gas pipeline network graph is constructed based on the pipeline pressures and gas flow rates of the plurality of pipeline nodes, and a plurality of suspected leakage points and leakage confidence levels are determined.

[0027] A pipeline node refers to a component or a critical position with a specific function in gas transmission or distribution. For example, a pipeline node can refer to a branching point or a converging point of a pipeline, a position where a valve is provided in a pipeline, etc. The management platform can obtain the pipeline nodes input by a technician.

[0028] The pipeline pressure and the gas flow rate can be obtained by a pressure sensor and a flow meter of the emergency supervision object platform.

[0029] The gas pipeline network graph refers to a graph that can reflect the connection relationship of a gas pipeline and valves, pressure sensors, flow meters, etc. connected to the pipeline. In some embodiments, the gas pipeline network graph is a data structure composed of nodes and edges, the edges connecting the nodes, and the nodes having node attributes and the edges having edge attributes.

[0030] In some embodiments, the nodes of the gas pipeline network graph can correspond to the pipeline nodes. For example, the nodes of the gas pipeline network graph correspond to the branching points of each pipeline or each valve. The node attributes can reflect the relevant characteristics of the corresponding pipeline or valve. For example, the node attributes include the pipeline pressure, the gas flow rate, and the valve opening degree of the node, etc. In some embodiments, the node attributes of the node include corrosion risk and deformation risk. For more information about the corrosion risk and the deformation risk, see later.

[0031] In some embodiments, the management platform can obtain the node attributes from the emergency supervision object platform.

[0032] In some embodiments, the edges of the gas pipeline network graph can correspond to the gas pipelines existing between the nodes. The edge is a directed edge, and the direction of the edge reflects the gas flow direction between the nodes. The edge attributes can include pipeline characteristics, such as pipeline length, pipeline material, pipeline age, diameter, and burial depth, etc. The edge attributes can be input by a technician.

[0033] A suspected leakage point refers to a pipeline node with data anomalies, which can possibly leak.

[0034] In some embodiments, the suspected leakage point is obtained by: determining the theoretical pressure and the theoretical flow rate of the next node of each node based on the gas pipeline network graph, and comparing the theoretical pressure and the theoretical flow rate with the actually measured pipeline pressure and gas flow rate of the next node to obtain a pressure difference value and a flow rate difference value; and in response to at least one of the pressure difference value and the flow rate difference value being greater than a corresponding difference threshold value, the management platform determines the next node as a suspected leakage point.

[0035] The next node and the previous node respectively refer to the next and previous pipeline node connected to the node. The next node and the previous node can be determined based on the edges in the gas pipeline network map. The calculation of the theoretical pressure and the theoretical flow rate can be achieved in various ways, for example, it can be simulated by fluid mechanics simulation software, or it can be calculated by fluid mechanics formula (such as Bernoulli equation, Hagen-Poiseuille equation, etc.). The difference threshold corresponding to the pressure difference and the flow rate difference can be preset according to experience.

[0036] If the difference between the actual pressure or flow rate of the gas node and the theoretical value is too large, it indicates that there may be a leak between the gas nodes.

[0037] The leak confidence refers to the possibility of real leak at the suspected leak point. The confidence can be represented by a real number, a percentage, or a level, etc. For example, the confidence can be a value between 0 and 1.

[0038] In some embodiments, after determining the suspected leak points, for each suspected leak point, the management platform constructs a feature vector based on the node attribute of the suspected leak point, the node attribute of the previous node of the suspected leak point, and the edge attribute between the suspected leak point and the previous node; determines the leak confidence of each suspected leak point based on the search result of the feature vector in the vector database.

[0039] The vector database includes a plurality of reference vectors and a leak confidence corresponding to each reference vector. The reference vector is constructed based on the node attribute of the suspected leak point, the node attribute of the previous node of the suspected leak point, and the edge attribute between the suspected leak point and the previous node in the historical data. The management platform can select the leak confidence corresponding to the reference vector with the smallest vector distance between the feature vector and the reference vector as the leak confidence.

[0040] In some embodiments, the vector database can be constructed based on historical pipeline leak information. For example, the leak confidence of the suspected leak point can be the ratio of the maximum number of leak occurrences to the total number of leaks in the historical pipeline leak information. For example, the maximum number of leak occurrences corresponding to all pipeline nodes is 10 times, and the number of leak occurrences of the suspected leak point is 8 times, then the leak confidence is 80%. The historical pipeline leak information can be directly obtained from the management platform.

[0041] In some embodiments, for each suspected leak point, the following evaluation steps 221-222 are performed.

[0042] Step 221, determine the sampling radius according to the leak confidence, and control the detection robot to drive to the suspected leak point to collect water body data, and control the unmanned aerial vehicle to move to the suspected leak point to collect image data.

[0043] The sampling radius refers to the collection range of data collection centered on the suspected leak point.

[0044] In some embodiments, the sampling radius is in a positive correlation with the leakage confidence. The specific proportional relationship can be preset; for example, the leakage confidence is 0.8, and the sampling radius is 30 meters; the leakage confidence is 0.2, and the sampling radius is 2 meters. The higher the leakage confidence, the greater the possibility of leakage at the point, and therefore a larger range of sampling is required, and the sampling radius is also larger accordingly.

[0045] In some embodiments, after determining the sampling radius, the management platform can determine the sampling quantity and the sampling position. For example, the sampling quantity can be a fixed value, and the sampling position is the position when the sampling quantity is arranged at equal intervals.

[0046] In some embodiments, the management platform determines the sampling quantity and the sampling position of the plurality of sampling points within the sampling radius based on the pipe material and the water soaking time of the suspected leakage point; automatically controls the detection robot and the unmanned aerial vehicle to sample based on the sampling quantity and the sampling position of the plurality of sampling points.

[0047] The water soaking time refers to the cumulative duration of the pipe being soaked by the accumulated water due to the flood disaster. In some embodiments, the water soaking time can be calculated through the data of the water level sensor. The emergency supervision sensor network platform is also deployed with a water level sensor (for example, a floating ball sensor), which is used to monitor the depth of accumulated water. When the sensor data of the water level sensor shows accumulated water, the management platform accumulates the time, thereby obtaining the water soaking time.

[0048] In some embodiments, the management platform can determine the sampling quantity by querying a first preset table based on the pipe material and the water soaking time. The first preset table includes the sampling quantity corresponding to different pipe materials and water soaking times. The first preset table can be constructed by technical personnel based on experience. For example, only as an example, when constructing the first preset table, the sampling quantity is in a positive correlation with the corrosion resistance of the pipe material and the water soaking time.

[0049] After determining the sampling quantity, the management platform sets the ratio of the sampling radius to the sampling quantity as the sampling interval, and then determines the sampling position of each sampling point accordingly.

[0050] In some embodiments of the present specification, by dynamically associating the sampling strategy with the pipe material and the water soaking time, intelligent and fine management of the data collection process is achieved, and the sampling efficiency is significantly improved.

[0051] The detection robot is a robot that can move in water or on land and collect water samples. For example, the detection robot is a robot with remote control, navigation, obstacle avoidance, and data transmission functions, and is provided with various water monitoring sensors. The detection robot is configured to obtain water data.

[0052] Water body data refers to data related to the water body of the flood. In some embodiments, the water body data can include water body conductivity, water body composition and proportion, water body depth distribution. The water body conductivity reflects the corrosion ability of the water body to the pipeline. The water body depth distribution (also known as water level depth distribution, underwater topographic map) can be drawn by the detection robot when moving on the water surface or underwater by constantly recording three-dimensional coordinate points.

[0053] The unmanned aerial vehicle is configured to acquire image data. The unmanned aerial vehicle includes a visible light camera and an infrared imager. The visible light camera is configured to acquire a sequence of water surface images. The infrared imager is configured to acquire an infrared thermal imaging image.

[0054] In some embodiments, the image data includes a sequence of water surface images taken by the visible light camera on the unmanned aerial vehicle, and an infrared thermal imaging image taken by the infrared imager.

[0055] In some embodiments, after determining the sampling position, the management platform controls the detection robot to drive to the sampling position and collect water body data; controls the unmanned aerial vehicle to move to the sampling position and collect image data.

