A data management method and management system for rail transit

Abstract

The present application discloses a data management method for rail transit. The method includes: based on the information of each trip of each subway line, inputting according to the working attributes into the corresponding pre-trained dwell time prediction model to obtain the initial dwell time sequence corresponding to the station sequence of this trip; based on the next station that this trip is heading for, obtaining the expected number of boarding passengers at this station; based on the expected number of boarding passengers and the initial dwell time of this station, obtaining the target dwell time of this station according to the first algorithm, replacing the original initial dwell time, and updating the expected arrival times of all the remaining stations of this trip; according to the pre-set global dwell time management mechanism, dynamically updating the initial dwell times of all the remaining stations of this trip, and repeating steps S102 to S104 until the next station that this trip is heading for is the terminal station. Thus, it is possible to continuously optimize the dwell time strategy of railway trips and improve the railway operation level and passenger experience.

Description

Technical Field

[0001] The present invention relates to the technical field of urban rail transit, and particularly to a data management method and a management system for rail transit. Background Art

[0002] With the rapid increase in the urban population, the resulting traffic congestion problem in big cities has emerged. The original basic transportation infrastructure can no longer meet the current traffic demand. Currently, the main approach is to build underground railway facilities. Subways, due to their advantages such as large passenger-carrying capacity, fast and punctual operation, and green and low-carbon travel, can solve the problem of insufficient transport capacity of traditional public transportation and relieve traffic pressure at the same time. They have been vigorously promoted and constructed in many cities and are also the primary solution to solve the traffic congestion problem in domestic big cities. In the existing technology, the dwell time of the subway at each subway station is usually fixed. For some subway stations with low or high passenger flow, this is likely to result in unreasonable dwell time. How to adjust the dwell time of the subway at each subway station according to the ticket sales data of each subway station, so that the dwell time of the subway at each subway station is more reasonable, is the problem we need to solve.

[0003] The Chinese invention patent with the patent application number 202211366062.5 discloses a subway rail transit vehicle operation and maintenance monitoring system, which dynamically optimizes the stop time of the train according to the number of people entering and leaving the subway and the load capacity of the train, making the dwell time of the subway at each subway station more reasonable and improving the train operation efficiency and passenger-carrying capacity.

[0004] However, for the special situation where most stations on a certain subway line are crowded during a certain special period, when dynamically adjusting the dwell time of each station, the overall time plan is not considered whether it meets the expected plan requirements. Blindly adjusting the dwell time of each station according to the passenger flow information ignores the time management of the overall subway shift operation for all stations included in each subway line, that is, from the perspective of global time management. Although the passenger experience is improved, it is likely to cause subway shift delays or affect the normal operation handover of the next shift. In addition, the passenger flow data in the above patent solution is obtained based on the subway tickets sold within the statistical period. However, with the development of technology, today's subway rail transit has gradually tended to be self-service card swiping or code scanning to enter the subway waiting area, and there are few cases of paper ticket purchases or other ticket purchase methods that can pre-enter the starting point and destination. It is impossible to pre-know the destination directions of the ticket-purchasing passengers. When counting the passenger flow of a certain subway shift, due to the inability to identify the destination directions of the passengers who have swiped their cards, the passenger flow data statistics for the corresponding station entrance of this shift are not accurate, with certain limitations and affecting the adjustment accuracy. Summary of the Invention

[0005] The present application provides a data management method for rail transit, which can continuously optimize the dwell time strategy of train schedules, improve the railway operation level and passenger experience.

[0006] The present application provides a data management method for rail transit, including:

[0007] S101, based on the information of each trip of each subway line, where the trip information includes working attributes, subway line number, start and end time periods, and station sequence, input according to the working attributes into the corresponding pre-trained dwell time prediction model to obtain the initial dwell time sequence corresponding to the station sequence of this trip;

[0008] S102, based on the next station that this trip is heading for, obtain the predicted number of boarding passengers at this station;

[0009] S103, based on the predicted number of boarding passengers and the initial dwell time at this station, obtain the target dwell time at this station according to the first algorithm, replace the original initial dwell time, and update the predicted arrival times of all the remaining stations of this trip;

[0010] S104, according to the pre-set global dwell time management mechanism, dynamically update the initial dwell times of all the remaining stations of this trip to generate a new initial dwell time sequence, and repeat steps S102 to S104 until the next station that this trip is heading for is the terminal station.

[0011] Preferably, the working attributes are set as weekdays, weekends, and holidays. Each working attribute corresponds to a dwell time prediction model, and the dwell time prediction model for each working attribute is trained according to the ideal dwell times corresponding to a large number of different trip information in history.

[0012] Preferably, the predicted number of boarding passengers is set as: the number of passengers who have successfully purchased tickets within the time period formed by the departure of the previous trip from this station and the predicted arrival time of this trip at this station.

[0013] Preferably, the first algorithm is set as:

[0014]

[0015] where, T o is the target dwell time at this station, T initial is the initial dwell time at this station, N is the predicted number of boarding passengers at this station, N ideal is the ideal number of boarding passengers at this station under the initial dwell time, and k is the pre-set adjustment factor.

[0016] Preferably, the ideal number of boarding passengers at this station under the initial dwell time is calculated according to the following formula:

[0017] N ideal = v × T initial

[0018] Wherein, N ideal is the ideal passenger boarding volume, T initial is the initial residence duration, and v is the predicted passenger boarding speed of this station;

[0019] The method for obtaining the predicted passenger boarding speed of this station includes:

[0020] A1. Collect the passenger boarding video streams of each carriage of this train at this station in history based on the monitoring equipment in the rail transit;

[0021] A2. Use the image frame-by-frame analysis technology to obtain the successful passenger boarding volume of each carriage during the historical residence duration of this train at this station, so as to calculate the total historical passenger boarding volume of this train during the corresponding historical residence duration;

[0022] A3. Calculate the corresponding historical passenger boarding speed based on the historical residence duration and the total historical passenger boarding volume;

[0023] A4. Statistically analyze the large number of historical passenger boarding speeds of this train at this station, and take the average to obtain the predicted passenger boarding speed of this station.

[0024] Preferably, the method for obtaining the pre-trained residence duration prediction model includes:

[0025] B1. For each work attribute, collect a large amount of historical train trip information and its corresponding actual residence duration sequences under this work attribute for different subway lines;

[0026] B2. For each historical train trip information, obtain the monitoring video stream at each station;

[0027] B3. Based on the monitoring video stream of each station, intercept the video stream after the train closes the door, determine whether there are associated passengers. If there are, identify the associated passengers of this historical train trip information. Based on the actual passenger boarding volume and the actual residence duration of this station, obtain the corresponding passenger boarding speed, determine the upward adjustment amount as the ratio of the number of associated passengers to the passenger boarding speed, and determine the ideal residence duration of the corresponding station as the sum of the corresponding actual residence duration and the upward adjustment amount; if not, execute step B4;

[0028] B4. In the monitoring video stream of each station, intercept the video stream before the train closes the door, identify the duration during which the bound passengers exist, determine the downward adjustment amount as the difference between the actual residence duration and the duration during which the bound passengers exist, and determine the ideal residence duration of the corresponding station as the difference between the corresponding actual residence duration and the downward adjustment amount;

[0029] B5. Generate an ideal dwell time sequence for the historical shift information based on the ideal dwell times of all stations in the historical shift information;

[0030] B6. Perform label annotation on all historical shift information, and set the content of the label annotation as the corresponding ideal duration sequence to form a training sample set;

[0031] B7. Sort each sample in the training sample set in the time dimension, and equally divide it into a short-time effect sample set, a medium-time effect sample set, and a long-time effect sample set, and assign corresponding training weight factors A1, A2, and A3 respectively, where A1 < A2 < A3 and A1 + A2 + A3 = 1;

[0032] B8. Use the training sample set to train a pre-determined neural network structure, perform training according to the training weight factors of different training samples during the training process, and optimize the model parameters to obtain the final dwell time prediction model;

[0033] Among them, the associated passengers are determined as: passengers in the video stream after the train closes the door, whose moving speed exceeds a preset speed threshold and is towards the waiting area of this train; the bound passengers are determined as: passengers in the video stream before the train closes the door, who are towards the waiting area of this train.

[0034] Preferably, S101 further includes:

[0035] Based on the shift information of each trip on each subway line, input it into the corresponding pre-trained dwell time prediction model according to its working attributes, and determine the initial dwell time sequence corresponding to the station sequence of this trip as the output ideal dwell time sequence;

[0036] If the sum of all elements in the initial dwell time sequence of this trip is greater than a preset standard value, then adaptively adjust the initial dwell time of each station according to the following formula to update the initial dwell time sequence:

[0037] The new initial dwell time of this station = (the original initial dwell time of this station / the sum of the original initial dwell times of all stations) × standard value.

[0038] Preferably, the global dwell time management mechanism specifically includes:

[0039] S201. Obtain the maximum available dwell time for all remaining stations of this shift, and determine whether to trigger the stop dynamic adjustment instruction. If so, return to step S103 to switch the target dwell time determined for the next station that this shift is heading to to the original initial dwell time. Otherwise, execute step S202. