Rail transit abnormal state detection method, system, device and storage medium

Abstract

The application provides a rail transit abnormal state detection method, system, device and storage medium. The method comprises the following steps: acquiring AFC entry and exit, line network topology, train working diagram and other multi-source data, and preprocessing the multi-source data; extracting passenger riding total travel time from the preprocessed data and splitting and decoupling the passenger riding total travel time into real-time values of entry walking, exit walking, platform waiting, riding and transfer walking; and finally analyzing each time component real-time value, identifying and quantifying abnormal events, and forming an abnormal judgment result set. The application solves the problem that existing AFC data analysis cannot accurately locate the cause of travel time increase by mixing and overlapping each travel time component, realizes accurate differentiation and attribution of train operation, station passenger flow organization and other different dimension abnormalities through fine splitting and decoupling of total travel time, makes the formulation of operation scheduling strategy more targeted, and provides accurate and specific decision basis for rail transit operation and scheduling.

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology for rail transit, and in particular to methods, systems, equipment and storage media for detecting abnormal conditions in rail transit. Background Technology

[0002] With the continuous expansion of urban rail transit networks and the sustained growth of passenger flow, the requirements for intelligent and refined rail transit operation and maintenance are increasing. Currently, anomaly monitoring in rail transit mainly relies on signaling systems and manual inspections. While signaling systems can accurately monitor train positions and speeds, they cannot detect passenger congestion within stations; manual inspections suffer from delayed response and limited coverage.

[0003] Existing AFC (Automatic Fare Collection) data analysis is mostly used to count the total passenger flow in and out of the station or the OD (Origin and Destination) flow. Although some technologies can calculate the average travel time, they often mix up the time components of walking into the station, waiting, boarding, and transferring, and cannot distinguish whether the increase in travel time is caused by train delays or by passenger flow organization measures in the station (such as flow restriction and detour), resulting in a lack of targeted scheduling strategies. Summary of the Invention

[0004] The technical problem this application aims to solve is that existing AFC (Automatic Fare Collection) data analysis is mostly used to count the total passenger flow in and out of the station or the OD (Origin and Destination) flow. Although some technologies can calculate the average travel time, they often mix up the time components of walking into the station, waiting, boarding, and transferring, and cannot distinguish whether the increase in travel time is caused by train delays or by passenger flow organization measures in the station (such as flow restriction and detour), resulting in a lack of targeted scheduling strategies.

[0005] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this application provides a method, system, equipment and storage medium for detecting abnormal conditions in rail transit.

[0006] In a first aspect, the present invention discloses a method for detecting abnormal conditions in rail transit, comprising,

[0007] Acquire multi-source data, preprocess the multi-source data to obtain a preprocessed dataset, which includes AFC entry and exit data, network topology data, and train timetable data.

[0008] The total travel time of passengers is obtained from multi-source data. The total travel time is decomposed and decoupled to obtain a set of real-time values ​​of travel component times. The set of real-time values ​​of travel component times includes real-time values ​​of walking time upon entering the station, walking time upon exiting the station, waiting time on the platform, boarding time, and walking time for transfers.

[0009] Each real-time value in the real-time value set of travel component time is analyzed and processed separately to obtain abnormal events, label the abnormal events and quantify the degree of abnormality, and obtain an abnormal judgment result data set.

[0010] Preferably, the following includes:

[0011] Based on the anomaly detection result dataset, the abnormal events are visualized on an electronic map to obtain a map with the abnormal events, and alarm information is output to the backend server in combination with the abnormal events.

[0012] Preferably, the process involves acquiring multi-source data, preprocessing the multi-source data to obtain a preprocessed dataset, specifically including the following steps:

[0013] AFC (Automatic Fare Collection) entry and exit data, network topology data, and train timetable data are collected from existing rail transit business systems, and preliminarily integrated according to preset field formats to establish a raw data pool and obtain multi-source data.

[0014] AFC entry and exit data includes core fields such as card number, entry and exit time, and station ID; network topology data includes core fields such as station connection relationship, mileage between stations, and transfer station identification; and train timetable data includes core fields such as theoretical departure interval, first and last train time, and theoretical running time of each section.

[0015] The data source data is integrated and processed for data format integration, verification and invalid data removal. Passenger OD path matching is completed based on the network topology data to obtain a preprocessed data set.

[0016] The preprocessed dataset is categorized by OD path and time dimension and stored in a designated database.

[0017] Preferably, the total travel time of passengers is obtained from multi-source data, and the total travel time is decomposed and decoupled to obtain a set of real-time values ​​of travel component times. Specifically, this includes the following steps:

[0018] Extract passenger card swipe data from multiple sources, including origin station, destination station, and transfer station. Calculate the total travel time for any origin-destination (OD) path based on the multi-source data.

[0019] The total travel time is combined with multi-source data for component decoupling processing to obtain the baseline value of the component time and the real-time value of the travel component time.

[0020] The component time base value and the travel component time real-time value are classified according to OD path and time dimension to obtain the component time base value set and the travel component time real-time value set. The component time base value set and the travel component time real-time value set are stored in the designated database.

[0021] Preferably, the total travel time is processed by combining multi-source data for component decoupling to obtain the baseline value of the component time and the real-time value of the travel component time. Specifically, this includes the following steps:

[0022] Define the time components of the total travel time for any OD path.

[0023] Using non-congested periods as the time benchmark, passengers with the shortest travel time along the same OD path are selected as unobstructed samples from AFC entry and exit data from multiple sources.

[0024] Based on unobstructed samples, the baseline values ​​of each rigid physical movement time component and the overall rigid physical movement time baseline value are calculated to obtain the baseline values ​​of the component times.

[0025] Based on the calculated rigid physical movement time benchmark value, combined with real-time AFC entry and exit data, the total real-time travel time of passengers is separated into components one by one, realizing the separation of rigid components and elastic components, as well as the individual quantification of each subdivided rigid component, to obtain the real-time value of the travel component time.

[0026] Preferably, based on the calculated rigid physical movement time baseline value and combined with real-time AFC effective flow data, the total real-time travel time of passengers is component-by-component, realizing the separation of rigid components and elastic components, and the individual quantization of each subdivided rigid component, to obtain the real-time value of the travel component time. Specifically, this includes the following steps:

[0027] By obtaining historical baseline values ​​and combining them with train operation data, including the actual travel time of trains in each section, the real-time value of transfer travel time can be obtained.

[0028] Obtain historical baseline values, card swipe records, boarding and alighting stations, boarding and alighting times. Plan passenger routes based on boarding and alighting stations. First, determine whether the passenger is a transfer passenger. For transfer passengers, combine transfer card swipe records, boarding card swipe records, and train departure and stop times from AFC entry and exit data to calculate the real-time transfer walking time.

[0029] Obtain historical baseline values ​​and real-time passenger flow density at passenger entry and exit stations. Use the baseline values ​​during off-peak hours and adjust according to passenger flow density during peak hours. Output the real-time values ​​according to the adjusted values ​​to obtain real-time values ​​of entry and exit walking time.