[0056] Step 222, based on the water body data and the image data, determining the corrosion risk and the deformation risk of the suspected leakage point.

[0057] The corrosion risk refers to the risk of pipeline damage caused by chemical factors.

[0058] In some embodiments, the management platform processes the sequence of water surface images through image recognition technology to determine the water surface bubble proportion; based on the water surface bubble proportion, the water body conductivity and the water body composition and proportion, the corrosion risk is determined by querying a second preset table.

[0059] The water surface bubble proportion refers to the proportion of the area of the water surface bubbles to the water surface. The more the water surface bubble proportion, the stronger the impact and corrosiveness of the surface flood. The image recognition technology is prior art and will not be described here.

[0060] The second preset table includes different water surface bubble proportions, water body conductivities and water body composition and proportions corresponding to the corrosion risk. The second preset table can be constructed by technicians based on experience. For example only, when constructing the second preset table, the corrosion risk and the water surface bubble proportion, the water body conductivity and the proportion of the corrosive component in the water body composition and proportion are all in a proportional relationship.

[0061] Through the water surface bubble proportion, the water body conductivity and other multi-modal data, a relatively accurate corrosion risk can be obtained.

[0062] The deformation risk refers to the risk of pipeline damage caused by physical factors. In some embodiments, the management platform can determine the deformation risk of the suspected leakage point in multiple ways based on the water body data and the image data.

[0063] In some embodiments, the management platform determines the deformation risk based on the water depth distribution and the infrared thermal imaging image.

[0064] In some embodiments, the management platform can determine the deformation risk by consulting a third preset table based on the water depth distribution and the infrared thermal imaging image. The third preset table includes different water depth distributions and infrared thermal imaging images and corresponding deformation risks. The third preset table can be constructed by technicians based on experience. By way of example only, the ground subsidence value is determined by comparing the water depth distribution in the historical data with the water depth distribution initially obtained, and the greater the ground subsidence value, the higher the deformation risk; when the area of the low-temperature region of the infrared thermal imaging image in the historical data and the temperature (the gas leakage point is characterized by local low temperature due to rapid expansion of gas and heat absorption) are the largest and the lowest, respectively, the deformation risk is the highest.

[0065] At step 230, the comprehensive risk value of the target pipeline node is determined by a prediction model based on the plurality of corrosion risks and the plurality of deformation risks of the plurality of suspected leakage points.

[0066] The prediction model is a model for predicting the comprehensive risk value of the target pipeline node. In some embodiments, the prediction model can be a graph neural network (GNN) model.

[0067] The target pipeline node refers to a pipeline node for which the comprehensive risk value is desired to be determined. The target pipeline node can include all nodes in the pipeline node except the suspected leakage points. In some embodiments, the target pipeline node can also refer to a supplementary acquisition node, and the two can be interchangeable. For more description of the supplementary acquisition node, see the following.

[0068] The comprehensive risk value refers to the possibility of damage (such as corrosion, deformation, etc.) of the pipeline under the current flood environment. The corrosion risk, the deformation risk, and the comprehensive risk value can all be expressed in percentage to quantify the risk. The corrosion risk can also be referred to as a corrosion risk value, and the deformation risk can also be referred to as a deformation risk value.

[0069] Figure 3 is an exemplary flowchart of the prediction model according to some embodiments of the present specification.

[0070] In some embodiments, the node attributes of the gas pipeline network graph also include corrosion risks and deformation risks. The management platform updates the gas pipeline network graph based on the plurality of corrosion risks and the plurality of deformation risks of the plurality of suspected leakage points. The management platform updates the corrosion risks and the deformation risks of the plurality of suspected leakage points determined in the previous step to the corresponding node attributes in the gas pipeline network graph as new node attributes.

[0071] The input of the prediction model 320 includes the updated gas pipeline network atlas 311 and the target pipeline node 312. In some embodiments, the input of the prediction model 320 can only include the updated gas pipeline network atlas 311, and the target pipeline node 312 is the entire pipeline node of the gas pipeline network atlas 311 except the suspected leakage point by default. In other embodiments, the target pipeline node 312 needs to be input separately, for example, the target pipeline node 312 is only the target pipeline node determined based on the node importance degree hereinafter. The output of the prediction model 320 includes the comprehensive risk value of the target pipeline node determined based on the node output.

[0072] In some embodiments, the prediction model can be trained by a large number of first training samples with first labels. For example, the management platform can input a plurality of first training samples into an initial prediction model, construct a loss function based on the output of the initial prediction model and the labels, and iteratively update the parameters of the initial prediction model based on the loss function. When the iteration completion condition is met, the iteration is ended, and the trained prediction model is obtained. The method of iterative update includes but is not limited to gradient descent method, and the iteration completion condition can be convergence of the loss function or the number of iterations reaching a threshold.

[0073] The first training sample can be obtained based on historical data. The first training sample includes a historical gas pipeline network atlas and a historical target pipeline node in the historical data. The first label is the actual comprehensive risk value corresponding to each node in the first training sample. The first label can be constructed by: normalizing the true corrosion depth and the true deformation size corresponding to each node in the first training sample into dimensionless values and then weighting and summing them to obtain the first label of each node. The true corrosion depth and the true deformation size can be obtained from historical pipeline leakage information or historical maintenance records.

[0074] In some embodiments, the comprehensive risk value can also include regional risk values of a plurality of regions. The regional risk value refers to the comprehensive risk value corresponding to each region. The plurality of regions can be pre-set by technical personnel based on administrative region division, etc. In some embodiments, the management platform determines the plurality of regions based on the plurality of corrosion risks, the plurality of deformation risks, and the gas pipeline network atlas. For more information about determining the plurality of regions, see later.

[0075] In some embodiments, the management platform is also configured to determine the comprehensive risk value of the target pipeline node simultaneously by the prediction model. The comprehensive risk value includes regional risk values of a plurality of regions. The input of the prediction model also includes the plurality of regions, and the output includes the regional risk values of the regions determined based on the node output. The first training sample also includes historical regions in the historical data; and the corresponding first label also includes the actual comprehensive risk value of the region, which can be the average value of the actual comprehensive risk values of a plurality of pipeline nodes corresponding to the region. For more information about the regions and the regional risk values, see later.

[0076] In response to the integrated risk value satisfying the drainage condition, a control signal is automatically sent to the drainage robot to drive the drainage robot to drain water.

[0077] The drainage condition refers to a condition for controlling the drainage robot to drain water. In some embodiments, the drainage condition is that the integrated risk value of the target pipeline node is higher than a risk threshold. The risk threshold can be preset.

[0078] In some embodiments, the control signal includes a drainage power of the drainage robot, and the drainage robot drains water based on the drainage power.

[0079] In some embodiments, the management platform determines a supplementary collection node in response to the number of suspected leakage points or the dispersion of the suspected leakage points satisfying a supplementary collection condition. For the supplementary collection node, the same evaluation steps as for the suspected leakage points are performed. For more information about the evaluation steps for the suspected leakage points, see the related description of steps 221-222.

[0080] The dispersion of the suspected leakage points refers to the degree of concentration of the suspected leakage points. The dispersion can be obtained by the management platform by calculating the variance of the coordinates of all suspected leakage points based on the coordinates of all suspected leakage points. The smaller the variance, the lower the dispersion.

[0081] The supplementary collection condition includes that the number of suspected leakage points is less than a preset number threshold, or the dispersion thereof is less than a preset dispersion threshold. The number threshold and the dispersion threshold can be set by technicians based on experience.

[0082] In some embodiments, the management platform sets nodes within a preset distance (such as 3-5 meters) around the suspected leakage points as the supplementary collection nodes, and the preset distance is set by technicians based on experience. In some embodiments, the management platform determines the supplementary collection nodes based on the node importance. For more information about the node importance and the supplementary collection nodes, see later.

[0083] Too few suspected leakage points or too concentrated distribution of the suspected leakage points means that collecting data only for these high-risk points may not provide sufficient comprehensive and representative information for subsequent risk inference of the entire pipe network. Therefore, when the supplementary collection condition is satisfied, the management platform can ensure global coverage of the collected data by determining the supplementary collection nodes, and avoid the accuracy of the subsequent prediction model from being reduced due to sample bias.

[0084] In some embodiments of the present specification, by updating the corrosion risk and deformation risk evaluated by field collection as dynamic features into the gas pipe network atlas, and using the GNN model for learning and reasoning, the topological structure information of the pipe network and the spatial correlation of the risk data can be fully utilized to accurately judge the risk.

[0085] In some embodiments, the management platform is further configured to determine at least one of the target pipeline node or the supplementary collection node based on the node importance.

[0086] The node importance refers to the importance of a certain pipeline node.

[0087] The atlas node importance refers to the importance of a node in the gas pipeline network atlas. The atlas node importance is related to the gas pipeline network atlas.

[0088] In some embodiments, the atlas node importance is in direct proportion to the number of edges in the gas pipeline network atlas and the number of paths passing through the node, and is in inverse proportion to the path length of the node. The specific proportional relationship can be preset. For example, when all the shortest paths are counted, if a node appears the most number of times, the number of edges connected to the node is the largest, and the total length of the path length of the node to all other nodes is the shortest, then the node importance of the node is the highest.

[0089] The node dynamic risk value refers to the risk value of the amplification of the negative impact of the node after the node fails under the flood environment. The node dynamic risk value can be determined by weighting the sum of the betweenness centrality of the node in the gas pipeline network atlas and the normalized processing of the node failure flood simulation to a dimensionless value. The normalization processing can include Min-Max normalization, etc.

[0090] The betweenness centrality of the node can indicate the influence of the node in the gas pipeline network atlas. The betweenness centrality of the node is determined by the following way: determining all node pairs in the gas pipeline network atlas (i.e., a pair of two nodes selected at random); for each node pair, determining the shortest path (i.e., the shortest path in the path of the two nodes selected at random); for a certain node, counting the number of times the node appears in the shortest path, and the betweenness centrality of the node is the ratio of the number of times to the number of paths of all shortest paths.

[0091] The node failure flood simulation refers to the simulation of the leakage range of the leaked gas after the node leaks under the change of the flood. The node failure flood simulation can be obtained by using the fluid mechanics simulation software in the prior art, such as ANSYS Fluent, Flow-3D, etc.

[0092] For example, for a certain pipeline node, the larger the betweenness centrality of the node and the larger the leakage range of the node failure flood simulation, the larger the node dynamic risk value.

[0093] The node importance is determined based on the atlas node importance and the node dynamic risk value. In some embodiments, the node importance is the sum of the atlas node importance and the node dynamic risk value. In other embodiments, the node importance is the average of the atlas node importance and the node dynamic risk value.

[0094] By considering the node dynamic risk value, not only a single node is considered, but also whether the negative influence of the node after failure will spread and amplify to the downstream or other areas through water flow and topology in the flood environment.

[0095] The target pipeline node or the supplementary collection node can be a node with high node importance. Considering the node importance, the risk of important nodes can be determined in priority or the important nodes can be collected in priority, thereby improving the efficiency.

[0096] In some embodiments, the management platform is further configured to determine at least one of the target pipeline node or the supplementary collection node based on the gas pipeline network graph, the map information, and the node dynamic risk value through a node selection model.

[0097] The node selection model refers to a model for determining the target pipeline node. In some embodiments, the node selection model is a graph neural network model.

[0098] The input of the node selection model includes the gas pipeline network graph and the map information, and the node attribute of the gas pipeline network graph contains the node dynamic risk value. The output of the node selection model includes one or more target pipeline nodes determined based on the node output. The map information includes the building density of each pipeline node in the plurality of pipeline nodes. The building density refers to the density of buildings around the node. The building density can be obtained based on public map information. For more information about the gas pipeline network graph, see Figure 2 .

[0099] In some embodiments, the node selection model can be trained through a plurality of second training samples with second labels. The training process of the node selection model is similar to the training process of the prediction model, see Figure 3 for related content.

[0100] The second training sample includes a sample gas pipeline network graph, sample map information, and sample node dynamic risk value, and the second training sample can be obtained based on historical data. The second label includes the target pipeline node and the supplementary collection node corresponding to the second training sample. The second label can be manually labeled. For example, a node actually leaking in the sample gas pipeline network graph is determined as the target pipeline node.

[0101] It can be understood that the way of determining the supplementary collection node can be the same as the way of determining the target pipeline node, which will not be repeated here.

[0102] In some embodiments, the second label is determined based on a relationship between a building density of the second training sample and a density threshold, and a relationship between a node dynamic risk value and a dynamic risk threshold. For example, a pipe node that satisfies both a condition that the building density of the node is greater than the density threshold and a condition that the node dynamic risk value is greater than the dynamic risk threshold is labeled as a target pipe node. For more information about the node dynamic risk value, see the foregoing.

[0103] In some embodiments, the density threshold and the dynamic risk threshold are inversely proportional to a historical rainfall of the second training sample and a number of abnormal times of a historical abnormal outside pressure of the second training sample. The specific proportional relationship can be preset. For example, for each node of the second training sample, the greater the historical rainfall corresponding to the node and the higher the number of abnormal times, the smaller the density threshold and the dynamic risk threshold. The historical rainfall can be obtained through public weather data.

[0104] The outside pressure of the pipe refers to the external force on the outer wall of the pipe. The outside pressure of the pipe can be obtained by a pressure sensor arranged on the outer wall of the pipe.

[0105] When the outside pressure exceeds a preset range, it is considered that the outside pressure of the pipe is abnormal. The preset range can be preset by a technician based on experience. The number of abnormal times of the historical abnormal outside pressure can be obtained based on the historical abnormal outside pressure statistics. For example, the statistics period can be hours or days, and 1 or 1 day is counted as 1 time.

[0106] By optimizing the second label of the second training sample, the node selection model can be optimized so that the node selection model outputs suitable target pipe nodes.

[0107] In some embodiments, the management platform is further configured to: determine a plurality of regions based on the plurality of corrosion risks, the plurality of deformation risks, and the gas pipe network map, and determine a comprehensive risk value of the target pipe node through the prediction model; and in response to the comprehensive risk value satisfying a drainage condition and a region risk value of the region being greater than a region risk value threshold, drive the drainage robot to move to a region center of the region in need of drainage to drain water. The comprehensive risk value of the target pipe node further includes the region risk value of the plurality of regions. The region and the region risk value are in one-to-one correspondence.

[0108] In some embodiments, the region includes a region center and a region internal node.

[0109] Determining a plurality of regions based on the plurality of corrosion risks, the plurality of deformation risks, and the gas pipe network map includes: determining a plurality of regions based on the plurality of corrosion risks and the plurality of deformation risks of the plurality of suspected leakage points in units of each node of the gas pipe network map; each cluster obtained by clustering is a region, and the cluster center of each cluster is the region center of the region. The nodes included in the region are the region internal nodes of the region.

[0110] In some embodiments, the clustering further comprises weighting the normalized corrosion risk and deformation risk of the suspected leakage points before clustering, and clustering based on the weighting of both. The clustering algorithm comprises K-Means clustering, density-based clustering method (DBSCAN), etc.

[0111] In the node comparison of the plurality of comprehensive risk values, only relying on the comprehensive risk value of the node may cause the drainage robot to be too concentrated; by dividing a plurality of regions and determining the region risk value corresponding to the plurality of regions, the region center can be quickly located, and the drainage robot can be reasonably distributed.

[0112] In some embodiments, the region risk value threshold corresponding to each region is different, and the region risk value threshold of each region is related to the pipeline age, the pipeline length and the historical rainfall of the region corresponding to the region.

[0113] The pipeline age and the pipeline length corresponding to the plurality of regions can be determined by the gas pipe network atlas. The pipeline age corresponding to each region can be the average pipeline age of the plurality of pipelines in the region. The pipeline length corresponding to each region can be the total length of the plurality of pipelines in the region.

[0114] For more description of this part, see Figure 2 and Figure 3 related content.

[0115] In some embodiments, the management platform can determine the region risk value threshold corresponding to the plurality of regions by querying a fourth preset table based on the pipeline age, the pipeline length and the historical rainfall of the region corresponding to the plurality of regions. The fourth preset table comprises the region risk value threshold corresponding to the pipeline age, the pipeline length and the historical rainfall of the plurality of regions. The fourth preset table can be constructed by technicians based on experience. For example only, when constructing the fourth preset table, the region risk value threshold of each region and the pipeline age, the pipeline length and the historical rainfall of the region corresponding to the region are in inverse relationship, that is, the longer the pipeline age, the longer the pipeline length and the greater the historical rainfall of the region corresponding to the region, the smaller the region risk value threshold of the region.

[0116] The longer the pipeline age, the longer the pipeline length, and the greater the historical rainfall, the more likely the region risk value of the region meets the condition, so that the drainage of these regions can be performed as soon as possible, which can avoid pipeline damage; at the same time, the region with large historical rainfall is likely to further increase the water accumulation, and the drainage of the region as soon as possible can prevent the expansion of flood disasters.

[0117] In some embodiments, the drainage power of the drainage robot is positively correlated with the comprehensive risk value of the target pipeline node and the area risk value of the area. The drainage power is positively proportional to the comprehensive risk value of the target pipeline node or the area risk value of the area. The specific proportional relationship can be preset. For example, when the comprehensive risk value of the target pipeline node or the area risk value of the area is the highest (e.g., 95%), the drainage power is the maximum drainage power.

[0118] By setting the drainage power matched with the risk, the drainage can be performed as early as possible for the high-risk area, thereby reducing the impact of flooding.

[0119] In some embodiments, the management platform is further configured to: in response to the corrosion risk of the suspected leakage point being greater than a corrosion threshold, control the drainage robot to move to the corresponding suspected leakage point and start a protection electrode; the protection electrode is electrically connected with the pipeline wall of the suspected leakage point. The corrosion threshold can be preset by a technician based on experience.

[0120] The protection electrode refers to a device for reducing corrosion through electrochemical means such as cathodic protection. The protection electrode forms a closed circuit with the pipeline wall of the suspected leakage point through electrical connection, thereby changing the electrochemical state of the pipeline metal surface and reducing the corrosion of the liquid to the pipeline wall. The protection electrode can be provided on the drainage robot. For example only, the protection electrode is provided on the side of the drainage robot, and when the drainage robot moves to the vicinity of the pipeline wall of the suspected leakage point, the protection electrode is in close contact with the pipeline wall, and the protection electrode is started after.

[0121] In some embodiments, the management platform is further configured to: before determining the plurality of suspected leakage points and the leakage confidence, in response to the outside pressure anomaly of the pipeline, send a control signal to a target valve to reduce the opening degree of the target valve to reduce the pipeline pressure; the target valve is located at an upstream node of the pressure sensor corresponding to the outside pressure anomaly.

[0122] The target valve refers to a valve whose pipeline pressure at the position corresponding to the outside pressure anomaly can be reduced after reducing the opening degree of the valve. The opening degree of the target valve refers to the opening size of the target valve, corresponding to the size of the pipeline flow.

[0123] The management platform can determine the nearest upstream node to the pressure sensor corresponding to the outside pressure anomaly among the nodes provided with the valve as the target valve. The nearest upstream node can be determined based on the gas pipeline map. For example, the gas flow direction is node A to node B to node C, the pressure sensor corresponding to the outside pressure anomaly is located at node C, node A is provided with a valve, and node B is not provided with a valve, then the target valve is node A.

[0124] In the early stage of flood, when the flood causes the soil of the buried pipeline to be loose or causes the pressure of the open pipeline to increase, the early warning of the outside pressure anomaly is triggered, and the pipeline network pressure is reduced or the valve in the high-risk area is closed in advance, so that the risk of pipeline rupture can be reduced.

[0125] The foregoing detailed description has set forth various embodiments of the application via the use of specific terminology. As such, the description herein is not intended to limit the application, but rather to describe various embodiments of the application. It is to be understood that the use of certain specific language (or the absence thereof) does not limit the scope of the application.

