Natural gas pipeline network collaborative planning method and system based on machine learning

Abstract

The application discloses a natural gas pipeline network collaborative planning method and system based on machine learning, and relates to the technical field of intelligent pipeline network planning. The method comprises the following steps: S1, collecting pipeline network expansion collaborative planning data and performing data preprocessing; S2, matching construction window reference operation data segments, evaluating node pressure margin, pipe segment load rate, remaining pressure regulating capacity of pressure regulating stations and construction window gas load level; updating pipeline network topology, identifying standby gas supply channels and evaluating the remaining capacity of the standby channels; S3, counting construction window reference operation characteristics, training a construction window gas supply disturbance risk prediction model, obtaining a comprehensive construction disturbance risk value, and identifying an optimization construction stage; S4, determining a recommended construction window and generating a valve switching collaborative scheme. The method solves the problem that the existing natural gas pipeline network expansion planning is difficult to quantify the construction window disturbance risk in the planning stage, resulting in that the static planning scheme is difficult to form an executable collaborative planning scheme in the actual construction stage.

Description

Technical Field

[0001] This invention relates to the field of intelligent pipeline planning technology, specifically to a collaborative planning method and system for natural gas pipelines based on machine learning. Background Technology

[0002] With the continuous advancement of digital construction of urban gas infrastructure, data on natural gas pipeline network planning, operation monitoring, engineering construction, and user gas consumption are gradually forming multi-source, large-scale, and online data resources. Relying on internet data services such as big data resource services, databases, and cloud database services, combined with cloud computing software, cloud-integrated application operation support platform software, and data mining software, information such as pipeline topology, operating conditions, construction plans, and user loads can be centrally managed and collaboratively analyzed. Simultaneously, the application of intelligent analysis technologies such as machine learning in complex engineering data modeling and risk identification is constantly deepening, providing an important data processing and intelligent decision-making foundation for collaborative planning of natural gas pipeline networks.

[0003] For example, Chinese patent CN109002419B discloses a method and apparatus for dynamic analysis of natural gas pipeline networks. The method includes: establishing node dynamic equations for a specified natural gas pipeline network based on the mass conservation equation, and establishing pipeline dynamic equations for the natural gas pipeline network based on the momentum conservation equation; determining a directed adjacency matrix for the nodes based on the node dynamic equations and the adjacency matrix, and determining a directed adjacency matrix for the pipelines based on the pipeline dynamic equations and the transpose of the adjacency matrix; coupling the node directed adjacency matrix and the pipeline directed adjacency matrix to obtain a state-space model of the natural gas pipeline network; and performing dynamic analysis of the natural gas pipeline network based on the state-space model. This application's embodiments can achieve dynamic analysis of large-scale complex pipeline networks with high efficiency.

[0004] For example, Chinese patent CN104731761B discloses a method and apparatus for simulating natural gas pipeline networks. The method includes: discretizing the control equations and boundary condition equations for natural gas flow and heat transfer in the natural gas pipeline network to obtain a system of algebraic equations; adjusting the grid points in the grid system at time i based on the comparison results of interpolated and numerical solutions of the flow parameter values ​​at the grid points at time i-1; adjusting the time step at time i based on the numerical solutions of the flow parameter values ​​at each grid point in the grid system from the previous three time steps and the time step; and solving the system of algebraic equations based on the grid points in the grid system at time i and the time step to determine the numerical solutions of the flow parameter values ​​at each grid point in the grid system at time i. This achieves adaptive adjustment of the time step and the spatial grid system, reducing waste of computer memory resources and computation time while ensuring computational accuracy.

[0005] However, existing natural gas pipeline expansion plans typically use predetermined demand as input, determining pipeline routes, diameters, and pressure regulating facility configurations based on hydraulic calculations under steady-state or peak operating conditions, and using the gas transmission capacity and pressure satisfaction during the planning period as the primary evaluation criteria. However, in the actual expansion implementation phase, the connection of new pipeline sections, the renovation of old pipelines, valve switching, and the expansion and renovation of pressure regulating stations often need to be completed in stages while the existing pipeline network continues to supply gas. Within the construction window, operational constraints such as temporary topology changes, valve position changes, and phased limitations on pressure regulating capacity are introduced. When the selected window coincides with peak gas consumption, insufficient pressure margin at key nodes, or limited backup channel capacity, failure modes such as exceeding minimum pressure limits at nodes, overloading of bottleneck sections due to detours in the gas supply path, insufficient backup channels, and expanded impact range of temporary gas outages are prone to occur. Because the aforementioned construction stage constraints and disturbance risks lack quantifiable assessment and comparison mechanisms during the planning stage, some planning schemes, while meeting gas transmission capacity requirements in static verification, struggle to form executable segmented connection and switching plans during the construction implementation phase, thus affecting project progress and gas supply reliability.

[0006] Therefore, in order to address the above problems, there is an urgent need for a collaborative planning method and system for natural gas pipeline networks based on machine learning. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a machine learning-based collaborative planning method and system for natural gas pipeline networks. This solves the problem that existing natural gas pipeline network expansion plans struggle to quantify construction window disturbance risks during the planning stage, making it difficult for static planning schemes to form executable segmented access and valve switching plans during the actual construction phase.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a machine learning-based collaborative planning method for natural gas pipeline networks, comprising the following steps: S1, collecting collaborative planning data for pipeline expansion within the same natural gas pipeline network and preprocessing the data; S2, extracting the time attributes of the construction phase from the collaborative planning data for pipeline expansion, matching reference operation data segments for the construction window based on the time attributes, evaluating node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, and gas load level of the construction window based on the reference operation data segments for the construction window; and updating the pipeline topology based on the collaborative planning data for pipeline expansion and identifying backup gas supply. S3, statistically analyze the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, gas load level of the construction window, and remaining capacity of the backup channel to obtain the reference operating characteristics of the construction window. Combine the pipeline expansion collaborative planning data corresponding to the historical construction stage with the reference operating characteristics of the construction window to train the gas supply disturbance risk prediction model of the construction window. Based on the gas supply disturbance risk prediction model of the construction window, obtain the comprehensive construction disturbance risk value and identify the construction stage to be optimized. S4, for the construction stage to be optimized, generate candidate construction windows and determine the recommended construction windows, and simultaneously screen candidate connecting valves to generate valve switching collaborative schemes.

[0011] Furthermore, the specific process of collecting pipeline expansion collaborative planning data within the same natural gas pipeline network and performing data preprocessing on the pipeline expansion collaborative planning data is as follows: Collecting pipeline expansion collaborative planning data for the same natural gas pipeline network includes: collecting basic topology data of the natural gas pipeline network based on the natural gas pipeline network GIS, which includes: pipeline segment number, pipeline segment starting node number, pipeline segment ending node number, node number, valve number, valve's associated pipeline segment number, pressure regulating station number, and pressure regulating station connection node number; collecting natural gas pipeline network operation data based on the SCADA system, pressure sensors, flow meters, and pressure regulating station monitoring system, which includes: node operating pressure, pipeline segment flow rate, pressure regulating station outlet pressure, pressure regulating station pressure regulating flow rate, and collection timestamp; and collecting user gas consumption data from the gas company's user management system, which includes: user node number, user gas consumption, user historical peak gas consumption, and collection timestamp. Expansion construction plan data is collected from the project planning management system. This data includes: the proposed expansion section number, the proposed connection node number, the construction stage number, the estimated construction start time, the estimated construction end time, the affected pipeline section number, the affected valve number, the planned opening / closing status of the affected valve, and the affected pressure regulating station number. Based on the pipeline section number, node number, and pressure regulating station number, they are associated with the corresponding natural gas pipeline network operation data. The natural gas pipeline network operation data and user gas consumption data are processed for time uniformity according to the collected timestamps, and a unified sampling time is generated according to the preset sampling period to form a unified time series data. When there are more than two records corresponding to the same object number in the same sampling period, the record closest to the sampling time is selected as the data for the corresponding sampling time for the object number. If the pressure and flow data exceed the range of the corresponding sensor, they are marked as outliers and removed. Missing data is filled in using linear interpolation.

[0012] Furthermore, the specific process of extracting the time attributes of the construction phase based on the pipeline expansion collaborative planning data, matching the construction window reference operation data segments according to the time attributes, and evaluating the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, and gas load level of the construction window based on the construction window reference operation data segments is as follows: Based on the construction phase number, the expected construction start time, and the expected construction end time, determine the expected construction window corresponding to each construction phase number, and extract the time attributes of the expected construction window. The time attributes include seasonal attributes, date type, and intraday time period. Based on the time attributes of the expected construction window, select historical data segments with the same time attributes as the expected construction window from the unified time series data to construct the construction window reference operation data segments corresponding to the construction phase. The operational data segment includes: node number, node operating pressure, pipe segment number, pipe segment flow rate, pressure regulating station number, pressure regulating station outlet pressure, pressure regulating station flow rate, user node number, and user gas consumption. Based on the operational data segment referenced during the construction window, the node pressure margin, pipe segment load rate, and remaining pressure regulating capacity of the pressure regulating station are calculated. Specifically, the node pressure margin is the difference between the node operating pressure and the node's minimum allowable pressure; the pipe segment load rate is the ratio between the pipe segment flow rate and the pipe segment's maximum allowable flow rate; and the remaining pressure regulating capacity of the pressure regulating station is the difference between the pressure regulating station's maximum pressure regulating capacity and its flow rate. Simultaneously, for each user node number in the operational data segment referenced during the construction window, the ratio between the average user gas consumption and the average of the user's historical peak gas consumption is calculated to obtain the gas load level for the construction window.

[0013] Furthermore, the specific process of updating the pipeline topology based on pipeline expansion collaborative planning data, identifying backup gas supply channels, and assessing the remaining capacity of backup channels is as follows: A natural gas pipeline topology map is constructed based on the pipe segment number, the starting node number of the pipe segment, and the ending node number of the pipe segment. The connectivity of the edges in the natural gas pipeline topology map is updated based on the pipe segment number to which the valve belongs, the valve number affected by construction, and the planned opening / closing status of the valve affected by construction. If the planned opening / closing status of the valve affected by construction is closed, the corresponding pipe segment number is marked as an unaccessible pipe segment in the natural gas pipeline topology map and removed, resulting in a ventilable topology map. Based on the pipe segment number affected by construction, the pipe segments affected by construction are numbered in the ventilable topology map. The pipe sections marked with numbers indicating construction restrictions are removed, resulting in a construction window reference topology map. Within this map, the connection node numbers of the pressure regulating station are used as the starting node set, and the node numbers to be connected are used as the target nodes. A breadth-first search algorithm is employed to find pipe path segments connected to the node numbers to be connected, and these found pipe path segments are designated as backup gas supply channels. For each backup gas supply channel, the pipe section numbers within the channel are read, and the difference between the maximum allowable flow rate and the flow rate of each segment is calculated to obtain the remaining flow capacity of each segment. The minimum remaining flow capacity among all segments within the same backup gas supply channel is taken as the remaining capacity of the backup gas supply channel.

[0014] Furthermore, the specific process for statistically analyzing node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, gas load level during construction windows, and remaining capacity of the backup channel to obtain the reference operating characteristics of the construction windows is as follows: Using the construction stage number as a unit, read the reference operating data segments of the N most recent construction windows corresponding to the construction stage number; take the starting node number and ending node number of the pipeline segment corresponding to the proposed access node number and the construction-affected pipeline segment number as the node statistical objects; for each node statistical object, calculate the node pressure margin at each unified sampling time, and take the minimum value to obtain the minimum node pressure margin; take the construction-affected pipeline segment number and the pipeline segment number within the backup gas supply channel as the pipeline segment statistical objects; for each pipeline segment statistical object, calculate the reference operating characteristics at each unified sampling time. The maximum pipe section load rate is obtained by taking the maximum value. The pressure regulating station numbers affected by construction are used as statistical objects for pressure regulating stations. For each statistical object, the remaining pressure regulating capacity of the pressure regulating station at each unified sampling time is calculated, and the minimum value is taken to obtain the minimum remaining pressure regulating capacity of the pressure regulating station. The set of user nodes affected by construction is determined based on the proposed connection node number, the pipe section number affected by construction, the backup gas supply channel, and the natural gas pipeline network topology. For each user node affected by construction, the gas load level of the construction window is calculated, and the average value is taken to obtain the average gas load level of the construction window. The backup gas supply channel is used as a statistical object for backup channels. For each statistical object for backup channels, the remaining capacity of the backup channel at each unified sampling time is calculated, and the minimum value is taken to obtain the minimum remaining capacity of the backup channel.

[0015] Furthermore, the specific process of training the gas supply disturbance risk prediction model for construction windows by combining the pipeline expansion collaborative planning data corresponding to historical construction stages and the reference operation characteristics of construction windows is as follows: Using construction stage numbers as units, a construction stage disturbance prediction feature vector is generated based on the most recent N construction window reference operation data segments. The construction stage disturbance prediction feature vector includes: construction stage number, proposed expansion segment number, proposed connection node number, construction-affected segment number, construction-affected valve number, planned opening / closing status of construction-affected valves, construction-affected pressure regulating station number, minimum node pressure margin, maximum segment load rate, minimum remaining pressure regulating capacity of the pressure regulating station, average gas load level of the construction window, number of backup gas supply channels, and minimum remaining capacity of the backup channel; based on the construction stage disturbance prediction feature vectors corresponding to historical construction stages, a construction disturbance training sample set is constructed; for each historical construction stage number in the construction disturbance training sample set, a training sample label is generated, and the training samples... The labels include node pressure exceeding limit labels, pipe segment overload labels, and insufficient backup channel labels. During the subsequent actual execution of the historical construction phase, if the node pressure margin of any node in the corresponding node statistics is less than the pressure margin threshold, the node pressure exceeding limit label of the corresponding historical construction phase number is marked as 1; otherwise, it is marked as 0. During the subsequent actual execution of the historical construction phase, if the load rate of any pipe segment in the corresponding pipe segment statistics is greater than the load threshold, the pipe segment overload label of the corresponding historical construction phase number is marked as 1; otherwise, it is marked as 0. If there is no backup gas supply channel in the historical construction phase, or if the minimum remaining capacity of the backup channel is less than the backup capacity threshold, the insufficient backup channel label of the corresponding historical construction phase number is marked as 1; otherwise, it is marked as 0. The construction disturbance training sample set is used as the model input feature, and the training sample labels are used as the model output labels. The gradient boosting tree classification model is trained to construct a construction window gas supply disturbance risk prediction model.

[0016] Furthermore, the process of obtaining the comprehensive construction disturbance risk value based on the construction window gas supply disturbance risk prediction model and identifying the construction stage to be optimized is as follows: The construction stage disturbance prediction feature vector of the current construction stage number is input into the construction window gas supply disturbance risk prediction model, which outputs the probability of node pressure exceeding limits, pipeline overload, and insufficient backup channels corresponding to the construction stage number. The maximum value among these three probabilities is taken as the comprehensive construction disturbance risk value. When the comprehensive construction disturbance risk value is less than or equal to the optimization trigger threshold, the expected construction start time and expected construction end time corresponding to the construction stage number remain unchanged, resulting in a recommended construction window, and the planned opening and closing status of the valves affected by the construction remains unchanged. When the comprehensive construction disturbance risk value is greater than the optimization trigger threshold, the construction stage number is marked as the construction stage to be optimized.

[0017] Furthermore, for the construction stage to be optimized, the specific process of generating candidate construction windows and determining recommended construction windows is as follows: For the construction stage to be optimized, a window adjustment plan is generated: Based on the expected construction start time, expected construction end time, and preset window adjustment step size, a set of candidate construction windows is generated; and each candidate construction window maintains the same duration as the original expected construction window; for each candidate construction window in the set of candidate construction windows, the N most recent construction window reference operation data segments with the same time attributes as the candidate construction window are selected from the unified time series data, and the corresponding construction stage disturbance prediction feature vector is generated, which is input into the gas supply disturbance risk prediction model of the construction window to obtain the corresponding comprehensive construction disturbance risk value; the candidate construction window with the smallest comprehensive construction disturbance risk value in the set of candidate construction windows is determined as the optimal candidate construction window, and the optimal candidate construction window is determined as the recommended construction window.

[0018] Furthermore, the specific process of simultaneously screening candidate connected valves to generate a valve switching coordination scheme is as follows: For the construction stage to be optimized, a valve switching coordination scheme is generated: valves in the backup gas supply channel that are not affected by the construction are selected as candidate connected valves and marked as openable; a construction coordination optimization scheme is generated and output based on the window adjustment scheme and the valve switching coordination scheme.

[0019] The second aspect of this invention provides a machine learning-based natural gas pipeline network collaborative planning system, comprising: a pipeline network data preprocessing module for collecting pipeline expansion collaborative planning data within the same natural gas pipeline network and performing data preprocessing on the pipeline expansion collaborative planning data; a window feature recognition module for extracting time attributes of the construction phase based on the pipeline expansion collaborative planning data, matching reference operation data segments for construction windows based on the time attributes, and evaluating node pressure margin, pipeline section load rate, remaining pressure regulating capacity of pressure regulating stations, and gas load level of construction windows based on the reference operation data segments for construction windows; updating the pipeline network topology based on the pipeline expansion collaborative planning data, identifying backup gas supply channels, and evaluating backup channels. The system includes: a pipeline remaining capacity module; a disturbance risk prediction module, which statistically analyzes node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating stations, gas load level of construction windows, and remaining capacity of standby channels to obtain reference operating characteristics of construction windows; a gas supply disturbance risk prediction model for construction windows, which is trained by combining historical pipeline expansion collaborative planning data corresponding to construction stages and reference operating characteristics of construction windows; a comprehensive construction disturbance risk value is obtained based on the gas supply disturbance risk prediction model for construction windows, and construction stages to be optimized are identified; and a window valve optimization module, which generates candidate construction windows and determines recommended construction windows for construction stages to be optimized, and simultaneously filters candidate connecting valves to generate valve switching collaborative schemes.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention collects basic topology data, operation data, user gas consumption data and expansion construction plan data within the same natural gas pipeline network and performs unified time series processing, so that pipeline expansion planning no longer relies solely on static design parameters, but can be combined with actual operating pressure, pipeline flow, pressure regulating station status and construction plan constraints for collaborative analysis, thereby improving the integrity and consistency of the expansion construction planning data base.

[0023] (2) According to the time attributes of the expected construction window, the reference operation data segment of the construction window is matched, and the reference operation characteristics of the construction window are formed by combining the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station and the gas load level of the construction window. It can identify the gas supply carrying capacity under different construction windows before construction, and avoid blindly implementing expansion and access during peak gas consumption or periods of insufficient pressure margin.

[0024] (3) The present invention updates the topology of the natural gas pipeline network based on the planned opening and closing status of valves affected by construction and the pipeline sections affected by construction, and identifies the backup gas supply channels and their remaining capacity through breadth-first search. It can quantify the detour gas supply capacity under temporary topology changes during construction, thereby discovering the risk of insufficient backup channels and overload of bottleneck pipeline sections in advance.

[0025] (4) This invention uses the construction stage disturbance prediction feature vector corresponding to the historical construction stage to train the gas supply disturbance risk prediction model, outputs the probability of node pressure exceeding the limit, the probability of pipeline overload and the probability of insufficient backup channel, and generates window adjustment scheme and valve switching coordination scheme based on the comprehensive construction disturbance risk value. It can directly transform the construction disturbance risk prediction results into an executable construction coordination optimization scheme, thereby improving the feasibility and gas supply reliability of the natural gas pipeline network expansion construction stage.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 The flowchart shows a machine learning-based collaborative planning method for natural gas pipeline networks.

[0028] Figure 2 This is a structural diagram of a machine learning-based collaborative planning system for natural gas pipeline networks.

[0029] Figure 3 A schematic diagram showing the ranking of comprehensive construction disturbance risk values ​​for candidate construction windows;

[0030] Figure 4 This is a flowchart of a machine learning-based collaborative planning process for natural gas pipeline networks. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. As those skilled in the art will understand, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figures 1-4 This invention provides a technical solution: a machine learning-based collaborative planning method for natural gas pipeline networks, such as... Figure 1 As shown, the process includes the following steps: S1, collecting pipeline expansion collaborative planning data within the same natural gas pipeline network and preprocessing the data; S2, extracting the time attributes of the construction phase based on the pipeline expansion collaborative planning data, matching the construction window reference operation data segments according to the time attributes, and evaluating the node pressure margin, pipeline section load rate, remaining pressure regulating capacity of the pressure regulating station, and gas load level of the construction window based on the construction window reference operation data segments; and updating the pipeline topology based on the pipeline expansion collaborative planning data, identifying backup gas supply channels, and evaluating the remaining capacity of the backup channels; S3, for nodes... The pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, gas load level of the construction window, and remaining capacity of the backup channel are statistically analyzed to obtain the reference operating characteristics of the construction window. Combined with the pipeline expansion collaborative planning data corresponding to the historical construction stage and the reference operating characteristics of the construction window, a gas supply disturbance risk prediction model for the construction window is trained. Based on the gas supply disturbance risk prediction model for the construction window, a comprehensive construction disturbance risk value is obtained, and the construction stage to be optimized is identified. S4, for the construction stage to be optimized, candidate construction windows are generated and recommended construction windows are determined. At the same time, candidate connecting valves are screened to generate valve switching collaborative schemes.

[0033] Specifically, the process of collecting pipeline expansion collaborative planning data within the same natural gas pipeline network and preprocessing this data is as follows: Collecting pipeline expansion collaborative planning data for the same natural gas pipeline network. The same natural gas pipeline network refers to a natural gas pipeline network topology system managed by the same gas supply dispatching system and possessing unified pipe segment numbers, node numbers, valve numbers, and pressure regulating station numbers. This ensures that subsequent data from different sources can be correlated and analyzed within the same pipeline network topology. Specifically, this includes: collecting basic topology data of the natural gas pipeline network based on the natural gas pipeline network GIS. This basic topology data includes: pipe segment number, pipe segment starting node number, and pipe segment ending node number. The data includes node numbers, valve numbers, valve segment numbers, pressure regulating station numbers, and pressure regulating station connection node numbers. The natural gas pipeline network GIS pre-stores the spatial relationships and connections between pipe segments, nodes, valves, and pressure regulating stations. Pipe segment numbers uniquely identify gas transmission segments within the natural gas pipeline network. The starting and ending node numbers of pipe segments determine their connection relationships within the network. Node numbers uniquely identify pipe segment junctions, user access points, pressure regulating station access points, and valve control association points. The valve segment number determines the location of the pipe segment controlled by the valve. The pressure regulating station connection node number determines the access location of the pressure regulating station within the natural gas pipeline network topology. This basic topology data forms the data foundation for subsequent pipeline topology construction, locating construction-affected pipe segments, and searching for backup gas supply channels. Natural gas pipeline network operation data is collected using a SCADA system, pressure sensors, flow meters, and a pressure regulating station monitoring system. This data includes: node operating pressure, pipeline segment flow, pressure regulating station outlet pressure, pressure regulating station regulating flow, and a collection timestamp. Node operating pressure is collected by pressure sensors installed at pipeline nodes and uploaded via the SCADA system. Pipeline segment flow is collected by flow meters installed on the corresponding pipeline segments and uploaded via the SCADA system. Pressure regulating station outlet pressure is collected by pressure sensors installed at the outlet pipeline of the pressure regulating station, and pressure regulating station regulating flow is collected by flow meters installed at the outlet pipeline of the pressure regulating station and uploaded via the pressure regulating station monitoring system. The pressure regulating station monitoring system summarizes and records the outlet pressure and regulating flow of the pressure regulating station, generating pressure regulating station operation data with collection timestamps. The collection timestamps represent the actual collection time of the corresponding operation data. Through this operation data, the pressure distribution, flow distribution, and pressure regulating station operation status of the same natural gas pipeline network at different sampling times can be obtained.Gas consumption data is collected from the gas company's user management system. This data includes user node number, user gas consumption, historical peak gas consumption, and collection timestamp. The user node number represents the user's gas supply access node in the natural gas pipeline network topology, and a correspondence is established between the user node number and the node number in the basic natural gas pipeline network topology data through a user node mapping relationship. This mapping relationship includes the correspondence between user numbers, user node numbers, and node numbers. The user node mapping relationship is obtained from the gas company's user management system, user metering records, and user access point information in the pipeline network GIS, and is used to determine the actual gas supply access location corresponding to each user number. When user gas consumption data is stored using user numbers as an index, the user number is first converted to a user node number according to the user node mapping relationship, and then the user node number is associated with the node number in the basic natural gas pipeline network topology data, thereby mapping the user's gas consumption to the corresponding pipeline node. Through this mapping relationship, the user's gas load can be accurately associated with the node location in the natural gas pipeline network topology, providing a data foundation for subsequent determination of the user node set affected by construction, calculation of gas load levels during construction windows, and prediction of construction disturbance risks. User gas consumption refers to the actual gas consumption of the corresponding user node number at the data collection timestamp; user historical peak gas consumption refers to the maximum historical gas consumption of the corresponding user node number within a preset historical statistical period. The preset historical statistical period is preferably 30 to 180 days before the current expected construction window, used to cover recent gas consumption fluctuations and seasonal gas consumption changes, and serves as a benchmark for comparing gas load levels in subsequent construction windows. Through the correspondence between user node numbers and node numbers, user gas consumption data can be mapped to specific gas supply locations in the natural gas pipeline network topology.The expansion construction plan data is collected from the project planning management system. This data includes: the proposed expansion section number, the proposed connection node number, the construction stage number, the estimated construction start time, the estimated construction end time, the construction-affected section number, the construction-affected valve number, the planned opening / closing status of the construction-affected valve, and the construction-affected pressure regulating station number. Specifically, the proposed expansion section number indicates the pipeline segment to be added, modified, or connected in this expansion project; the proposed connection node number indicates the location where the proposed expansion section will connect to the existing natural gas pipeline network; the construction stage number distinguishes different construction stages within the same expansion project; the estimated construction start time and estimated construction end time determine the expected construction window for the corresponding construction stage; the construction-affected section number, construction-affected valve number, and construction-affected pressure regulating station number limit the pipeline network objects that will cause topology changes and operational capacity changes during the corresponding construction stage. These numbers are pre-determined by the project planning management system during the construction plan preparation stage. The planned opening / closing status of the construction-affected valve is the valve's planned status during the construction period, pre-determined by the project planning management system based on the construction organization plan and scheduling operation plan. The construction-affected pipeline segment number is pre-determined based on the proposed expansion pipeline segment number, the proposed connection node number, the existing pipeline connection relationship, and the construction sequence. It indicates existing pipeline segments that require connection, cutting, modification, shutdown, pressure reduction, detour gas supply, or temporary gas supply restriction during the corresponding construction phase. The construction-affected valve number is pre-determined based on the construction-affected pipeline segment number, the valve's corresponding pipeline segment number, the connection relationship between the nodes at both ends of the construction-affected pipeline segment, the connection relationship between adjacent pipeline segments, and construction isolation requirements. It indicates valves that need to be opened or closed during the corresponding construction phase. The construction-affected valve number includes the valve number on the upstream side of the construction-affected pipeline segment, the valve number on the downstream side of the construction-affected pipeline segment, and the branch valve number adjacent to the construction-affected pipeline segment and requiring isolation control. The construction-affected pressure regulating station number is pre-determined based on the proposed connection node number, the construction-affected pipeline segment number, the pressure regulating station connection node number, the upstream gas supply path, and the downstream gas supply range. It indicates pressure regulating stations whose outlet pressure, pressure regulating flow, and gas supply range are affected by construction during the corresponding construction phase. The aforementioned expansion construction plan data allows for the mapping of construction stages, construction windows, affected objects, and pipeline topology objects. Based on pipe segment numbers, node numbers, and pressure regulating station numbers, corresponding natural gas pipeline operation data are associated. Specifically, pipe segment numbers are associated with pipe segment flow rates, node numbers with node operating pressures, and pressure regulating station numbers with pressure regulating station outlet pressures and pressure regulating flow rates, thus forming operation data records under the same object number.The natural gas pipeline network operation data and user gas consumption data are processed for time uniformity based on the collection timestamp, and a unified sampling time is generated according to the preset sampling period to form unified time series data. The preset sampling period is set according to the natural gas pipeline network operation monitoring frequency, preferably 5 to 30 minutes, suitable for the collaborative analysis of pipeline operation pressure, flow, pressure regulating station data, and user gas consumption data. The unified sampling time is used to align data from different sources and with different sampling frequencies to the same time reference. When there are more than two records corresponding to the same object number within the same sampling period, the record closest to the sampling time is selected as the data for the object number at the corresponding sampling time. The object number includes the node number, pipeline segment number, pressure regulating station number, and user node number. When the time distance between two or more records and the sampling time is the same, the record with an earlier collection time and the closest sampling time is selected first. If there are still two or more records that meet the above conditions, the record with the earlier data entry time is selected as the data for the object number at the corresponding sampling time. The above methods avoid duplicate or conflicting calculations caused by multiple records of the same object at the same sampling time, and ensure the uniqueness and reproducibility of data preprocessing results in scenarios with records at the same distance. If pressure and flow data exceed the corresponding sensor range, they are marked as outliers and removed. The corresponding sensor range is derived from the measurement point configuration parameters in the sensor equipment ledger. The pressure and flow data include node operating pressure, pipe flow, pressure regulating station outlet pressure, and pressure regulating station flow. The corresponding sensor range includes pressure sensor range and flow meter range, used to limit the effective acquisition range of node operating pressure, pipe flow, pressure regulating station outlet pressure, and pressure regulating station flow. Missing data is filled using linear interpolation. When there is missing data between adjacent valid sampling times, linear interpolation is performed according to the changing trend of the corresponding data at adjacent valid sampling times to obtain the completed data at the corresponding unified sampling time, thereby ensuring the continuity and computability of the unified time series data. Linear interpolation is applicable when valid sampled data exist for the same object number before and after the missing sampling time, and the continuous missing duration is less than or equal to the maximum continuous missing duration threshold. The maximum continuous missing duration threshold is set based on the unified sampling period, the normal sensor upload interval, and the allowable completion accuracy of the pipeline operation data, preferably 2 to 6 times the unified sampling period. When the continuous missing duration exceeds the maximum continuous missing duration threshold, linear interpolation is not performed on the corresponding missing data, and the data of the corresponding object number within that missing period is marked as non-interpolable missing data to avoid the completion result deviating from the actual operating state due to long-term missingness. When the missing data is located at the starting boundary of the unified time series data, the data from the next valid sampling time is used for forward boundary completion. When the missing data is located at the ending boundary of the unified time series data, the data from the previous valid sampling time is used for backward boundary completion.For object numbers and sampling times marked as non-interpolable missing data, they will not be included in the calculation of node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating station and gas load level of construction window at the corresponding sampling time. When the proportion of non-interpolable missing data in a certain construction window reference operation data segment is greater than the preset missing proportion threshold, the construction window reference operation data segment will be removed and will not be used as the data basis for subsequent construction window reference operation characteristic statistics and gas supply disturbance risk prediction model training.

[0034] This implementation plan, by limiting the scope to the same natural gas pipeline network, uniformly collects data on basic topology, operating conditions, user gas consumption, and expansion construction plans. It also performs object association, time alignment, anomaly removal, and missing data completion on multi-source data, forming a continuous, unified, and computable data foundation for collaborative planning of pipeline expansion. Furthermore, it clarifies the collection sources, object numbering relationships, sampling period ranges, and effective measurement ranges of various data types, avoiding problems such as difficulty in associating data from different systems, inconsistent sampling times, and interference from abnormal data in subsequent analysis. This improves the accuracy, completeness, and feasibility of construction window disturbance risk assessment and collaborative optimization calculations.

[0035] Specifically, the process of extracting the time attributes of the construction phase based on the collaborative planning data of pipeline expansion, matching the reference operation data segments of the construction window according to the time attributes, and evaluating the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, and gas load level of the construction window based on the reference operation data segments of the construction window is as follows: According to the construction phase number, the expected construction start time, and the expected construction end time, the expected construction window corresponding to each construction phase number is determined. For the same construction phase number, the expected construction start time recorded in the project planning management system is used as the start time of the expected construction window, and the expected construction end time recorded in the project planning management system is used as the end time of the expected construction window. The expected construction window for the corresponding construction phase number is formed from the start time to the end time of the expected construction window. The expected construction window is a continuous time interval for the planned construction operation of the corresponding construction phase, which is used to define the time benchmark for subsequent historical operation data matching, operation feature extraction, and construction disturbance risk prediction. The time attributes of the expected construction window are extracted, including seasonal attributes, date type, and intraday time period. The seasonal attribute is determined according to the month in which the expected construction window begins: preferably March to May is spring, June to August is summer, September to November is autumn, and December to February of the following year is winter. The date type is determined according to the calendar attribute corresponding to the expected construction window start date, preferably including weekdays and non-working days, with non-working days including rest days and statutory holidays. The intraday time period is determined according to the time period to which the expected construction window begins, including the hour and minute information corresponding to the expected construction window start time and the hour and minute information corresponding to the expected construction window end time. Based on the time attributes of the expected construction window, historical data segments with the same time attributes as the expected construction window are selected from the unified time series data. These historical data segments are preferably those prior to the expected construction start time and sharing the same seasonal attributes, date type, and intraday time period as the expected construction window. The historical data segments should maintain the same duration as the expected construction window to construct a reference operational data segment for the construction window corresponding to the construction phase. For example, if the expected construction window for a certain construction phase is from 8:00 to 18:00 on July 15, 2026, then the duration of this expected construction window is 10 hours, its seasonal attribute is summer, and its date type is weekday. Therefore, historical data segments prior to the expected construction start time, belonging to summer and a weekday from 8:00 to 18:00, are selected from the unified time series data, and each historical data segment is truncated to a duration of 10 hours to serve as the reference operational data segment for the construction window. The reference operation data segments for the construction window include: node number, node operating pressure, pipe segment number, pipe segment flow rate, pressure regulating station number, pressure regulating station outlet pressure, pressure regulating station pressure regulating flow rate, user node number, and user gas consumption.The construction window reference operation data segment is used to characterize the historical pipeline network operation status under similar time attributes to the expected construction window, and serves as the reference data basis for predicting construction disturbance risks during the current construction phase. Based on the construction window reference operation data segment, node pressure margin, pipeline section load rate, and remaining pressure regulating capacity of the pressure regulating station are calculated. The node pressure margin is the difference between the node's operating pressure and its minimum allowable pressure. The minimum allowable pressure is obtained by reading the operation constraint parameters of the corresponding node number in the pipeline network operation configuration parameter table. The minimum allowable pressure represents the minimum pressure limit that the corresponding node must maintain to meet normal gas supply to users and ensure safe operation of the pipeline network. The node pressure margin characterizes the safety margin of the corresponding node from the minimum gas supply pressure requirement. The pipeline network operation configuration parameter table is a parameter table pre-established by the gas company based on pipeline design data, equipment ledgers, operation scheduling procedures, and safety operation standards. It stores the operation constraint parameters corresponding to each pipeline object. The pipeline network operation configuration parameter table uses node number, pipeline section number, and pressure regulating station number as index fields, recording the minimum allowable pressure of the node, the maximum allowable flow rate of the pipeline section, and the maximum pressure regulating capacity of the pressure regulating station. The minimum allowable pressure of a node is set according to the user's gas supply pressure requirements and the pipeline network's safe operation requirements. The maximum allowable flow rate of a pipe section is set according to the pipe diameter, design pressure rating, material, allowable flow velocity, and operational safety margin. The maximum pressure regulating capacity of a pressure regulating station is set according to the rated flow rate of the pressure regulating station equipment, the outlet pressure control range, and the operational safety margin. The pipe section load rate is the ratio between the pipe section flow rate and the maximum allowable flow rate of the pipe section. The maximum allowable flow rate of the pipe section is obtained by reading the operational constraint parameters of the corresponding pipe section number in the pipeline network operation configuration parameter table. The maximum allowable flow rate of the pipe section represents the maximum flow rate that the corresponding pipe section is allowed to carry under the conditions of meeting the design pressure rating, pipe diameter, material, and operational safety requirements. The pipe section load rate is used to characterize the degree of occupancy of the corresponding pipe section's transport capacity. The remaining pressure regulating capacity of a pressure regulating station is the difference between its maximum pressure regulating capacity and its pressure regulating flow rate. The maximum pressure regulating capacity is obtained by reading the operating constraint parameters of the corresponding pressure regulating station number in the pipeline network operation configuration parameter table. The maximum pressure regulating capacity represents the maximum pressure-regulated gas supply flow rate that the corresponding pressure regulating station can handle under conditions that meet equipment rated capacity, outlet pressure control requirements, and safe operation. The remaining pressure regulating capacity of a pressure regulating station characterizes the margin of capacity that the corresponding pressure regulating station can continue to undertake pressure regulation and gas supply adjustment within the construction window. Simultaneously, for each user node number in the reference operation data segment of the construction window, the ratio between the average user gas consumption and the average of the user's historical peak gas consumption is calculated to obtain the gas load level of the construction window.Among them, the average gas consumption of users is the average gas consumption of users at each unified sampling time within the reference operation data segment of the construction window for the corresponding user node number; the historical peak gas consumption of users is the maximum value of user gas consumption at each sampling time within the preset historical statistical period according to the unified sampling period for the corresponding user node number; the average historical peak gas consumption of users is the average historical peak gas consumption of users at each unified sampling time within the reference operation data segment of the construction window for the corresponding user node number; the window gas load level is used to characterize the degree to which the average gas load of the corresponding user node under similar time conditions in the expected construction window is close to the recent historical peak gas consumption state; when the gas load level of the construction window is larger, it indicates that the gas demand of the corresponding user node in the construction window is closer to the recent historical high load state.