[0030] The real-time values ​​of transfer walking time, station entry walking time, and station exit walking time are summed to obtain the rigid component sum. Based on the real-time passenger flow data changes in the multi-source data, the line congestion time is obtained. The real-time waiting time at the platform is calculated by combining the total duration with the rigid component sum.

[0031] Preferably, the real-time data of each travel component time in the real-time value set are analyzed and processed separately to obtain abnormal events, label the abnormal events and quantify the degree of abnormality, and obtain an abnormality judgment result data set, specifically including the following steps:

[0032] Based on the collected real-time transfer walking time, the interval travel time is calculated according to the station. The mean and variance of the interval travel time for all passengers are calculated. The interval travel time is compared with the historical benchmark. When the mean is larger than the historical benchmark and the variance is smaller than the historical benchmark, it is determined that the interval train operation efficiency is abnormal.

[0033] Based on the real-time values ​​of waiting time on the platform collected, frequency domain analysis or histogram statistical analysis is performed on the real-time values ​​of waiting time on the platform of each station. Comb-like multi-peak characteristics are identified in the frequency domain analysis data or histogram, and the peak interval is approximately equal to the departure interval. This indicates that there is a vehicle platform congestion, and the average number of congestion cycles is calculated.

[0034] Based on the collected real-time transfer walking time values, compared with historical baseline values, if the real-time transfer walking time value is greater than the historical baseline value, and cross-validation shows no abnormalities in the real-time transfer walking time value and platform waiting time value of the same OD path, it is determined that the transfer path has physically changed and the transfer path is abnormal.

[0035] Extract passenger flow data for each section and match data between actual passenger routes and regular routes. If passenger flow in the downstream section is zero, or a large number of passengers exit at intermediate stations and then re-enter, it is determined to be an abnormal route execution.

[0036] For the identified abnormal events, the abnormality type, location, and time of occurrence are marked, and the degree of abnormality is calculated to form an abnormality judgment result data set.

[0037] Secondly, the present invention discloses a rail transit abnormal state detection system, which includes the above-mentioned rail transit abnormal state detection method.

[0038] Thirdly, the present invention discloses an electronic device, which includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0039] Memory, used to store computer programs;

[0040] When the processor executes the program stored in the memory, it implements the steps of the above-described method for detecting abnormal states in rail transit.

[0041] Fourthly, the present invention discloses a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for detecting abnormal states in rail transit.

[0042] The technical solution provided in this application has the following advantages compared with the prior art:

[0043] The method, system, equipment, and storage medium for detecting abnormal states in rail transit provided in this application describe a process whereby the following steps are taken: First, multi-source data, including AFC (Automatic Fare Collection) entry / exit data, network topology, and train timetables, are acquired and preprocessed. Then, the total passenger travel time is extracted from the preprocessed data and decoupled into real-time values ​​for walking into the station, walking out of the station, waiting on the platform, boarding the train, and walking for transfers. Finally, the real-time values ​​of each time component are analyzed separately to identify and quantify abnormal events, forming an abnormality judgment result set. This solves the problem of existing AFC data analysis involving overlapping travel time components and the inability to accurately pinpoint the causes of increased travel time. By finely decomposing and decoupling the total travel time, it achieves accurate differentiation and attribution of abnormalities in different dimensions, such as train operation and passenger flow organization within stations. This makes the formulation of operation and scheduling strategies more targeted and provides accurate and specific decision-making basis for rail transit operation and maintenance scheduling.

[0044] Furthermore, in the existing technology, when passengers' travel time becomes longer, it is difficult to determine whether the problem is due to equipment failure (the train is running slowly) or management factors (transfer channels are changed to detours, security checks are upgraded), resulting in a lack of targeted scheduling strategies. This invention can accurately distinguish between train problems (speed limits) and station problems (detours / delays), providing a direct basis for operational scheduling decisions.

[0045] Furthermore, existing technologies make it difficult to quantify exactly how many trains passengers waited to board, and cannot accurately assess the severity of platform congestion. Comb distribution analysis, however, can specifically quantify how many trains passengers were delayed, reflecting the true passenger experience better than a simple percentage of crowding.

[0046] Furthermore, existing technologies lack automatic sensing methods for physical isolation measures of transfer channels (such as temporarily changing platform transfers to concourse transfers), and usually rely on manual reporting, resulting in a delayed response. The solution proposes a method to use AFC data to infer changes in physical transfer routes, filling the gap in the lack of digital supervision of temporary passenger flow organization measures (such as detours by barricades and changes in transfer channels).

[0047] Furthermore, without any additional hardware investment, second-level monitoring of the entire network can be achieved simply by leveraging the value of existing data. Attached Figure Description

[0048] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 A flowchart of the steps for a method for detecting abnormal conditions in rail transit provided in this application;

[0051] Figure 2 A flowchart illustrating a method for detecting abnormal conditions in rail transit provided in this application;

[0052] Figure 3 A flowchart illustrating the specific steps of step S1 in the rail transit abnormal state detection method provided in this application;

[0053] Figure 4 A flowchart detailing step S2 of the method for detecting abnormal conditions in rail transit provided in this application;

[0054] Figure 5 A flowchart illustrating step S2 of the method for detecting abnormal states in rail transit provided in this application;

[0055] Figure 6 A flowchart detailing step S3 of the rail transit abnormal state detection method provided in this application;

[0056] Figure 7 A flowchart illustrating step S3 of the method for detecting abnormal conditions in rail transit provided in this application;

[0057] Figure 8 This application provides a system architecture diagram of an abnormal state detection system for rail transit. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] Firstly, see Figures 1-7 This invention discloses a method for detecting abnormal conditions in rail transit, comprising,

[0060] Step S1: Obtain multi-source data, preprocess the multi-source data to obtain a preprocessed data set. The multi-source data includes AFC entry and exit data, network topology data, and train timetable data.

[0061] Step S2: Obtain the total travel time of passengers from multi-source data, decompose and decouple the total travel time to obtain a set of real-time values ​​of travel component times. The set of real-time values ​​of travel component times includes real-time values ​​of walking time upon entering the station, walking time upon exiting the station, waiting time on the platform, boarding time, and walking time for transfers.

[0062] Step S3: Analyze and process each real-time value data in the real-time value set of travel component time, obtain abnormal events, label abnormal events and quantify the degree of abnormality, and obtain an abnormal judgment result data set.

[0063] Specifically, in step S1, AFC entry and exit data, network topology data, and train timetable data are collected from the existing rail transit business system. The data is then cleaned (invalid card swipes and extreme value anomalies are removed), integrated (OD paths are associated with network topology), and standardized (field formats and storage structures are unified) to form a preprocessed data set that can be directly accessed. Through multi-source data fusion and standardization, the consistency and availability of the data are ensured, laying a solid data foundation for the entire detection process and avoiding subsequent analysis deviations due to data quality issues.

[0064] Specifically, in step S2, the total travel time of passengers is extracted from the preprocessed data. Based on unobstructed samples, the baseline values ​​of each time component are extracted. The total travel time is decoupled into a set of real-time values ​​for station entry walking time, station exit walking time, platform waiting time, boarding time, and transfer walking time through an algorithm. The general total travel time is refined into rigid physical movement time and flexible waiting time, clarifying the physical meaning and boundaries of each component. This breaks through the limitation of time component overlap in the existing technology, realizes accurate attribution of travel time, provides key data dimensions for distinguishing between train operation abnormalities and station passenger flow organization abnormalities, and improves the granularity and accuracy of abnormality identification.

[0065] Specifically, in step S3, the real-time values ​​of each time component are compared with the historical benchmark values ​​in the benchmark model library. Through the judgment rules of the anomaly identification logic library, the fluctuation characteristics and distribution patterns of each component are analyzed, abnormal events are identified and labeled with their types and the degree of anomaly is quantified, forming an anomaly judgment result data set. This accurately locates anomalies in different dimensions such as train operation efficiency, platform capacity matching, passenger flow organization and physical paths, and route execution, providing targeted basis for operation scheduling. It realizes accurate attribution and quantitative analysis of anomalies, solves the problems of ambiguous anomaly attribution and lack of targeted scheduling in existing technologies, makes operation decisions more scientific and operable, and significantly improves the intelligent level of rail transit operation and maintenance.

[0066] The approach involves first acquiring multi-source data such as AFC (Automatic Fare Collection) entry / exit data, network topology, and train timetables. This data is then preprocessed. Next, the total passenger travel time is extracted from the preprocessed data and decoupled into real-time values ​​for walking into the station, walking out of the station, waiting on the platform, boarding the train, and walking for transfers. Finally, the real-time values ​​of each time component are analyzed separately to identify and quantify abnormal events, forming an anomaly determination result set. This solves the problem of existing AFC data analysis where various travel time components are mixed and the causes of increased travel time cannot be accurately located. By finely decomposing and decoupling the total travel time, it achieves accurate differentiation and attribution of anomalies in different dimensions such as train operation and passenger flow organization within stations. This makes the formulation of operation and scheduling strategies more targeted and provides accurate and specific decision-making basis for rail transit operation and maintenance scheduling.

[0067] Furthermore, in the existing technology, when passengers' travel time becomes longer, it is difficult to determine whether the problem is due to equipment failure (the train is running slowly) or management factors (transfer channels are changed to detours, security checks are upgraded), resulting in a lack of targeted scheduling strategies. This invention can accurately distinguish between train problems (speed limits) and station problems (detours / delays), providing a direct basis for operational scheduling decisions.

[0068] Furthermore, existing technologies make it difficult to quantify exactly how many trains passengers waited to board, and cannot accurately assess the severity of platform congestion. Comb distribution analysis, however, can specifically quantify how many trains passengers were delayed, reflecting the true passenger experience better than a simple percentage of crowding.

[0069] Furthermore, existing technologies lack automatic sensing methods for physical isolation measures of transfer channels (such as temporarily changing platform transfers to concourse transfers), and usually rely on manual reporting, resulting in a delayed response. The solution proposes a method to use AFC data to infer changes in physical transfer routes, filling the gap in the lack of digital supervision of temporary passenger flow organization measures (such as detours by barricades and changes in transfer channels).

[0070] Furthermore, without any additional hardware investment, second-level monitoring of the entire network can be achieved simply by leveraging the value of existing data.

[0071] Step S3 is followed by:

[0072] Step S4: Based on the anomaly determination result data set, visualize the anomaly events on the electronic map to obtain a map with the anomaly events, and output alarm information to the backend server in combination with the anomaly events.

[0073] Specifically, the abnormal locations (stations, sections, transfer stations) in the anomaly judgment result data set are first precisely bound to the geographic coordinates of the rail transit electronic map to establish a correspondence between abnormal events and geographic space. Then, the visualization is rendered according to preset rules, with different colors used to indicate the severity of the anomaly (e.g., red for severe, yellow for moderate) and exclusive icons to distinguish the anomaly type (e.g., snail icon for train speed limit, fence icon for transfer detour, crowd icon for platform congestion), generating an electronic map with the abnormal events. At the same time, core information (anomaly location, type, severity, occurrence time, and scale of passenger flow involved) is extracted from the anomaly judgment results to generate structured alarm information. Finally, the visualization results are displayed in real time through the electronic map monitoring screen, and the alarm information is pushed to the backend server and operation dispatch terminal through system messages, SMS, API interfaces, etc., supporting clicking on the anomaly icon to view detailed judgment basis and related data. This step transforms abstract, structured anomaly data into intuitive, interactive visualizations, enabling spatial and concrete presentation of anomaly events. This allows operations and dispatch personnel to quickly grasp the distribution and status of anomalies across the network. Simultaneously, multi-channel alarm push notifications ensure timely delivery of anomaly information, preventing the omission of critical events and providing direct support for rapid response and precise dispatch. This approach breaks through the limitations of traditional pure data reports in information transmission. Visualization significantly improves the readability and transmission efficiency of anomaly information. Precise visual identifiers (colors, icons) reduce the probability of misjudgment and improve the accuracy of dispatch decisions. Multiple output formats (map display, alarm push notifications, and detailed linkage) cover the usage needs of different scenarios, significantly improving the response speed and intelligence level of rail transit operations and maintenance.

[0074] Step S1 specifically includes the following steps:

[0075] Step S11: Collect AFC entry and exit data, network topology data, and train timetable data from the existing rail transit business system, perform preliminary integration according to the preset field format, establish the raw data pool, and obtain multi-source data;

[0076] Step S12: AFC entry and exit data includes core fields such as card number, entry and exit time, and station ID; network topology data includes core fields such as station connection relationship, mileage between stations, and transfer station identification; and train timetable data includes core fields such as theoretical departure interval, first and last train time, and theoretical running time of each section.

[0077] Step S13: Perform data format integration, verification and invalid data removal on the multi-source data, complete passenger OD path matching based on the network topology data, and obtain the preprocessed data set;

[0078] Step S14: Classify the preprocessed data set according to OD path and time dimension, and store it in the specified database.

[0079] Specifically, AFC entry and exit data, including card number, entry and exit time, and station ID, is collected uniformly from the existing rail transit business system; network topology data, including station connection relationships, inter-station mileage, and transfer station identification, is collected; and train timetable data, including theoretical departure intervals, first and last train times, and theoretical running time of sections, is collected. The three types of data are first initially integrated according to preset field formats to establish an original data pool. Then, the multi-source data in the pool is integrated in a unified format across all dimensions, and validity is verified and invalid data is removed. At the same time, the network topology data is used to complete the accurate matching of passenger origin-destination (OD) paths, forming a standardized pre-processed data set. Finally, the dataset is classified and sorted according to OD path and time dimension, and uniformly stored in a designated database to complete the standardized management of multi-source data. This step achieved unified collection and integration of multi-source heterogeneous business data in rail transit. Through standardized format processing and rigorous invalid data cleaning, the integrity, accuracy, and consistency of the basic data were effectively guaranteed. Based on the network topology, OD path matching was completed, establishing a precise correlation between passenger journeys and the network, laying a data foundation that is associative and traceable for subsequent time component decoupling. The classification and storage method according to OD path and time dimension not only improved the efficiency of subsequent data retrieval and calculation, but also made data management more organized. At the same time, the entire process was based on the existing business system of rail transit, without the need for additional hardware investment, thus balancing the professionalism of data processing with the low cost of implementation.

[0080] The process involves integrating and processing multi-source data, verifying and removing invalid data, and matching passenger origin-destination (OD) paths based on the network topology data. This includes the following steps: validating AFC data by removing invalid card swipes such as missing card numbers, reversed entry / exit times, and invalid station IDs; setting a time threshold (staying longer than 4 hours); filtering out extreme abnormal data in AFC entry / exit data where passenger single-trip time exceeds a reasonable range; associating each passenger's entry and exit stations based on the network topology to match a unique OD path, thus binding passenger trips to the network; and storing the processed data in a unified format in a database (Oracle / MySQL) to provide a directly callable dataset for subsequent steps.

[0081] Step S2 specifically includes the following steps:

[0082] Step S21: Extract the card swiping data of the passenger's origin station, destination station, and transfer station from the multi-source data, and calculate the passenger's total travel time for any OD path based on the multi-source data;

[0083] Step S22: Perform component decoupling processing on the total travel time by combining multi-source data to obtain the baseline value of the component time and the real-time value of the travel component time;

[0084] Step S23: Classify the component time base value and the travel component time real-time value according to OD path and time dimension to obtain the component time base value set and the travel component time real-time value set. Store the component time base value set and the travel component time real-time value set in the designated database.

[0085] Specifically, the card-swiping data of passengers at the origin, destination, and transfer stations are accurately extracted from the preprocessed multi-source data. Based on this core data, the total travel time for a single trip of passengers along any OD (origin-destination) path of the rail transit system is calculated uniformly. Subsequently, combined with multi-source data such as the network topology and train timetable, the calculated total travel time undergoes professional component decoupling processing, breaking it down into time components for entering the station, exiting the station, waiting on the platform, boarding the train, and transferring. At the same time, the historical baseline value and real-time actual value of each component time are calculated simultaneously. Finally, the time baseline value of each component and the real-time value of the travel component time are systematically classified and sorted according to the OD path and time dimension, forming a set of component time baseline values ​​and a set of travel component time real-time values, respectively. These two sets of data are uniformly stored in a designated database, completing the standardized retention and management of the decoupled data, and providing directly accessible quantitative data for subsequent anomaly analysis. This step overcomes the limitations of existing technologies that lump together various travel time components. It achieves refined and precise decoupling of total passenger travel time for any OD path, while simultaneously acquiring baseline and real-time values ​​for each component. This establishes a core reference system for quantitative comparison in subsequent anomaly detection, effectively distinguishing between train operation-related time and station passenger flow organization-related time from a data perspective, laying a crucial foundation for accurate anomaly attribution. By classifying the decoupled data according to OD path and time dimension and forming a standardized set, the efficiency of subsequent data retrieval and comparative analysis is significantly improved, adapting to the monitoring needs of multiple paths and time periods in the rail transit network. The baseline and real-time value sets are uniformly stored in a designated database, achieving standardized and systematic management of decoupled data, ensuring data traceability and reusability, and connecting with the storage system of pre-processed data, making the data flow of the entire system more coherent.

[0086] The specific steps for step S2 are as follows:

[0087] First, for any origin-destination (OD) path in rail transit (originating station O, destination station D), the total travel time for a single trip is defined as:

[0088] ;

[0089] In the formula, The station entry walking time component refers to the pure walking time from when a passenger swipes their card at the origin station to when they arrive at the corresponding platform. The platform waiting time component (elastic component) refers to the waiting time from when a passenger arrives at the platform until the train departs, including the total waiting time at platforms of all lines in the case of transfers. The travel time component refers to the pure travel time from the departure of the passenger train to the arrival of the train at the drop-off station / transfer station / terminal station, including the total travel time of each line in the case of transfer. The transfer walking time component refers to the pure walking time for passengers to complete the line change, and is divided into same-station transfer walking time according to the transfer type. Travel time between different stations ; The exit walking time component refers to the pure walking time of a passenger from the time the train arrives at the platform to the time they swipe their card at the final station, defining rigid physical movement time. (Pure travel time with no waiting or congestion) and flexible waiting time (Including various types of waiting, congestion, and delays), the formula is:

[0090] ;

[0091] .

[0092] Then, using the non-congested operating hours of rail transit as the time base, passenger journey samples under any OD path are extracted from the AFC entry and exit data. The 5%-10% of samples with the shortest total journey time are selected as the unobstructed sample set. Interference factors such as waiting, congestion, and temporary delays are excluded. Based on the unobstructed sample set, the baseline values ​​(denoted as superscript 0) of each rigid physical movement time component, the overall rigid physical movement time baseline value, and the station entry walking time baseline value are calculated. Baseline value of walking time upon exiting the station Travel time benchmark Baseline value of walking time for transfers at the same station (Calculated based on the combination of transfer stations and lines), baseline values ​​for walking time at different stations. (Calculated based on "transfer station pair + transfer method"), the formula for calculating the overall rigid physical movement time reference value is as follows:

[0093] ;

[0094] In the formula The value is determined based on whether the transfer is at the same station: For transfers at the same station, the value is... When transferring between different stations, take .

[0095] Next, using the reference values ​​of each rigid physical movement time component as the anchoring basis, and combining real-time AFC entry and exit data, network topology data, and train timetable data, the total real-time travel time of passengers under any OD path is calculated. Component-by-component stripping was performed to obtain the real-time values ​​of each time component (denoted by the superscript t), and differentiated processing was applied for scenarios with no transfer, same-station transfer, and different-station transfer. Among them, the real-time value of the station entry walking time Calculation with Based on the baseline, adjustments are made using real-time passenger flow density at the originating station (adjustment factor). ∈[1.0,1.2]), that is Based on the network topology, passenger journeys are divided into several inter-station sections, and the information from each section is extracted. By combining the actual train travel time corrections with the train timetable, the real-time value of the entire journey travel time is obtained. ;by Based on the baseline, adjustments are made by combining the real-time passenger flow density at the terminal station (adjustment factor). ∈[1.0,1.2]), which is the real-time value of the exit walking time. ;

[0096] Real-time value of transfer walking time It is necessary to decouple the data based on the transfer type. For same-station transfer scenarios, passengers complete line switching within the same station, without exiting or moving across stations, generating only one entry and one exit card swipe record. First, extract the corresponding "transfer station + line combination" data for this scenario. Then, match the passenger's disembarkation time from the train timetable before the transfer. and the time of boarding the train after the transfer Calculate the time difference Excluding waiting time at transfer stations ,get ;like and If the deviation is within a reasonable threshold of ±10%, take... These are real-time values; if they exceed the specified value, the actual calculated value will be used. For different station transfer scenarios, passengers switch lines across stations, including cross-station travel / short-distance travel, generating multiple entry / exit card swipe records. First, extract the transfer station pairs and corresponding transfer methods for that scenario. Then, match the passenger's exit time at transfer station 1 with the AFC entry and exit data. Entry time at transfer station 2 Calculate the time difference between stations After removing cross-station travel time (transfer by train), invalid walking time (walking transfer), and gate card swiping time, the non-walking time is obtained. ;like and If the deviation is within a reasonable threshold of ±10%, take... This is a real-time value; if it exceeds this value, the actual calculated value will be used. Finally, the real-time value of the platform waiting time. The final separation of the elastic and rigid components is achieved by using the difference between the total travel time and the real-time value of the rigid physical movement time. The formula is as follows:

[0097] ;

[0098] In the formula Based on whether the transfer is at the same station, it is divided into two categories: when transferring at the same station, take... Take when transferring between different stations After completing the above decoupling, it is necessary to verify the rationality of the obtained real-time values ​​of each time component and validate the formula. If the sum of the components is true, then... If the deviation is ≤60 seconds, the decoupling result is considered valid; if the deviation exceeds the threshold, ... Perform difference correction to ensure the accuracy and completeness of the decoupling results.

[0099] Step S22 specifically includes the following steps:

[0100] Step S221: Define the time components of the total travel time for any OD path;

[0101] Step S222: Using non-congested periods as the time base, select passengers with the shortest travel time under the same OD path from the AFC entry and exit data of multi-source data as unobstructed samples.

[0102] Step S223: Based on the unobstructed sample, calculate the reference value of each rigid physical movement time component and the overall rigid physical movement time reference value to obtain the reference value of the component time;

[0103] Step S224: Based on the calculated rigid physical movement time reference value, combined with real-time AFC entry and exit data, the total real-time travel time of passengers is separated into components to achieve the separation of rigid components and elastic components, as well as the individual quantification of each subdivided rigid component, to obtain the real-time value of the travel component time.

[0104] Specifically, for any origin-destination (OD) path in rail transit, the specific connotations and boundaries of the rigid physical movement time components (entry walking, exit walking, boarding, and transfer walking) and the flexible waiting time components (platform waiting) within the total passenger travel time are clearly defined. Then, using non-congested periods as a reference, the passenger journey with the shortest travel time under the same OD path is selected from AFC entry and exit data in multi-source data as an unobstructed sample, excluding interference factors such as waiting and congestion. Next, based on this unobstructed sample, the individual benchmark values ​​of each rigid physical movement time component and the comprehensive benchmark value of the overall rigid physical movement time are accurately calculated, forming component time benchmark values ​​that can be used as a reference for anomaly detection. Finally, using the calculated rigid physical movement time benchmark values ​​as anchoring criteria, combined with real-time AFC entry and exit data, the total real-time travel time of passengers is decomposed and separated component by component, achieving effective separation of rigid physical movement components and flexible waiting components. Simultaneously, the individual quantification of each subdivided rigid physical movement time component is completed, ultimately obtaining real-time travel component time values ​​that reflect the real-time operational status. This step first clearly defines each time component, clarifying the boundaries of different travel times from the root and avoiding component confusion. It then selects the shortest unobstructed sample along the same OD path during non-congested periods to calculate a baseline value, ensuring the objectivity and accuracy of the baseline value and providing a reliable quantitative reference standard for subsequent anomaly identification. By anchoring the baseline value and combining it with real-time data, component-by-component stripping is performed, achieving precise separation of rigid and flexible time components and individual quantification of each subdivided rigid component. This completely overcomes the limitation of existing technologies that conflate various travel time components, clearly distinguishing the time consumption of different dimensions such as train operation, station walking, and transfers, laying a core data foundation for accurately determining the causes of increased travel time. Simultaneously, the entire process is conducted for any OD path, adapting to the full-path monitoring needs of the rail transit network. The decoupled real-time values ​​accurately reflect the actual operational status of each link, improving the granularity and accuracy of subsequent anomaly identification.

[0105] Step S224 specifically includes the following steps:

[0106] Step S2241: Obtain historical benchmark values, combine them with train operation data, including the actual travel time of the train in each section, and combine the historical benchmark values ​​with the train operation data to obtain the real-time value of transfer travel time.

[0107] Step S2242: Obtain historical baseline values, obtain card swipe records, boarding and alighting stations, boarding and alighting times, plan the passenger's travel route based on the boarding and alighting stations, first determine whether the passenger is a transfer passenger, for transfer passengers, combine the transfer card swipe records, travel card swipe records, and train departure and stop times of the AFC entry and exit data to calculate the real-time value of the transfer travel time.

[0108] Step S2243: Obtain historical baseline values ​​and real-time passenger flow density at passenger entry and exit stations. Use the baseline values ​​during off-peak hours and adjust according to passenger flow density during peak hours. Output the real-time values ​​according to the adjusted values ​​to obtain the real-time values ​​of entry and exit walking time.

[0109] Step S2244: The real-time values ​​of transfer walking time, station entry walking time and exit walking time are summed to obtain the rigid component sum. The line congestion time is obtained based on the real-time passenger flow data changes in the multi-source data. The platform waiting time is calculated by combining the total duration and the rigid component sum.

[0110] Specifically, based on historical benchmark values ​​and combined with multi-source data such as train operation, AFC card swiping, and passenger flow density, the various time components are decoupled step by step: First, the historical benchmark values ​​are read and integrated with the actual running time of each section of the train to calculate the real-time transfer walking time; then, based on card swiping records, boarding and alighting stations, and time planning, the passenger route is planned to determine whether the passenger is transferring. For transferring passengers, the real-time transfer walking time is calculated by combining transfer records, boarding records, and train arrival and departure times; subsequently, the historical benchmark values ​​are dynamically corrected according to the real-time passenger flow density of the entry and exit stations. During off-peak hours, the benchmark values ​​are directly used, while during peak hours, the values ​​are adaptively adjusted according to passenger flow density to obtain the real-time entry and exit walking times; finally, the real-time transfer walking time, entry walking time, and exit walking time are summed to obtain the sum of the rigid components. Combined with the real-time passenger flow changes of the line, the congestion time is obtained, and the total travel time is subtracted from the sum of the rigid components to finally calculate the real-time platform waiting time, thus completing the decoupling and quantification of all time components. This step achieves refined and quantifiable decoupling of total travel time. By linking baseline values ​​with real-time data, it ensures the accuracy and robustness of each component. It can automatically distinguish between transfer and non-transfer scenarios, accurately identify and calculate transfer times, and adapt to complex origin-destination (OD) travel routes. It introduces passenger flow density to dynamically correct travel time, making the results more consistent with the actual on-site operation status. It obtains platform waiting time by summing and differing rigid components, which is logically rigorous and highly computable. It fundamentally distinguishes different sources of delays, such as train operation delays, station walking, transfer organization, and platform waiting, providing reliable data support for subsequent accurate identification and attribution of anomalies.

[0111] Step S3 specifically includes the following steps:

[0112] Step S31: Based on the collected real-time transfer walking time, calculate the real-time interval travel value according to the station, calculate the mean and variance of the real-time interval travel value for all passengers, and compare the real-time interval travel value with the historical benchmark value. When the mean is larger than the historical benchmark value and the variance is smaller than the historical benchmark value, it is determined that the interval train operation efficiency is abnormal.

[0113] Step S32: Based on the collected real-time values ​​of platform waiting time, perform frequency domain analysis or draw histogram statistical analysis on the real-time values ​​of platform waiting time for each station. In the frequency domain analysis data or histogram, a comb-like multi-peak feature is identified, and the peak interval is approximately equal to the departure interval. It is determined that there is a vehicle platform congestion, and the average number of congestion cycles is calculated.

[0114] Step S33: Based on the collected real-time transfer walking time value, compare it with the historical baseline value. If the real-time transfer walking time value is greater than the historical baseline value, and the cross-validation of the real-time transfer walking time value and the real-time waiting time value of the same OD path is normal, it is determined that the transfer path has been physically changed and the transfer path is abnormal.

[0115] Step S34: Extract passenger flow data for each section and matching data between actual passenger routes and regular routes. If passenger flow in the downstream section is zero, or a large number of passengers exit at intermediate stations and then re-enter, it is determined to be an abnormal route execution.

[0116] Step S35: For the identified abnormal events, mark the abnormality type, location, and time of occurrence, and calculate the degree of abnormality to form an abnormality judgment result data set.

[0117] Specifically, based on the real-time values ​​of each time component obtained by decoupling the travel time components, combined with historical benchmark values, network passenger flow data, and path matching data, abnormal events are identified, judged, and quantified in multiple dimensions: First, for the real-time value of transfer walking time, the real-time value of boarding in each section is calculated by station, and the mean and variance of the real-time values ​​of all passengers in that section are statistically analyzed. The real-time values ​​of boarding in each section are compared with historical benchmark values. When the mean increases compared to the historical benchmark value and the variance decreases compared to the historical benchmark value, it is judged that the train operation efficiency of the section is abnormal. Then, frequency domain analysis or histogram statistical analysis is performed on the real-time values ​​of platform waiting time at each station. If a comb-like multi-peak feature is identified and the peak interval is approximately equal to the train departure interval, it is judged that there is a platform vehicle congestion and the average congestion is calculated. The system counts the number of travel cycles. Then, it compares the real-time transfer walking time with historical baselines. If the real-time transfer walking time is greater than the historical baseline, and cross-validation shows no anomalies in the real-time transfer walking time and platform waiting time along the same OD route, it is determined that the transfer route has physically changed, i.e., the transfer route is abnormal. Simultaneously, it extracts passenger flow data for each section and matching data between actual passenger routes and regular routes. If downstream passenger flow drops to zero, or a large number of passengers exit and re-enter the station at intermediate stations, it is determined that the route execution is abnormal. Finally, for all the abnormal events identified above, it labels the anomaly type, location, and time of occurrence, quantifies the degree of anomaly (e.g., the percentage increase in travel time, average number of travel cycles), and integrates them to form a structured set of anomaly judgment results. This step designs differentiated judgment logic for different anomaly types, accurately covering four core operational scenarios: train operation, platform capacity, transfer organization, and route execution, to achieve the classification, identification, and accurate attribution of abnormal events. Through various analytical methods such as mean-variance comparison, frequency domain / histogram analysis, and cross-validation, the accuracy of anomaly judgment is improved, effectively avoiding misjudgments and omissions. Abnormal events are comprehensively labeled and quantified, providing clear and specific basis for subsequent anomaly visualization, alarms, and scheduling decisions, completely solving the pain points of existing AFC data analysis's inability to distinguish anomaly sources and the lack of targeted scheduling. The entire judgment process relies on decoupled, subdivided time components, with rigorous and highly operable judgment logic, adapting to the anomaly monitoring needs of multiple sections and scenarios in the rail transit network, and improving the intelligence and precision of operation and maintenance scheduling.

[0118] Secondly, see Figure 8 The present invention discloses a rail transit abnormal state detection system, which includes the rail transit abnormal state detection method disclosed in the first aspect above.

[0119] Specifically, the system implements the rail transit anomaly detection method disclosed in the first aspect. The method involves first acquiring multi-source data such as AFC (Automatic Fare Collection) entry / exit data, network topology, and train timetables, and preprocessing this data. Then, it extracts the total passenger travel time from the preprocessed data and decouples it into real-time values ​​for walking into the station, walking out of the station, waiting on the platform, boarding the train, and walking for transfers. Finally, it analyzes the real-time values ​​of each time component separately, identifies and quantifies abnormal events, and forms an anomaly judgment result set. This solves the problem of existing AFC data analysis involving overlapping travel time components and the inability to accurately pinpoint the causes of increased travel time. Through refined decomposition and decoupling of the total travel time, it achieves accurate differentiation and attribution of anomalies in different dimensions such as train operation and passenger flow organization within stations, making the formulation of operation and scheduling strategies more targeted and providing accurate and specific decision-making basis for rail transit operation and maintenance scheduling.

[0120] Furthermore, in the existing technology, when passengers' travel time becomes longer, it is difficult to determine whether the problem is due to equipment failure (the train is running slowly) or management factors (transfer channels are changed to detours, security checks are upgraded), resulting in a lack of targeted scheduling strategies. This invention can accurately distinguish between train problems (speed limits) and station problems (detours / delays), providing a direct basis for operational scheduling decisions.

[0121] Furthermore, existing technologies make it difficult to quantify exactly how many trains passengers waited to board, and cannot accurately assess the severity of platform congestion. Comb distribution analysis, however, can specifically quantify how many trains passengers were delayed, reflecting the true passenger experience better than a simple percentage of crowding.

[0122] Furthermore, existing technologies lack automatic sensing methods for physical isolation measures of transfer channels (such as temporarily changing platform transfers to concourse transfers), and usually rely on manual reporting, resulting in a delayed response. The solution proposes a method to use AFC data to infer changes in physical transfer routes, filling the gap in the lack of digital supervision of temporary passenger flow organization measures (such as detours by barricades and changes in transfer channels).

[0123] Furthermore, without any additional hardware investment, second-level monitoring of the entire network can be achieved simply by leveraging the value of existing data.

[0124] The system comprises an application presentation layer, a computing engine layer, and a data access layer. The data access layer interfaces with existing rail transit business systems through a data synchronization interface to complete the unified collection and standardized storage of multi-source data. Then, it pushes pre-processed AFC entry and exit data, network topology, and train timetable data to the computing engine layer via a RESTful API or real-time data synchronization interface. After receiving the underlying data, the computing engine layer uses a decoupling algorithm engine to perform time component decoupling based on a benchmark model library and calls an anomaly identification logic library for judgment and analysis. Finally, it transmits core calculation signals such as the anomaly judgment result set and real-time values ​​of time components to the application presentation layer through a real-time message queue or batch data interface. The application presentation layer receives the signals from the computing layer and performs visualization rendering on the electronic map monitoring screen, alarm list, and statistical reports. It also supports the feedback of user interaction commands to the computing engine layer.

[0125] The data access layer includes an AFC database, an Oracle / MySQL general database, and a network basic parameter library. The AFC database specifically stores real-time AFC entry and exit flow data, while the network basic parameter library stores static data such as network topology, station attributes, and section mileage. The Oracle / MySQL database is used to integrate and store dynamic and historical data such as train timetables and historical operating benchmarks. The databases work together through a data fusion interface to integrate scattered multi-source data into a structured, directly callable dataset, and output preprocessed basic data to the computing engine layer as needed.

[0126] The computational engine layer consists of a decoupling algorithm engine, a benchmark model library, and an anomaly recognition logic library. The decoupling algorithm engine is the core execution unit for decoupling the travel time components, responsible for accurately breaking down the total travel time of passengers into real-time values ​​of time components such as walking to and from the station, waiting on the platform, boarding the train, and walking for transfers. The benchmark model library stores historical benchmark values ​​and reasonable fluctuation thresholds for each time component, providing quantitative standards for anomaly judgment. The anomaly recognition logic library has built-in judgment rules and algorithms for various anomalies, analyzes the real-time values ​​of each time component separately, and realizes accurate identification, type labeling, and degree quantification of abnormal events, ultimately forming a structured anomaly judgment result dataset.

[0127] The application presentation layer includes an electronic map monitoring screen, an alarm list, and statistical reports. The electronic map monitoring screen binds abnormal events to the geographic information of the rail transit network, using color to indicate the degree of abnormality and unique icons to distinguish the type of abnormality, thus providing a visual and intuitive display of the location and status of abnormalities. The alarm list structures abnormal events by time, location, and type, supporting quick retrieval and detailed viewing. The statistical reports perform multi-dimensional statistical analysis on abnormal data, outputting regularized data such as the number of abnormalities, type distribution, and frequency of occurrence. These three sub-units work together to achieve multi-format display and refined management of abnormal information, supporting dispatchers in rapid response and accurate decision-making.

[0128] Thirdly, the present invention discloses an electronic device, which includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0129] Memory, used to store computer programs;

[0130] When the processor executes the program stored in the memory, it implements the steps of the rail transit abnormal state detection method disclosed in the first aspect above.

[0131] Specifically, the processor implements the first aspect of the disclosed rail transit anomaly detection method stored in the memory. The method involves first acquiring multi-source data such as AFC (Automatic Fare Collection) entry / exit data, network topology, and train timetables, and preprocessing this data. Then, it extracts the total passenger travel time from the preprocessed data and decouples it into real-time values ​​for walking into the station, walking out of the station, waiting on the platform, boarding the train, and walking for transfers. Finally, it analyzes the real-time values ​​of each time component separately, identifies and quantifies abnormal events, and forms an anomaly judgment result set. This solves the problem of existing AFC data analysis involving overlapping travel time components and the inability to accurately pinpoint the causes of increased travel time. Through refined decomposition and decoupling of the total travel time, it achieves accurate differentiation and attribution of anomalies in different dimensions such as train operation and passenger flow organization within stations, making the formulation of operation and scheduling strategies more targeted and providing accurate and specific decision-making basis for rail transit operation and maintenance scheduling.

[0132] Furthermore, in the existing technology, when passengers' travel time becomes longer, it is difficult to determine whether the problem is due to equipment failure (the train is running slowly) or management factors (transfer channels are changed to detours, security checks are upgraded), resulting in a lack of targeted scheduling strategies. This invention can accurately distinguish between train problems (speed limits) and station problems (detours / delays), providing a direct basis for operational scheduling decisions.

[0133] Furthermore, existing technologies make it difficult to quantify exactly how many trains passengers waited to board, and cannot accurately assess the severity of platform congestion. Comb distribution analysis, however, can specifically quantify how many trains passengers were delayed, reflecting the true passenger experience better than a simple percentage of crowding.

[0134] Furthermore, existing technologies lack automatic sensing methods for physical isolation measures of transfer channels (such as temporarily changing platform transfers to concourse transfers), and usually rely on manual reporting, resulting in a delayed response. The solution proposes a method to use AFC data to infer changes in physical transfer routes, filling the gap in the lack of digital supervision of temporary passenger flow organization measures (such as detours by barricades and changes in transfer channels).

[0135] Furthermore, without any additional hardware investment, second-level monitoring of the entire network can be achieved simply by leveraging the value of existing data.

[0136] Fourthly, the present invention discloses a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the rail transit abnormal state detection method disclosed in the first aspect.

[0137] Specifically, the storage medium stores the rail transit anomaly detection method provided in the first aspect. The method involves first acquiring multi-source data such as AFC (Automatic Fare Collection) entry / exit data, network topology, and train timetables, and preprocessing this data. Then, it extracts the total passenger travel time from the preprocessed data and decouples it into real-time values ​​for walking into the station, walking out of the station, waiting on the platform, boarding the train, and walking for transfers. Finally, it analyzes the real-time values ​​of each time component separately, identifies and quantifies abnormal events, and forms an anomaly judgment result set. This solves the problem of existing AFC data analysis involving overlapping travel time components and the inability to accurately pinpoint the causes of increased travel time. Through refined decomposition and decoupling of the total travel time, it achieves accurate differentiation and attribution of anomalies in different dimensions such as train operation and passenger flow organization within stations, making the formulation of operation and scheduling strategies more targeted and providing accurate and specific decision-making basis for rail transit operation and maintenance scheduling.

[0138] Furthermore, in the existing technology, when passengers' travel time becomes longer, it is difficult to determine whether the problem is due to equipment failure (the train is running slowly) or management factors (transfer channels are changed to detours, security checks are upgraded), resulting in a lack of targeted scheduling strategies. This invention can accurately distinguish between train problems (speed limits) and station problems (detours / delays), providing a direct basis for operational scheduling decisions.

[0139] Furthermore, existing technologies make it difficult to quantify exactly how many trains passengers waited to board, and cannot accurately assess the severity of platform congestion. Comb distribution analysis, however, can specifically quantify how many trains passengers were delayed, reflecting the true passenger experience better than a simple percentage of crowding.

[0140] Furthermore, existing technologies lack automatic sensing methods for physical isolation measures of transfer channels (such as temporarily changing platform transfers to concourse transfers), and usually rely on manual reporting, resulting in a delayed response. The solution proposes a method to use AFC data to infer changes in physical transfer routes, filling the gap in the lack of digital supervision of temporary passenger flow organization measures (such as detours by barricades and changes in transfer channels).

[0141] Furthermore, without any additional hardware investment, second-level monitoring of the entire network can be achieved simply by leveraging the value of existing data.

[0142] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0143] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0144] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0145] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0146] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0147] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. The illustrative expressions of the above terms in this specification should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0148] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0149] The above description describes specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting abnormal conditions in rail transit, characterized in that, include, Acquire multi-source data, preprocess the multi-source data to obtain a preprocessed dataset, which includes AFC entry and exit data, network topology data, and train timetable data. The total travel time of passengers is obtained from multi-source data. The total travel time is then decomposed and decoupled to obtain a set of real-time values ​​of travel component times. The set of real-time values ​​of travel component times includes real-time values ​​of walking time upon entering the station, walking time upon exiting the station, waiting time on the platform, boarding time, and walking time for transfers. Specifically, this includes defining each time component within the total travel time for any OD path. Using non-congested periods as the time benchmark, passengers with the shortest travel time along the same OD path are selected as unobstructed samples from AFC entry and exit data from multiple sources. Based on unobstructed samples, the baseline values ​​of each rigid physical movement time component and the overall rigid physical movement time baseline value are calculated to obtain the baseline values ​​of the component times. Based on the calculated rigid physical movement time reference value, combined with real-time AFC entry and exit data, the total real-time travel time of passengers is separated into components to achieve the separation of rigid components and elastic components, as well as the individual quantification of each subdivided rigid component, to obtain the real-time value of the travel component time. The system analyzes and processes each real-time value in the travel component time real-time value set to identify abnormal events, label these events and quantify their severity, resulting in an anomaly determination result dataset. Specifically, this includes... Based on the collected real-time transfer walking time, the interval travel time is calculated according to the station. The mean and variance of the interval travel time for all passengers are calculated. The interval travel time is compared with the historical benchmark. When the mean is larger than the historical benchmark and the variance is smaller than the historical benchmark, it is determined that the interval train operation efficiency is abnormal. Based on the real-time values ​​of waiting time on the platform collected, frequency domain analysis or histogram statistical analysis is performed on the real-time values ​​of waiting time on the platform of each station. Comb-like multi-peak characteristics are identified in the frequency domain analysis data or histogram, and the peak interval is approximately equal to the departure interval. This indicates that there is a vehicle platform congestion, and the average number of congestion cycles is calculated. Based on the collected real-time transfer walking time values, compared with historical baseline values, if the real-time transfer walking time value is greater than the historical baseline value, and cross-validation shows no abnormalities in the real-time transfer walking time value and platform waiting time value of the same OD path, it is determined that the transfer path has physically changed and the transfer path is abnormal. Extract passenger flow data for each section and match data between actual passenger routes and regular routes. If passenger flow in the downstream section is zero, or a large number of passengers exit at intermediate stations and then re-enter, it is determined to be an abnormal route execution. For the identified abnormal events, the abnormality type, location, and time of occurrence are marked, and the degree of abnormality is calculated to form an abnormality judgment result data set.

2. The method according to claim 1, characterized in that, The following includes: Based on the anomaly detection result dataset, the abnormal events are visualized on an electronic map to obtain a map with the abnormal events, and alarm information is output to the backend server in combination with the abnormal events.

3. The method according to claim 1, characterized in that, Acquire multi-source data, preprocess the multi-source data to obtain a preprocessed dataset, specifically including the following steps: AFC (Automatic Fare Collection) entry and exit data, network topology data, and train timetable data are collected from existing rail transit business systems, and preliminarily integrated according to preset field formats to establish a raw data pool and obtain multi-source data. AFC entry and exit data includes core fields such as card number, entry and exit time, and station ID; network topology data includes core fields such as station connection relationship, mileage between stations, and transfer station identification; and train timetable data includes core fields such as theoretical departure interval, first and last train time, and theoretical running time of each section. The data source data is integrated and processed for data format integration, verification and invalid data removal. Passenger OD path matching is completed based on the network topology data to obtain a preprocessed data set. The preprocessed dataset is categorized by OD path and time dimension and stored in a designated database.

4. The method according to claim 1, characterized in that, The total travel time of passengers is obtained from multi-source data. The total travel time is then broken down and decoupled to obtain a set of real-time values ​​of travel component times. The specific steps include: Extract passenger card swipe data from multiple sources, including origin station, destination station, and transfer station. Calculate the total travel time for any origin-destination (OD) path based on the multi-source data. The total travel time is combined with multi-source data for component decoupling processing to obtain the baseline value of the component time and the real-time value of the travel component time. The component time base value and the travel component time real-time value are classified according to OD path and time dimension to obtain the component time base value set and the travel component time real-time value set. The component time base value set and the travel component time real-time value set are stored in the designated database.

5. The method according to claim 1, characterized in that, Based on the calculated rigid physical movement time baseline value, combined with real-time AFC effective flow data, the total real-time travel time of passengers is component-by-component, realizing the separation of rigid components and elastic components, and the individual quantization of each subdivided rigid component, to obtain the real-time value of the travel component time. The specific steps include: By obtaining historical baseline values ​​and combining them with train operation data, including the actual travel time of trains in each section, the real-time value of transfer travel time can be obtained. Obtain historical baseline values, card swipe records, boarding and alighting stations, boarding and alighting times. Plan passenger routes based on boarding and alighting stations. First, determine whether the passenger is a transfer passenger. For transfer passengers, combine transfer card swipe records, boarding card swipe records, and train departure and stop times from AFC entry and exit data to calculate the real-time transfer walking time. Obtain historical baseline values ​​and real-time passenger flow density at passenger entry and exit stations. Use the baseline values ​​during off-peak hours and adjust according to passenger flow density during peak hours. Output the real-time values ​​according to the adjusted values ​​to obtain real-time values ​​of entry and exit walking time. The real-time values ​​of transfer walking time, station entry walking time, and station exit walking time are summed to obtain the rigid component sum. Based on the real-time passenger flow data changes in the multi-source data, the line congestion time is obtained. The real-time waiting time at the platform is calculated by combining the total duration with the rigid component sum.

6. A rail transit abnormal state detection system, characterized in that, The method for detecting abnormal conditions in rail transit as described in any one of claims 1-5.

7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in a memory, it implements the steps of the rail transit abnormality detection method according to any one of claims 1-5.

8. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the rail transit abnormal state detection method as described in any one of claims 1-5.

Images