Claims

1. A large model-based urban lifeline gas pipeline flood emergency prevention and control system, characterized in that, The system comprises an emergency supervision management platform; The emergency supervision management platform is configured to: construct a gas pipeline network graph based on pipeline pressure and gas flow rate of a plurality of pipeline nodes, and determine a plurality of suspected leakage points and leakage confidence; for each suspected leakage point, perform the following evaluation steps: determine a sampling radius according to the leakage confidence, and control a detection robot to drive to the suspected leakage point to collect water body data and control a drone to move to the suspected leakage point to collect image data; determine corrosion risk and deformation risk of the suspected leakage point based on the water body data and the image data; in response to the number of suspected leakage points or the dispersion degree of the suspected leakage points satisfying a supplementary collection condition, determine a supplementary collection node; for the supplementary collection node, perform the same evaluation steps as for the suspected leakage point; determine a comprehensive risk value of a target pipeline node through a prediction model based on a plurality of corrosion risks and a plurality of deformation risks of a plurality of suspected leakage points; in response to the comprehensive risk value of the target pipeline node satisfying a drainage condition, automatically send a control signal to a drainage robot to drive the drainage robot to drain water.

2. The system of claim 1, wherein, The emergency supervision management platform is further configured to: determine at least one of the target pipeline node or the supplementary collection node based on node importance; wherein the node importance is determined based on graph node importance and node dynamic risk value.

3. The system of claim 2, wherein, The emergency supervision management platform is further configured to: determine at least one of the target pipeline node or the supplementary collection node based on the gas pipeline network graph, map information, and the node dynamic risk value through a node selection model; wherein the map information includes building density of each pipeline node in the plurality of pipeline nodes; The node selection model is a graph neural network model.

4. The system of claim 1, wherein, The gas pipeline network graph includes the plurality of pipeline nodes; The emergency supervision management platform is further configured to: update the gas pipeline network graph based on the plurality of corrosion risks and the plurality of deformation risks of a plurality of suspected leakage points; the prediction model is a graph neural network model; and the input of the prediction model includes the updated gas pipeline network graph and the target pipeline node.

5. The system of claim 1, wherein, The emergency supervision management platform is further configured to: determine a plurality of regions based on the plurality of corrosion risks, the plurality of deformation risks, and the gas pipeline network graph, and determine the comprehensive risk value of the target pipeline node through the prediction model; the comprehensive risk value includes a plurality of regional risk values of a plurality of regions; in response to the comprehensive risk value satisfying the drainage condition and the regional risk value of the plurality of regions being greater than a regional risk value threshold, drive the drainage robot to move to a regional center of a region requiring drainage to drain water.

6. The system of claim 5, wherein, The regional risk value threshold corresponding to each region is different, and the regional risk value threshold is related to pipeline age, pipeline length, and historical rainfall corresponding to the region.

7. A large model-based urban lifeline gas pipeline flood emergency prevention and control method, executed by an emergency supervision and management platform, characterized in that, The method comprises: constructing a gas pipeline network graph based on pipeline pressure and gas flow rate of a plurality of pipeline nodes, and determining a plurality of suspected leakage points and leakage confidence; for each suspected leakage point, performing the following steps: According to the leakage confidence, a sampling radius is determined, and a detection robot is controlled to drive to the suspected leakage point to collect water body data, and a UAV is controlled to move to the suspected leakage point to collect image data; Based on the water body data and the image data, corrosion risk and deformation risk of the suspected leakage point are determined; In response to the number of suspected leakage points or the dispersion degree of the suspected leakage points satisfying a supplementary collection condition, a supplementary collection node is determined, and for each supplementary collection node, the same steps as each of the suspected leakage points are performed; Based on a plurality of corrosion risks and a plurality of deformation risks of a plurality of suspected leakage points, a comprehensive risk value of a target pipeline node is determined through a prediction model, and in response to the comprehensive risk value satisfying a drainage condition, a control signal is automatically sent to a drainage robot to drive the drainage robot to drain water.

8. The method of claim 7, wherein, At least one of the target pipeline node or the supplementary collection node is determined based on a node importance, and the node importance is determined based on a graph node importance and a node dynamic risk value.

Images