[0036] In this implementation plan, by determining the expected construction window according to the construction stage and matching historical data segments with the same time attributes and duration based on seasonal attributes, date type, and intraday time period, a reference operation data segment for the construction window that is closer to the actual gas consumption pattern can be constructed for the current construction stage. At the same time, based on the reference operation data segment, the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, and gas consumption load level of the construction window are evaluated. This allows the safety margin of gas supply pressure, pipeline load status, pressure regulating capacity margin, and the degree to which user load approaches peak under the construction window to be quantified and characterized, thereby improving the data relevance and reliability of the prediction of construction disturbance risk.

[0037] Specifically, the process of updating the pipeline topology based on pipeline expansion collaborative planning data, identifying backup gas supply channels, and assessing the remaining capacity of backup channels is as follows: A natural gas pipeline topology map is constructed based on the pipeline segment number, the starting node number of the pipeline segment, and the ending node number of the pipeline segment. The natural gas pipeline topology map uses node numbers as graph nodes and pipeline segment numbers as graph edges connecting two graph nodes. The starting node number and the ending node number of the pipeline segment are used to determine the connection relationship between the two ends of the corresponding graph edge. This topology map can characterize the connectivity structure between pipeline segments and nodes within the same natural gas pipeline network, providing a basis for identifying gas supply paths during subsequent construction. The connectivity status of the edges in the natural gas pipeline network topology map is updated based on the pipe segment number to which the valve belongs, the valve number affected by construction, and the planned opening / closing status of the valve affected by construction. The pipe segment number to which the valve belongs is used to determine the location of the pipe segment controlled by the valve affected by construction, and the planned opening / closing status of the valve affected by construction indicates the planned status of the valve affected by construction in the construction organization plan during the corresponding construction stage. When updating the connectivity status of the edges, the pipe segment number to which the valve belongs corresponding to the valve affected by construction is used as the object to be updated, and the planned opening / closing status of the valve affected by construction is used as the basis for judging whether the edge is ventilable. If the planned opening / closing state of the valve affected by construction is closed, the corresponding pipe segment number is marked as a non-ventilable pipe segment and removed from the natural gas pipeline network topology map to obtain a ventilable topology map. The pipe segment corresponding to the valve's pipe segment number is the pipe segment with the construction-affected valve as the ventilation control boundary, and the closure of the construction-affected valve can block the effective ventilation connection of the corresponding pipe segment within the construction window. Specifically, when the construction-affected valve is located at the starting end, ending end, or middle isolation position of the corresponding pipe segment, and there is no bypass branch that can bypass the construction-affected valve to maintain connection between the two ends of the same pipe segment, it is determined that closing the construction-affected valve can make the corresponding pipe segment non-ventilable, and the corresponding valve's pipe segment number is marked as a non-ventilable pipe segment and removed from the natural gas pipeline network topology map. When two or more valves are installed on the same pipe segment, the ventilability of the corresponding pipe segment is determined by jointly considering the valve number affected by construction, the pipe segment number to which the valve belongs, the valve's installation position within the pipe segment, and the planned opening / closing status of the valve affected by construction. If a closed valve affected by construction prevents the formation of a continuous ventilation path from the starting node number to the ending node number within the construction window, the pipe segment is marked as a non-ventilable segment. If a valid continuous ventilation path still exists within the pipe segment via a bypass branch or other open valves, the entire pipe segment is not excluded; instead, only the corresponding ventilation boundary blocked by the closed valve is marked as a non-ventilable boundary. This determination rule avoids mechanically presuming the entire pipe segment as non-ventilable based solely on the closed status of a single valve, ensuring that the ventilable topology within the construction window matches the actual valve arrangement and pipe segment connectivity.If the planned opening / closing state of a valve affected by construction is "open," the corresponding pipe segment number will be retained as a passable pipe segment in the natural gas pipeline network topology diagram. This method ensures the passable topology diagram reflects the temporary gas connection relationships formed within the expected construction window according to the planned opening / closing state of the valves affected by construction. Based on the pipe segment numbers affected by construction, these segments are marked as construction-restricted segments and removed from the passable topology diagram to obtain a construction window reference topology diagram. Construction-restricted segments are those that require connection, cutting, modification, shutdown, pressure reduction, or temporary gas restriction during the corresponding construction phase. Removing these construction-restricted segments from the passable topology diagram simulates the state where construction-affected segments cannot participate in gas supply connectivity as normal gas paths during this construction phase, thus obtaining a construction window reference topology diagram that matches the construction constraints of the expected construction window. In the construction window reference topology diagram, using the pressure regulating station connection node number as the starting node set and the proposed access node number as the target node, a breadth-first search algorithm is used to find the pipeline path connected to the proposed access node number, and the found pipeline path is determined as the backup gas supply channel. The pressure regulating station connection node number is the gas supply-side node capable of providing pressure-regulating gas to the pipeline network, and the starting node set includes one or more pressure regulating station connection node numbers within the same natural gas pipeline network. The breadth-first search algorithm uses the pressure regulating station connection node number in the starting node set as the search starting point and searches layer by layer along the ventilated pipeline segments in the construction window reference topology diagram until the proposed access node number is reached. If one or more pipeline paths from the pressure regulating station connection node number to the proposed access node number are found, the corresponding pipeline path is determined as the backup gas supply channel; if no connected path is found, it is determined that there is no backup gas supply channel for the corresponding construction stage. In the process of using a breadth-first search algorithm to find backup gas supply channels, a path queue is used to store currently expanded candidate paths. Each candidate path consists of the node numbers and pipe segment numbers that are sequentially passed from the starting node to the current node. Each time, a candidate path is taken from the path queue and expanded along the adjacent, unremoved, ventilable pipe segments. If the node number after expansion already exists in the candidate path, the path is not expanded further to avoid creating duplicate loops. If the expanded node number is the number of the node to be connected, the candidate path is recorded as a candidate backup gas supply channel. For each backup gas supply channel, the pipe segment numbers contained within the backup gas supply channel are read, and the difference between the maximum allowable flow rate and the flow rate of each pipe segment is calculated to obtain the remaining flow capacity of each pipe segment. The minimum remaining flow capacity among all pipe segments within the same backup gas supply channel is taken as the remaining capacity of the backup gas supply channel.The maximum allowable flow rate of a pipe segment is obtained by reading the operation constraint parameters of the corresponding pipe segment number in the pipeline network operation configuration parameter table. The pipe segment flow rate comes from the pipe segment flow rate of the corresponding pipe segment number at the corresponding unified sampling time. The remaining flow capacity of a pipe segment is used to characterize the flow margin that the corresponding pipe segment can still bear under the reference operation state. Since the gas supply capacity of the same backup gas supply channel is limited by the bottleneck pipe segment with the smallest remaining capacity in the channel, the minimum value of the remaining flow capacity of each pipe segment in the same backup gas supply channel is taken as the backup channel remaining capacity of the backup gas supply channel, which is used to characterize the minimum available bypass gas supply capacity of the backup gas supply channel under the reference state of the construction window.

[0038] In this implementation plan, the natural gas pipeline network topology is updated by influencing the planned opening and closing status of valves and the affected pipeline sections during construction. This allows for the simulation of temporary topological changes such as valve closure and pipeline section constraints within the construction window. Furthermore, breadth-first search is used to identify backup gas supply channels between the pressure regulating station connection node and the intended access node. Simultaneously, the detour gas supply capacity is assessed by evaluating the remaining flow capacity of pipeline sections and the remaining capacity of backup channels. This enables the early detection of potential gas supply channel interruptions, insufficient backup paths, and bottleneck pipeline capacity limitations during construction. This provides quantifiable topological connectivity and backup channel remaining capacity data for subsequent construction disturbance risk prediction and window optimization.

[0039] Specifically, the process of statistically analyzing node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating stations, gas load level during construction windows, and remaining capacity of backup channels to obtain the reference operating characteristics of construction windows is as follows: Using construction stage numbers as units, read the N most recent construction window reference operating data segments corresponding to the construction stage number; where N is the number of construction window reference operating data segments, a preset positive integer, preferably set based on the expected construction window's time attribute, the number of matching historical data segments in the unified time series data, and the accuracy requirements for predicting gas supply disturbance risks, preferably 3 to 10; the N most recent construction window reference operating data segments are historical reference segments matched according to the expected construction window's time attribute, used to characterize the historical pipeline network operating status under similar seasonal attributes, date types, and intraday time periods during this construction stage; reading by construction stage number ensures that subsequent statistical objects are consistent with the proposed access node number, construction-affected pipeline segment number, construction-affected pressure regulating station number, and backup gas supply channel for the corresponding construction stage. The starting and ending node numbers of the pipe segments corresponding to the proposed connection node number and the construction-affected pipe segment number are used as the node statistical objects. The proposed connection node number represents the location where the proposed expansion pipe segment connects to the existing pipeline network, while the starting and ending node numbers of the pipe segments corresponding to the construction-affected pipe segment number represent the pressure-sensitive nodes at both ends of the construction-affected pipe segment. Using these nodes as the node statistical objects covers the expansion connection location and the key pressure monitoring locations at both ends of the construction-affected pipe segment. For each node statistical object, the node pressure margin is calculated at each unified sampling time, and the minimum value is taken as the minimum node pressure margin. The minimum node pressure margin represents the pressure safety margin closest to the minimum gas supply pressure requirement among the relevant nodes in the reference operation data segments of the most recent N construction windows. The smaller the minimum node pressure margin, the higher the risk of insufficient node pressure under similar operating conditions during the corresponding construction stage. The pipeline segment numbers affected by construction and the pipeline segment numbers within the backup gas supply channel are used as the statistical objects for pipeline segments. The construction-affected pipeline segment numbers represent existing pipeline segments directly restricted during the construction phase, while the pipeline segment numbers within the backup gas supply channel represent pipeline segments that may be relocated to other gas supply routes after the construction-affected pipeline segments become restricted. Using both as the statistical objects for pipeline segments simultaneously covers both directly restricted pipeline segments and bottleneck pipeline segments in the relocation gas supply paths. For each pipeline segment statistical object, the pipeline segment load rate is calculated at each unified sampling time, and the maximum value is taken as the maximum pipeline segment load rate. The maximum pipeline segment load rate represents the state with the highest transport capacity occupancy among the relevant pipeline segments within the reference operation data segments of the most recent N construction windows. The higher the maximum pipeline segment load rate, the higher the probability of pipeline overload or relocation bottlenecks during construction.The pressure regulating station numbers affected by construction are used as statistical objects for pressure regulating stations. These numbers characterize the pressure regulating stations whose outlet pressure, pressure regulating flow, or gas supply range may be affected during the corresponding construction phase. Using these numbers as statistical objects allows for the extraction of the pressure regulating capacity margin of relevant pressure regulating stations during the construction phase. For each statistical object, the remaining pressure regulating capacity of the pressure regulating stations at each unified sampling time is calculated, and the minimum value is taken as the minimum remaining pressure regulating capacity of the pressure regulating station. This minimum remaining pressure regulating capacity characterizes the minimum capacity margin that relevant pressure regulating stations can continue to undertake pressure regulation and gas supply adjustment within the reference operating data segments of the most recent N construction windows. The smaller the minimum remaining pressure regulating capacity, the more limited the adjustment space of the pressure regulating station within the construction window. The set of user nodes affected by construction is determined based on the proposed access node number, the affected pipeline segment number, the backup gas supply channel, and the natural gas pipeline network topology. This set includes user node numbers in the natural gas pipeline network topology that are connected to the proposed access node number, the affected pipeline segment number, and the backup gas supply channel. It is used to limit the range of user-side nodes that may be affected by changes in gas supply pressure, pipeline detours, and construction access during the current construction phase. The set of user nodes affected by construction is determined through the correspondence between user node numbers and node numbers in the basic topology data of the natural gas pipeline network. For each user node affected by construction, the gas load level within the construction window is calculated, and the average value is obtained. This average value characterizes the degree to which the gas load of the set of user nodes affected by construction approaches historical high load levels within the overall dimension of the reference operating data segments of the most recent N construction windows, avoiding the use of gas load fluctuations of a single user node to represent the gas load status of the entire construction impact area. The backup gas supply channel is used as the statistical object for backup channels. The backup gas supply channel is the passable pipeline path found in the construction window reference topology map, searched from the pressure regulating station connection node number to the proposed access node number. The backup channel statistical object is used to characterize the channel capacity that can be used for bypass gas supply after the pipeline segment is restricted due to construction impact. For the backup channel statistical object, the remaining capacity of the backup channel at each unified sampling time is calculated, and the minimum value is taken as the minimum remaining capacity of the backup channel. Specifically, the remaining capacity of the backup channel at each unified sampling time is the minimum value of the remaining flow capacity of each pipeline segment within the corresponding unified sampling time, and the minimum remaining capacity of the backup channel is the minimum value of the remaining capacity of the backup channel at each unified sampling time within the most recent N construction window reference operation data segments. Through this statistical method, the weakest gas supply capacity of the backup gas supply channel at different sampling times can be used as a feature for subsequent risk prediction, avoiding the omission of short-term bottleneck states within the construction window.

[0040] In this implementation plan, by reading the reference operation data segments of the most recent N construction windows in units of construction stage number, and performing targeted statistics on nodes, pipe sections, pressure regulating stations, user nodes, and backup gas supply channels, the pressure safety margin, peak load of pipe sections, pressure regulating capacity margin, user load level, and weakest capacity of backup channels under the construction window of complex pipeline networks can be uniformly transformed into reference operation characteristics of the construction window. This method not only covers the nodes to be connected, pipe sections affected by construction, detour gas supply paths, and user scope affected by construction, but also avoids the one-sided characteristics caused by a single sampling time or a single object, thereby improving the accuracy of subsequent gas supply disturbance risk prediction in identifying the operating status of complex pipeline networks and short-term bottleneck risks.

[0041] Specifically, the process of training the gas supply disturbance risk prediction model for construction windows by combining the pipeline expansion collaborative planning data corresponding to historical construction stages and the reference operation characteristics of construction windows is as follows: Using the construction stage number as a unit, a construction stage disturbance prediction feature vector is generated based on the reference operation data fragments of the most recent N construction windows. This feature vector is used to uniformly convert the construction object, construction plan status, reference operation status, and standby gas supply capacity of the corresponding construction stage into structured input data that the gas supply disturbance risk prediction model can recognize. The construction stage disturbance prediction feature vector includes: construction stage number, proposed expansion pipeline segment number, proposed connection node number, construction-affected pipeline segment number, construction-affected valve number, planned opening / closing status of construction-affected valves, construction-affected pressure regulating station number, and minimum node pressure. The parameters include: gas supply margin, maximum pipeline load rate, minimum remaining pressure regulating capacity of the pressure regulating station, average gas load level during the construction window, number of backup gas supply channels, and minimum remaining capacity of the backup channel. Among these, the construction stage number distinguishes different construction stages; the proposed expansion pipeline section number and the proposed connection node number characterize the expansion connection object and connection location; the construction-affected pipeline section number, construction-affected valve number, planned opening / closing status of construction-affected valves, and construction-affected pressure regulating station number characterize the temporary topology changes and operational capacity changes caused by the corresponding construction stage; and the minimum node pressure margin, maximum pipeline load rate, minimum remaining pressure regulating capacity of the pressure regulating station, average gas load level during the construction window, number of backup gas supply channels, and minimum remaining capacity of the backup channel characterize the gas supply disturbance sensitivity state of the corresponding construction stage under similar historical operating conditions. When there is no backup gas supply channel for the corresponding construction stage, the number of backup gas supply channels is set to 0, and the minimum remaining capacity of the backup channel is set to 0. Based on the construction stage disturbance prediction feature vectors corresponding to historical construction stages, a construction disturbance training sample set is constructed. Historical construction stages are those that have been completed and whose operational data during execution has been recorded. Each training sample in the training sample set corresponds to a historical construction stage number. Each training sample uses the construction stage disturbance prediction feature vector generated before construction in that historical construction stage as its sample feature, and the gas supply disturbance results that occurred during the actual execution of that historical construction stage as its sample label source. For each historical construction stage number in the training sample set, training sample labels are generated, including node pressure exceeding limits, pipeline overload, and insufficient backup channels. The node pressure exceeding limits label indicates whether node pressures fell below safety constraints during the actual execution of the historical construction stage; the pipeline overload label indicates whether pipeline capacity was excessively occupied during the actual execution of the historical construction stage; and the insufficient backup channels label indicates whether there were missing backup gas supply paths or insufficient backup gas supply capacity during the actual execution of the historical construction stage.During the subsequent actual execution period of a historical construction phase, if the pressure margin of any node in the corresponding node statistics is less than the pressure margin threshold, the node pressure exceeding the limit label of the corresponding historical construction phase number will be marked as 1; otherwise, it will be marked as 0. The subsequent actual execution period of a historical construction phase refers to the continuous construction execution time interval between the actual start time and the actual end time of the corresponding historical construction phase. The pressure margin threshold is a preset pressure safety judgment threshold, set according to the minimum allowable pressure of the corresponding node in the pipeline operation configuration parameter table, the pressure control requirements during construction as specified in the gas company's dispatching procedures, and the normal fluctuation range of node pressure in historical operation data. It is used to determine whether the node pressure margin has fallen below the allowable safety margin requirement during construction. When the node pressure margin is less than the pressure margin threshold, it indicates that the node has a risk of insufficient pressure during construction execution. During the subsequent actual execution of the historical construction phase, when the load rate of any pipe segment in the corresponding statistical object exceeds the load threshold, the overload label of the pipe segment corresponding to the historical construction phase number will be marked as 1; otherwise, it will be marked as 0. The load threshold is a preset pipe segment load judgment threshold, which is set according to the maximum allowable flow rate of the corresponding pipe segment in the pipeline network operation configuration parameter table, the pipe segment load control requirements during construction as specified in the gas company's dispatching procedures, and the fluctuation range of the pipe segment load rate in the historical normal operation data. It is used to determine whether the flow occupancy of the pipe segment exceeds the allowable load level during construction. When the pipe segment load rate is greater than the load threshold, it indicates that the corresponding pipe segment has a tight transmission capacity and an overload risk during the construction execution period. When there is no backup gas supply channel during a historical construction phase, or when the remaining capacity of the minimum backup channel is less than the backup capacity threshold, the backup channel insufficiency label corresponding to the historical construction phase number is marked as 1; otherwise, it is marked as 0. The backup capacity threshold is a preset backup gas supply capacity judgment threshold, set based on the expected detour gas demand that the backup gas supply channel needs to handle after the construction-affected pipeline section is restricted, the backup gas supply safety margin specified in the gas company's dispatching procedures, and the fluctuation range of the remaining flow capacity of the backup gas supply channel in historical normal operation data. It is used to determine whether the backup gas supply channel can provide the remaining capacity to meet the detour gas supply needs when the construction-affected pipeline section is restricted. When there is no backup gas supply channel during a historical construction phase, it indicates that no connecting path for detour gas supply has been formed in the construction window reference topology map. When the remaining capacity of the minimum backup channel is less than the backup capacity threshold, it indicates that although a backup gas supply channel exists, there is insufficient remaining capacity in the bottleneck section of that channel, making it difficult to stably handle the detour gas supply demand. Using the construction disturbance training sample set as the model input features and the training sample labels as the model output labels, a gradient boosting tree classification model is trained to construct a risk prediction model for gas supply disturbance during construction windows.The gradient boosting tree classification model is a gradient boosting decision tree classification model, also known as the GBDT classification model. The GBDT model uses a classification regression tree as the base learner, iteratively constructing multiple decision trees round by round, and then weighting and combining the outputs of these decision trees to obtain the classification probability of the corresponding risk label. Specifically, the gradient boosting tree classification model uses each feature field from the construction phase disturbance prediction feature vector as input, and uses the node pressure exceeding limit label, pipe section overload label, and insufficient backup channel label as supervised training objectives. Since these labels are all binary labels (0 or 1), GBDT binary classification sub-models for node pressure exceeding limit, pipe section overload, and insufficient backup channel are constructed separately. Each GBDT binary classification sub-model uses a logarithmic loss function as the model training loss function to measure the deviation between the predicted probability and the corresponding binary classification label. After training, the construction window gas supply disturbance risk prediction model is used to output the corresponding node pressure exceeding probability, pipeline overload probability, and backup channel insufficient probability based on the construction stage disturbance prediction feature vector of the current construction stage. This provides a model basis for subsequent determination of comprehensive construction disturbance risk value and identification of construction stages to be optimized. Specifically, during training, the feature fields in the construction stage disturbance prediction feature vector are preprocessed for model input. Specifically, the construction stage number, proposed expansion pipeline number, proposed connection node number, construction-affected pipeline number, construction-affected valve number, planned opening / closing status of construction-affected valves, and construction-affected pressure regulating station number are categorized and coded. The minimum node pressure margin, maximum pipeline load rate, minimum remaining pressure regulating station capacity, average gas load level of the construction window, number of backup gas supply channels, and minimum remaining backup channel capacity are numerically normalized to obtain a standardized training feature matrix. The construction disturbance training sample set is divided into a training subset and a validation subset according to a preset ratio, preferably 8:2. The gradient boosting tree classification model is trained using the training subset, and the model output is validated using the validation subset. The number of iterations, learning rate, maximum depth of a single decision tree, and minimum number of leaf node samples in the GBDT classification model are set based on the prediction performance on the validation subset; preferably, the number of iterations is 50 to 300, the learning rate is 0.01 to 0.1, the maximum depth of a single decision tree is 3 to 6, and the minimum number of leaf node samples is 5 to 30. To reduce the risk of overfitting, an L2 regularization term is set during model training, and an early stopping rule is set based on the log loss function value on the validation subset; when the log loss function value on the validation subset no longer decreases within a preset number of consecutive rounds, iterative training is stopped.Since the training sample labels include labels for node pressure exceeding limits, pipeline overload, and insufficient backup channels, these labels can be used as supervised training objectives to train the node pressure exceeding limit prediction sub-model, pipeline overload prediction sub-model, and insufficient backup channel prediction sub-model, respectively. These three sub-models together constitute the gas supply disturbance risk prediction model for the construction window. Specifically, the node pressure exceeding limit prediction sub-model, pipeline overload prediction model, and insufficient backup channel prediction model are all GBDT binary classification sub-models, each outputting a probability value of 1 for its corresponding label. During model training, the gradient boosting tree classification model fits the prediction error of the previous round by constructing decision trees round by round, and learns the impact of different construction plan states, pipeline operation states, and backup gas supply capacity on gas supply disturbance risk based on the mapping relationship between the construction stage disturbance prediction feature vector and the corresponding training sample labels. After training, the feature vector of the current construction phase disturbance prediction is input into the gas supply disturbance risk prediction model for the construction window. This yields the probability of node pressure exceeding the limit (output by the node pressure exceeding the limit prediction sub-model), the probability of pipe section overload (output by the pipe section overload prediction sub-model), and the probability of insufficient backup channels (output by the insufficient backup channels prediction sub-model). The values ​​for these probabilities range from 0 to 1; higher values ​​indicate a higher probability of the corresponding disturbance risk occurring.

[0042] In this implementation plan, by converting the construction objects, valve planned status, reference operating characteristics, and backup gas supply capacity of historical construction phases into construction phase disturbance prediction feature vectors, and generating supervised training labels based on the results of node pressure exceeding limits, pipeline overload, and insufficient backup channels during actual construction execution, the gas supply disturbance risk prediction model can learn the correspondence between construction plan status and pipeline operation disturbances. At the same time, by constructing a GBDT binary classification sub-model to output the probability of node pressure exceeding limits, pipeline overload, and insufficient backup channels, the construction disturbance risk is transformed from empirical judgment into quantifiable probability prediction results, thereby improving the accuracy of identifying the construction phase to be optimized and the reliability of subsequent construction window adjustment and valve switching collaborative optimization.

[0043] Specifically, the process of obtaining the comprehensive construction disturbance risk value based on the construction window gas supply disturbance risk prediction model and identifying the construction stage to be optimized is as follows: The construction stage disturbance prediction feature vector of the current construction stage number is input into the construction window gas supply disturbance risk prediction model. Here, the current construction stage number is the construction stage number to be evaluated for the construction plan, and the construction stage disturbance prediction feature vector is the model input feature generated based on the pipeline expansion collaborative planning data corresponding to the current construction stage, the construction window reference operation characteristics, and the remaining capacity of the backup gas supply backup channel. Before inputting into the construction window gas supply disturbance risk prediction model, the construction stage disturbance prediction feature vector of the current construction stage is preprocessed according to the same category coding rules and numerical feature normalization rules as in the model training stage to ensure that the model input feature format is consistent. The construction window gas supply disturbance risk prediction model outputs the probability of node pressure exceeding limits, pipeline overload, and insufficient backup channels corresponding to the construction stage number. The probability of node pressure exceeding limits represents the likelihood that the node pressure margin will fall below the pressure margin threshold within the expected construction window during the current construction stage. The probability of pipeline overload represents the likelihood that the pipeline load rate will exceed the load threshold within the expected construction window during the current construction stage. The probability of insufficient backup channels represents the likelihood that the remaining capacity of the backup gas supply channels will be insufficient when the pipeline is restricted due to construction impact. All three probabilities range from 0 to 1. The maximum value among these three probabilities is taken as the comprehensive construction disturbance risk value. This comprehensive construction disturbance risk value represents the highest disturbance risk level among node pressure, pipeline load, and backup channels during the current construction stage. Using the maximum value as the comprehensive construction disturbance risk value ensures that any high gas supply disturbance risk will trigger subsequent optimization processing, avoiding the masking of a single high-risk issue by averaging multiple risk probabilities. The probabilities of node pressure exceeding limits, pipeline overload, and insufficient backup channels are all output by the GBDT binary classification sub-model. After model training, these three probabilities undergo probability calibration using the same validation subset, mapping each probability to a risk occurrence probability range of 0 to 1. The probability calibration process includes correcting the output probabilities of each sub-model based on the deviation between the statistically predicted probabilities of the validation subset and the actual frequency of occurrence, ensuring a consistent comparison scale for the probabilities of node pressure exceeding limits, pipeline overload, and insufficient backup channels. This probability calibration process avoids distortion of the overall construction disturbance risk value due to differences in the output probability distributions of different sub-models. Furthermore, using the maximum value as the basis for the overall construction disturbance risk value is based on the fact that node pressure exceeding limits, pipeline overload, and insufficient backup channels are all risk events within the construction window that may directly affect gas supply safety. When any of these risk events reaches a high probability, it may lead to adjustments in the construction window or coordinated valve switching at the corresponding construction stage.Therefore, using the maximum value among the three calibrated probabilities as the comprehensive construction disturbance risk value can retain the most unfavorable risk item according to the principle of safety conservatism, which is suitable for pipeline expansion collaborative planning scenarios with high gas supply guarantee requirements during construction. When the comprehensive construction disturbance risk value is less than or equal to the optimization trigger threshold, the expected construction start time and expected construction end time corresponding to the construction stage number remain unchanged to obtain the recommended construction window, and the planned opening and closing status of the construction-affected valves remains unchanged. Among them, the optimization trigger threshold is a preset construction disturbance risk judgment threshold, which is set according to the gas company's construction scheduling procedures, the correspondence between historical construction stage disturbance risk prediction results and actual disturbance occurrence results, and the gas supply guarantee requirements during construction. When the comprehensive construction disturbance risk value is less than or equal to the optimization trigger threshold, it means that the node pressure, pipeline load, and backup channel capacity under the current expected construction window have not reached the risk level that requires adjustment of the construction plan, so there is no need to change the original construction window and the original planned opening and closing status of the construction-affected valves. When the comprehensive construction disturbance risk value is greater than the optimization trigger threshold, the construction stage number is marked as the construction stage to be optimized. The "Construction Stage to be Optimized" indicates that the current construction stage has a high risk of gas supply disturbance under the expected construction window, and needs to enter the process of generating subsequent window adjustment schemes and valve switching collaborative schemes. This marker can establish a clear correspondence between the risk prediction results and the subsequent construction collaborative optimization steps.

[0044] In this implementation plan, the maximum value among the probability of node pressure exceeding limits, pipeline overload, and insufficient backup channels is used as the comprehensive construction disturbance risk value. This highlights the weakest gas supply risk link in the current construction phase and avoids the average masking of a single high risk. At the same time, the optimization trigger threshold is used to determine whether to maintain the original construction window and valve plan status, or to mark the corresponding construction phase as a construction phase to be optimized. This allows the gas supply disturbance risk prediction results to be directly converted into the basis for determining whether the construction plan needs to be adjusted, improving the pertinence and reliability of construction window recommendations and subsequent valve switching collaborative optimization.

[0045] Specifically, the process of generating candidate construction windows and determining recommended construction windows for the construction phase to be optimized is as follows: For the construction phase to be optimized, a window adjustment scheme is generated. The construction phase to be optimized is one where the comprehensive construction disturbance risk value exceeds the optimization trigger threshold. The window adjustment scheme is used to adjust the start and end times of construction without changing the construction content or the original expected duration of the construction window, thereby reducing the gas supply disturbance risk of the corresponding construction phase within the expected construction window. A set of candidate construction windows is generated based on the expected start time, expected end time, and preset window adjustment step size. The preset window adjustment step size limits the movement interval of the candidate construction windows relative to the original expected construction window. This preset step size is set based on the gas company's construction scheduling cycle, the adjustable granularity of the construction organization plan, and the pipeline operation data sampling cycle, preferably between 1 hour and 24 hours. The start time of the candidate construction window is obtained by moving it bidirectionally before and after the original expected start time according to the preset window adjustment step size. The end time of the candidate construction window is determined based on its start time and the original expected duration of the construction window. The candidate construction windows are generated using a bidirectional movement method. Based on the original projected construction window, they are gradually shifted forward and backward according to preset window adjustment steps, one before the projected start time and the other after. The set of candidate construction windows includes the original projected construction window, candidate construction windows obtained by forward shifting, and candidate construction windows obtained by backward shifting. The maximum forward shift is the integer ratio between the preset maximum allowable advance adjustment time and the preset window adjustment step size; the maximum backward shift is the integer ratio between the preset maximum allowable delay adjustment time and the preset window adjustment step size. The preset maximum allowable advance adjustment time and preset maximum allowable delay adjustment time are set based on the construction permit time, the permitted period for pipeline shutdown or pressure reduction operations, the construction team's shift schedule, user gas supply guarantee requirements, and the earliest and latest allowable construction times in the project planning management system. Preferably, these are 1 to 7 days, and the preset maximum allowable advance adjustment time must not exceed the current time. The preset maximum allowable advance adjustment time and preset maximum allowable delay adjustment time can be the same or set separately according to construction organization constraints. For candidate construction windows with the same start and end times, only one is retained. Furthermore, each candidate construction window maintains the same duration as the original planned construction window; by maintaining the same duration, it is possible to ensure that the construction workload corresponding to different candidate construction windows is consistent, and to avoid the incomparability of disturbance risk prediction results due to changes in the duration of candidate construction windows.Optionally, candidate construction windows whose start or end time falls within a preset nighttime restricted construction period are eliminated. The preset nighttime restricted construction period is set according to the gas company's construction management regulations, urban road construction permit requirements, user gas supply guarantee requirements, and on-site safety operation requirements, preferably from 22:00 to 6:00 the next day. If a candidate construction window spans the preset nighttime restricted construction period, it is marked as a nighttime restricted candidate construction window. This rule avoids adjusting regular expansion construction windows to the nighttime restricted period, ensuring that the set of candidate construction windows conforms to actual construction permits and on-site operation requirements. For each candidate construction window in the candidate construction window set, the N most recent construction window reference operation data segments with the same time attributes as the candidate construction window are selected from the unified time series data, and the corresponding construction stage disturbance prediction feature vector is generated. This feature vector is then input into the gas supply disturbance risk prediction model for the construction window to obtain the corresponding comprehensive construction disturbance risk value. The time attributes of the candidate construction window include seasonal attributes, date type, and intraday time period. The selection method of the N most recent construction window reference operation data segments is consistent with the selection method of the reference operation data segments corresponding to the originally expected construction window. The construction stage disturbance prediction feature vector is regenerated for each candidate construction window to ensure that the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, gas load level of the construction window, and remaining capacity of the standby channel are updated synchronously with the changes in the time attributes of the candidate construction window, thereby ensuring that the risk prediction results of different candidate construction windows are supported by corresponding historical operation data. The candidate construction window with the lowest comprehensive construction disturbance risk value in the candidate construction window set is determined as the optimal candidate construction window. When two or more candidate construction windows have the same comprehensive construction disturbance risk value, priority is determined according to the adjustment range of the candidate construction windows relative to the original planned construction window, from smallest to largest, with the candidate construction window with the smallest adjustment range being selected as the optimal candidate construction window. When the adjustment ranges are still the same, the candidate construction window with the smaller average gas load level is selected first. When the average gas load level of the construction windows is still the same, the candidate construction window with the earlier start time is selected first. Through these rules, the optimal candidate construction window can be uniquely determined when the comprehensive construction disturbance risk value is the same, avoiding inconsistent candidate window selection results. It is then determined whether the comprehensive construction disturbance risk value of the optimal candidate construction window is less than or equal to the optimization trigger threshold. If so, the optimal candidate construction window is determined as the recommended construction window. Otherwise, the optimal candidate construction window is marked as a remaining disturbance risk construction stage and a prompt is output for manual scheduling review, and can serve as an alternative output result for subsequent construction plan adjustments.Once recommended construction windows have been determined for multiple construction stages, time conflict checks are performed on each recommended construction window. A time conflict refers to an overlap between the start and end times of two or more recommended construction windows. If a time conflict exists, the multiple construction stages with time conflicts are grouped into the same conflicting construction stage group. Within each conflicting construction stage group, the comprehensive construction disturbance risk value, the number of user nodes affected, the number of pipe segments affected, and the adjustable range of candidate construction windows are retrieved. For the same conflicting construction stage group, window retention priority is determined according to the comprehensive construction disturbance risk value from highest to lowest. When the comprehensive construction disturbance risk values ​​are the same, the recommended construction window corresponding to the construction stage with a larger number of user nodes affected is retained first. When the number of user nodes affected is still the same, the recommended construction window corresponding to the construction stage with a larger number of pipe segments affected is retained first. When the number of pipe segments affected is still the same, the recommended construction window corresponding to the construction stage with a smaller adjustable range of candidate construction windows is retained first. For each conflicting construction stage group, the optimal candidate construction window corresponding to the highest priority construction stage remains unchanged. For the remaining construction stages in the same conflicting construction stage group, the retained recommended construction windows are used as the avoidance targets. The recommended construction windows for the corresponding construction stages are shifted forward and backward according to a preset window adjustment step size, and new avoidance candidate construction windows are generated. For the newly generated avoidance candidate construction windows, the corresponding construction window reference running data segments are rematched, the construction stage disturbance prediction feature vector is regenerated, and the comprehensive construction disturbance risk value is recalculated. The avoidance candidate construction window with no time conflict and the lowest comprehensive construction disturbance risk value is determined as the recommended construction window for the corresponding construction stage. When a construction stage cannot obtain an avoidance candidate construction window without time conflict within the preset maximum allowable advance adjustment time and preset maximum allowable delay adjustment time, the construction stage is marked as a window conflict construction stage awaiting review, and a prompt is output. Optionally, throughout the entire process of generating candidate construction windows, determining the optimal candidate construction window, and adjusting conflicts among the optimal candidate construction windows for multiple construction stages, the construction sequence between each construction stage remains unchanged. The construction sequence is determined based on the construction stage number, construction procedure dependencies, and the original expected sequence of construction windows recorded in the project planning management system. For any construction stage, its candidate construction window, optimal candidate construction window, avoidance candidate construction window, and recommended construction window must not be earlier than the end time of the recommended construction window of its preceding construction stage, nor later than the start time of the recommended construction window of its subsequent construction stage, thus disrupting the original construction procedure dependencies. If a candidate construction window causes the above-mentioned construction sequence to be disrupted, then the candidate construction window is marked as a sequence conflict window and is not used as a recommended construction window, and a corresponding prompt is output.By using the above methods, it can be ensured that the window adjustment only changes the specific execution time of the construction phase, without changing the sequential relationship between the various construction phases in the original construction organization plan.

[0046] In this embodiment, Table 1 is a risk assessment data table for candidate construction windows. The table details the candidate construction window time period, the probability of pressure exceeding the limit at key points, the probability of pipe section overload, the probability of insufficient backup access, and the overall construction disturbance risk value corresponding to the five candidate window numbers. Specifically, the candidate construction window time period corresponding to candidate window number W1 is July 17th, 06:00-16:00, with a key pressure exceeding the limit probability of 0.4, a pipe section overload probability of 0.50, a insufficient backup access probability of 0.38, and an overall construction disturbance risk value of 0.50; the candidate construction window time period corresponding to candidate window number W2 is July 18th, 08:00-18:00, with a key pressure exceeding the limit probability of 0.54, a pipe section overload probability of 0.59, a insufficient backup access probability of 0.46, and an overall construction disturbance risk value of 0.59. Candidate window number W3 corresponds to the candidate construction window period of July 19th, 07:00-17:00, with a node pressure exceeding limit probability of 0.34, a pipe section overload probability of 0.39, a backup channel insufficiency probability of 0.31, and a comprehensive construction disturbance risk value of 0.39; Candidate window number W4 corresponds to the candidate construction window period of July 20th, 09:00-19:00, with a node pressure exceeding limit probability of 0.58, a pipe section overload probability of 0.68, a backup channel insufficiency probability of 0.49, and a comprehensive construction disturbance risk value of 0.68.

[0047] surface Risk assessment data table for candidate construction windows

[0048] Candidate window number Candidate construction window Node pressure exceeding limit probability Pipeline overload probability Insufficient backup channels Comprehensive construction disturbance risk value W1 July 17, 06:00-16:00 0.43 0.50 0.38 0.50 W2 July 18, 08:00-18:00 0.54 0.59 0.46 0.59 W3 July 19, 07:00-17:00 0.34 0.39 0.31 0.39 W4 July 20, 09:00-19:00 0.58 0.68 0.49 0.68

[0049] like Figure 3 The diagram shows the ranking of candidate construction disturbance risk values ​​for each window. The diagram displays the risk ranking of the four candidate construction windows from lowest to highest comprehensive construction disturbance risk value. W3, corresponding to July 19th, 07:00-17:00, has the lowest comprehensive construction disturbance risk value at 0.39, indicating that the comprehensive risk of node pressure exceeding limits, pipeline overload, and insufficient backup channels is minimal under this window, making it a recommended construction window. W1, corresponding to July 17th, 06:00-16:00, has a risk value of 0.50, placing it at a selectable level. W2, corresponding to July 18th, 08:00-18:00, has a risk value of 0.59, indicating a further increase in risk. W4, corresponding to July 20th, 09:00-19:00, has the highest risk value at 0.68, indicating that construction within this window will have the greatest disturbance impact on the gas supply safety of the pipeline network. Therefore, this diagram can intuitively reflect the risk differences among different candidate construction windows, providing a visual basis for selecting low-risk recommended construction windows.

[0050] In this implementation plan, candidate construction windows are generated by adjusting the step size according to a preset window without changing the construction content and duration. The comprehensive construction disturbance risk value corresponding to each candidate construction window is predicted one by one. This allows the selection of the recommended construction window with the lowest gas supply disturbance risk from multiple feasible construction time intervals. At the same time, by setting the maximum allowable adjustment range and the remaining disturbance risk marker, the infinite expansion of the window is avoided, which would make the construction plan unexecutable. Furthermore, when the recommended construction windows of multiple construction stages overlap, the retention priority is determined by the comprehensive construction disturbance risk value, the number of user nodes affected by the construction, and the adjustable range of the window. The stages that are not retained are re-searched for avoidance, thereby reducing the risk of gas supply disturbance superposition caused by multiple construction stages running in parallel and improving the feasibility, coordination, and safety of the construction window adjustment scheme.

[0051] Specifically, the process of synchronously screening candidate connecting valves and generating valve switching coordination schemes is as follows: For the construction stage to be optimized, a valve switching coordination scheme is generated. This scheme coordinates the planned opening and closing status of valves affected by construction within the recommended construction window, and collaboratively marks connectable valves in the backup gas supply channel. This ensures that even after the affected pipeline section is restricted, the gas supply connection between the intended access node and the affected user node can still be maintained through the backup gas supply channel. The valve switching coordination scheme is generated synchronously with the window adjustment scheme to ensure consistency between construction time adjustments and pipeline gas path adjustments under the same construction stage number. Valves in the backup gas supply channel that do not belong to the construction-affected valve number are designated as candidate connecting valves and marked as openable. The valves in the backup gas supply channel are determined based on the correspondence between the pipeline segment number within the backup gas supply channel and the valve's corresponding pipeline segment number. Valves not belonging to the construction-affected valve number refer to valves not included in the construction isolation, shutdown, pressure reduction, or safety closure control scope by the engineering planning management system. Using these valves as candidate connection valves avoids altering the construction safety isolation boundary formed by the planned opening and closing status of valves affected by construction, while providing dispatchable objects for connecting backup gas supply channels. Marking a candidate connection valve as "openable" means that after meeting dispatch safety verification and on-site operational confirmation, it can participate in valve switching operations as a backup gas supply connection valve within the recommended construction window; it does not imply direct opening without verification and confirmation. Before actually executing valve switching, confirmation must be made based on the current valve status of the candidate connection valve, the planned opening and closing status of valves affected by construction, the connection relationship of the backup gas supply channel, and the gas company's valve operating procedures. Valves already in the open state and meeting the backup gas supply channel connection requirements should remain open; candidate connection valves in the closed state that, when opened, can form a continuous gas-flowing path in the backup gas supply channel should be considered openable valves; valves within the construction isolation range or whose opening may disrupt the construction safety isolation boundary should not be considered candidate connection valves. This approach avoids simply opening all non-construction-affected valves uniformly, ensuring that the valve switching coordination scheme meets actual dispatch operation requirements. When generating a valve switching coordination scheme, a connectivity check can be optionally performed on candidate connecting valves. This connectivity check includes: simulating the candidate connecting valves in an open state in the construction window reference topology diagram, and determining whether a continuous, permeable path is formed between the pressure regulating station connection node number and the proposed access node number. If a continuous, permeable path is formed, the corresponding candidate connecting valve is retained; if no continuous, permeable path is formed, the corresponding candidate connecting valve is not considered a valid candidate connecting valve. This check ensures that the candidate connecting valves can indeed function to connect the backup gas supply channel.Optionally, channel capacity verification is performed on the retained candidate connecting valves. This verification includes: reading the remaining capacity of the backup gas supply channel where the candidate connecting valve is located under the reference operating data segments of the most recent N construction windows corresponding to the recommended construction window, and determining whether the remaining capacity of the backup channel is greater than or equal to the backup capacity threshold. Preferably, the remaining capacity of the backup channel is the minimum value of the remaining capacity of the backup channel at each uniform sampling time within the reference operating data segments of the most recent N construction windows corresponding to the recommended construction window. If the remaining capacity of the backup channel is greater than or equal to the backup capacity threshold, the corresponding candidate connecting valve is retained in the valve switching coordination scheme. If the remaining capacity of the backup channel is less than the backup capacity threshold, the corresponding candidate connecting valve is marked as a candidate connecting valve with insufficient capacity, indicating that although the channel can be connected, the gas supply margin is insufficient under similar historical operating conditions, requiring manual scheduling verification or processing with other backup channels. A construction coordination optimization scheme is generated and output based on the window adjustment scheme and the valve switching coordination scheme. The construction collaboration optimization scheme includes the construction stage number, the proposed expansion pipeline segment number, the proposed connection node number, the recommended construction window, the construction-affected pipeline segment number, the construction-affected valve number, the planned opening and closing status of the construction-affected valve, the backup gas supply channel, the candidate connecting valve number, the candidate connecting valve's openable status, the remaining capacity of the backup channel, the optimized comprehensive construction disturbance risk value, and the scheme generation time. The construction collaboration optimization scheme is output to the engineering planning management system and the gas dispatch management system to enable the collaborative execution of construction plan adjustments, valve switching preparations, and gas supply dispatch verification based on the same construction stage number.

[0052] In this implementation plan, by simultaneously screening candidate connecting valves within the recommended construction window to generate a valve switching coordination scheme, it is possible to identify candidate connecting valves that can be used for connection from the backup gas supply channels without changing the planned opening and closing status of valves affected by construction and the construction safety isolation boundary. The plan also determines whether they can form an effective bypass gas supply path through connectivity verification and channel capacity verification. At the same time, by evaluating the remaining capacity of the backup channels based on historical similar operation data corresponding to the recommended construction window, it is possible to avoid misjudging the feasibility of gas supply based solely on topological connectivity. This allows construction time adjustment, valve switching preparation, and gas supply scheduling verification to be output collaboratively under the same construction stage number, thereby improving the executability, safety, and reliability of the construction collaborative optimization scheme.

[0053] Reference Figure 2As shown, the second aspect of the present invention provides a machine learning-based natural gas pipeline network collaborative planning system, applied to the aforementioned machine learning-based natural gas pipeline network collaborative planning method, comprising: a pipeline network data preprocessing module, used to collect pipeline network expansion collaborative planning data within the same natural gas pipeline network and perform data preprocessing on the pipeline network expansion collaborative planning data; a window feature recognition module, used to extract the time attributes of the construction phase based on the pipeline network expansion collaborative planning data, match the construction window reference operation data segment according to the time attributes, and evaluate the node pressure margin, pipeline section load rate, remaining pressure regulating capacity of the pressure regulating station, and gas load level of the construction window based on the construction window reference operation data segment; and update the pipeline network topology based on the pipeline network expansion collaborative planning data, identifying... The system includes a backup gas supply channel and an assessment of its remaining capacity. A disturbance risk prediction module is used to statistically analyze node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, gas load level during construction windows, and remaining capacity of backup channels to obtain reference operating characteristics for construction windows. This is combined with historical pipeline expansion collaborative planning data for different construction phases and the reference operating characteristics of construction windows to train a gas supply disturbance risk prediction model for construction windows. Based on this model, a comprehensive construction disturbance risk value is obtained, and construction phases requiring optimization are identified. A window valve optimization module is used to generate candidate construction windows and determine recommended construction windows for construction phases requiring optimization, while simultaneously screening candidate connecting valves to generate valve switching collaborative schemes.

[0054] like Figure 4The diagram shows the workflow of a machine learning-based collaborative planning method for natural gas pipeline networks. It illustrates the overall execution process of this method. First, it collects pipeline GIS topology, SCADA operation data, user gas consumption data, and expansion construction plan data, performing anomaly removal, missing data completion, and time alignment to form a unified data foundation for collaborative planning of pipeline expansion. Next, it extracts time attributes such as seasonality, date type, and intraday time period from the construction window, matches similar historical operation data segments, and calculates node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating stations, and user load levels. Simultaneously, it updates the pipeline topology based on the valves and pipeline segments affected by construction, searches for alternative gas supply channels, and assesses the remaining capacity of these channels. Finally, it statistically analyzes the above operational characteristics to construct a feature vector for predicting disturbances during the construction phase, and calculates the comprehensive construction disturbance risk value using the GBDT disturbance risk prediction model. When the risk value is less than or equal to the optimization trigger threshold, the original construction window and valve scheme are used, and the final scheme is output. When the risk value is greater than the optimization trigger threshold, the corresponding construction stage is marked as a construction stage to be optimized. Multiple candidate construction windows are generated, and the risk value is calculated for each. The recommended construction window with the lowest risk and that meets the constraints is selected, and candidate connected valves are matched simultaneously to generate a valve switching scheme. Finally, the construction collaborative optimization results are output by combining the construction window and valve scheme, realizing a closed-loop planning from multi-source data perception and risk prediction to window and valve collaborative adjustment.

[0055] In this implementation plan, a unified, continuous, and computable data foundation for collaborative planning of pipeline expansion is formed through a pipeline data preprocessing module. A window feature recognition module matches the operating status of similar construction windows and identifies backup gas supply channels. A disturbance risk prediction module quantifies the comprehensive construction disturbance risk corresponding to node pressure, pipeline load, pressure regulation capacity, user load, and remaining capacity of backup channels. Finally, a window valve optimization module generates construction window adjustment schemes and valve switching coordination schemes. This enables closed-loop processing from multi-source data acquisition and construction window risk identification to collaborative optimization of construction plans, improving the feasibility, gas supply security, and scheduling coordination of natural gas pipeline expansion construction planning.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0057] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. As those skilled in the art will understand, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A machine learning-based collaborative planning method for natural gas pipeline networks, characterized in that, Includes the following steps: S1, collect pipeline expansion collaborative planning data within the same natural gas pipeline network, and perform data preprocessing on the pipeline expansion collaborative planning data; S2 extracts the time attributes of the construction phase based on the pipeline expansion collaborative planning data, matches the construction window reference operation data segments according to the time attributes, and evaluates the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station and the gas load level of the construction window based on the construction window reference operation data segments; and updates the pipeline topology based on the pipeline expansion collaborative planning data, identifies the backup gas supply channel and evaluates the remaining capacity of the backup channel. S3. Statistical analysis is performed on node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating stations, gas load level of construction windows, and remaining capacity of standby channels to obtain reference operating characteristics of construction windows. Combined with pipeline expansion collaborative planning data corresponding to historical construction stages and reference operating characteristics of construction windows, a risk prediction model for gas supply disturbance in construction windows is trained. Based on the risk prediction model for gas supply disturbance in construction windows, a comprehensive construction disturbance risk value is obtained, and construction stages to be optimized are identified. S4: For the construction phase to be optimized, generate candidate construction windows and determine recommended construction windows, and simultaneously filter candidate connected valves to generate valve switching coordination schemes.

2. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 1, characterized in that, The specific process of collecting pipeline expansion collaborative planning data within the same natural gas pipeline network and performing data preprocessing on the pipeline expansion collaborative planning data is as follows: Data collection for collaborative planning of pipeline expansion within the same natural gas pipeline network includes: collecting basic topology data of the natural gas pipeline network based on a natural gas pipeline network GIS, including: pipeline segment number, starting node number, ending node number, node number, valve number, pipeline segment number to which the valve belongs, pressure regulating station number, and pressure regulating station connection node number; collecting operational data of the natural gas pipeline network based on a SCADA system, pressure sensors, flow meters, and pressure regulating station monitoring system, including: node operating pressure, pipeline flow rate, pressure regulating station outlet pressure, pressure regulating station flow rate, and collection timestamp; collecting user gas consumption data from the gas company's user management system, including: user node number, user gas consumption, user historical peak gas consumption, and collection timestamp; and collecting expansion construction plan data from the project planning management system, including: proposed expansion pipeline segment number, proposed connection node number, construction stage number, estimated construction start time, estimated construction end time, affected pipeline segment number, affected valve number, planned opening / closing status of affected valves, and affected pressure regulating station number. Based on the pipe section number, node number, and pressure regulating station number, the data is associated with the corresponding natural gas pipeline network operation data. The natural gas pipeline network operation data and user gas consumption data are processed to unify time according to the collection timestamp, and a unified sampling time is generated according to the preset sampling period, forming a unified time series data. When there are more than two records corresponding to the same object number within the same sampling period, the record closest to the sampling time is selected as the data for the corresponding object number at that sampling time. If the pressure and flow data exceed the range of the corresponding sensor, they are marked as outliers and removed. Missing data is filled in using linear interpolation.

3. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 2, characterized in that, The specific process of extracting the time attributes of the construction phase based on pipeline expansion collaborative planning data, matching the construction window reference operation data segments based on the time attributes, and evaluating the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station, and gas consumption load level of the construction window based on the construction window reference operation data segments is as follows: Based on the construction stage number, the expected start time and the expected end time, determine the expected construction window corresponding to each construction stage number, and extract the time attributes of the expected construction window, including seasonal attributes, date type and intraday time period; Based on the time attributes of the expected construction window, historical data segments with the same time attributes as the expected construction window are selected from the unified time series data to construct the construction window reference operation data segment corresponding to the construction stage. The construction window reference operation data segment includes: node number, node operating pressure, pipe segment number, pipe segment flow, pressure regulating station number, pressure regulating station outlet pressure, pressure regulating station pressure regulating flow, user node number, and user gas consumption. Based on the reference operation data segment of the construction window, the node pressure margin, pipeline load rate, and remaining pressure regulating capacity of the pressure regulating station are calculated. Among them, the node pressure margin is the difference between the node operating pressure and the node minimum allowable pressure, the pipeline load rate is the ratio between the pipeline flow rate and the pipeline maximum allowable flow rate, and the remaining pressure regulating capacity of the pressure regulating station is the difference between the maximum pressure regulating capacity of the pressure regulating station and the pressure regulating flow rate of the pressure regulating station. At the same time, for each user node number in the reference operation data segment of the construction window, the ratio between the average gas consumption of the user and the average historical peak gas consumption of the user is calculated to obtain the gas load level of the construction window.

4. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 1, characterized in that, The specific process of updating the pipeline topology based on pipeline expansion collaborative planning data, identifying backup gas supply channels, and assessing the remaining capacity of backup channels is as follows: A natural gas pipeline network topology map is constructed based on the pipeline segment number, the starting node number of the pipeline segment, and the ending node number of the pipeline segment. The connectivity of the edges in the natural gas pipeline network topology map is updated based on the pipeline segment number to which the valve belongs, the valve number affected by construction, and the planned opening and closing status of the valve affected by construction. If the planned opening and closing status of the valve affected by construction is closed, the corresponding pipeline segment number to which the valve belongs is marked as an unventilable pipeline segment in the natural gas pipeline network topology map and removed, resulting in a ventilable topology map. Based on the construction-affected pipe segment number, the construction-affected pipe segment number is marked as a construction-restricted pipe segment and removed in the ventilable topology map to obtain the construction window reference topology map. In the construction window reference topology diagram, the pressure regulating station connection node number is used as the starting node set, and the proposed connection node number is used as the target node. A breadth-first search algorithm is used to find the pipeline path connected to the proposed connection node number, and the found pipeline path is determined as the backup gas supply channel. For each backup gas supply channel, read the pipe segment number contained in the backup gas supply channel, calculate the difference between the maximum allowable flow rate of each pipe segment and the pipe segment flow rate, obtain the remaining flow capacity of each pipe segment, and take the minimum value of the remaining flow capacity of each pipe segment in the same backup gas supply channel as the backup channel remaining capacity of the backup gas supply channel.

5. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 1, characterized in that, The specific process of statistically analyzing node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating stations, gas load level of construction windows, and remaining capacity of standby channels to obtain the reference operating characteristics of construction windows is as follows: Read the reference operation data segments of the N most recent construction windows corresponding to the construction stage number, using the construction stage number as the unit; The starting node number and ending node number of the pipe segment corresponding to the proposed access node number and the construction-affected pipe segment number are taken as the node statistical objects; for the node statistical objects, the node pressure margin at each unified sampling time is calculated, and the minimum value is taken to obtain the minimum node pressure margin; The pipe section numbers affected by construction and the pipe section numbers within the backup gas supply channel are used as the pipe section statistics objects; for each pipe section statistics object, the pipe section load rate at each unified sampling time is calculated, and the maximum value is taken to obtain the maximum pipe section load rate; The number of the pressure regulating station affected by construction is used as the statistical object of the pressure regulating station; for the statistical object of the pressure regulating station, the remaining pressure regulating capacity of the pressure regulating station at each unified sampling time is calculated, and the minimum value is taken to obtain the remaining pressure regulating capacity of the minimum pressure regulating station. The set of user nodes affected by construction is determined based on the proposed access node number, the pipeline segment number affected by construction, the backup gas supply channel, and the natural gas pipeline network topology. For each user node affected by construction, the gas load level of the construction window is calculated, and the average value is taken to obtain the average gas load level of the construction window. The backup gas supply channel is used as the statistical object of the backup channel; for the statistical object of the backup channel, the remaining capacity of the backup channel at each unified sampling time is calculated, and the minimum value is taken to obtain the minimum remaining capacity of the backup channel.

6. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 5, characterized in that, The specific process of training the gas supply disturbance risk prediction model for the construction window by combining the pipeline expansion collaborative planning data corresponding to the historical construction stage and the reference operation characteristics of the construction window is as follows: Using the construction stage number as a unit, a construction stage disturbance prediction feature vector is generated based on the reference operation data segments of the most recent N construction windows. The construction stage disturbance prediction feature vector includes: construction stage number, proposed expansion section number, proposed connection node number, construction-affected section number, construction-affected valve number, planned opening and closing status of construction-affected valves, construction-affected pressure regulating station number, minimum node pressure margin, maximum section load rate, minimum remaining pressure regulating capacity of pressure regulating station, average gas load level of construction window, number of backup gas supply channels, and minimum remaining capacity of backup channels. Based on the construction stage disturbance prediction feature vector corresponding to the historical construction stage, a construction disturbance training sample set is constructed; for each historical construction stage number in the construction disturbance training sample set, a training sample label is generated, which includes node pressure exceeding limit label, pipe section overload label, and insufficient backup channel label. During the subsequent actual execution of the historical construction phase, if the pressure margin of any node in the corresponding node statistics object is less than the pressure margin threshold, the node pressure exceeding the limit label of the corresponding historical construction phase number will be marked as 1; otherwise, it will be marked as 0. During the subsequent actual execution of the historical construction phase, when the load rate of any pipe segment in the corresponding pipe segment statistics object is greater than the load threshold, the overload label of the pipe segment corresponding to the historical construction phase number will be marked as 1; otherwise, it will be marked as 0. When there is no backup gas supply channel in the historical construction phase, or when the remaining capacity of the minimum backup channel is less than the backup capacity threshold, the backup channel shortage label corresponding to the historical construction phase number is marked as 1; otherwise, it is marked as 0. Using the construction disturbance training sample set as the model input features and the training sample labels as the model output labels, a gradient boosting tree classification model is trained to construct a risk prediction model for gas supply disturbance during construction windows.

7. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 6, characterized in that, The process of obtaining the comprehensive construction disturbance risk value and identifying the construction stage to be optimized based on the construction window gas supply disturbance risk prediction model is as follows: Input the construction stage disturbance prediction feature vector of the current construction stage number into the gas supply disturbance risk prediction model of the construction window, and output the node pressure exceeding the limit probability, pipeline overload probability and backup channel insufficient probability of the corresponding construction stage number. The maximum value among the probability of node pressure exceeding the limit, the probability of pipe section overload, and the probability of insufficient backup channels is taken as the comprehensive construction disturbance risk value. When the comprehensive construction disturbance risk value is less than or equal to the optimization trigger threshold, the expected construction start time and expected construction end time corresponding to the construction stage number remain unchanged, the recommended construction window is obtained, and the planned opening and closing status of the construction-affected valve remains unchanged. When the overall construction disturbance risk value exceeds the optimization trigger threshold, the construction stage number will be marked as the construction stage to be optimized.

8. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 1, characterized in that, The specific process of generating candidate construction windows and determining recommended construction windows for the construction stage to be optimized is as follows: For the construction phases to be optimized, a window adjustment plan is generated: based on the expected construction start time, expected construction end time, and preset window adjustment step size, a set of candidate construction windows is generated; and each candidate construction window maintains the same duration as the original expected construction window. For each candidate construction window in the candidate construction window set, select the N most recent construction window reference operation data segments with the same time attributes as the candidate construction window from the unified time series data, generate the corresponding construction stage disturbance prediction feature vector, input the gas supply disturbance risk prediction model of the construction window, and obtain the corresponding comprehensive construction disturbance risk value. The candidate construction window with the lowest comprehensive construction disturbance risk value in the candidate construction window set is determined as the optimal candidate construction window, and the optimal candidate construction window is determined as the recommended construction window.

9. The machine learning-based collaborative planning method for natural gas pipeline networks according to claim 1, characterized in that, The specific process of synchronously screening candidate connected valves to generate a valve switching coordination scheme is as follows: For the construction phase to be optimized, a valve switching coordination scheme is generated: valves in the backup gas supply channel that are not affected by the construction are designated as candidate connected valves and marked as openable; a construction coordination optimization scheme is generated and output based on the window adjustment scheme and the valve switching coordination scheme.

10. A machine learning-based collaborative planning system for natural gas pipeline networks, characterized in that, include: The pipeline network data preprocessing module is used to collect pipeline expansion collaborative planning data within the same natural gas pipeline network and to preprocess the pipeline expansion collaborative planning data. The window feature recognition module is used to extract the time attributes of the construction phase based on the pipeline expansion collaborative planning data, match the construction window reference operation data segments according to the time attributes, and evaluate the node pressure margin, pipeline load rate, remaining pressure regulating capacity of the pressure regulating station and the gas load level of the construction window based on the construction window reference operation data segments; update the pipeline topology based on the pipeline expansion collaborative planning data, identify the backup gas supply channel and evaluate the remaining capacity of the backup channel; The disturbance risk prediction module is used to statistically analyze node pressure margin, pipeline load rate, remaining pressure regulating capacity of pressure regulating station, gas load level of construction window and remaining capacity of standby channel to obtain reference operating characteristics of construction window. Combined with the pipeline expansion collaborative planning data corresponding to historical construction stages and the reference operating characteristics of construction window, the gas supply disturbance risk prediction model of construction window is trained. Based on the gas supply disturbance risk prediction model of construction window, the comprehensive construction disturbance risk value is obtained and the construction stage to be optimized is identified. The window valve optimization module is used to generate candidate construction windows and determine recommended construction windows for the construction stage to be optimized, and simultaneously filter candidate connected valves to generate valve switching coordination schemes.

Images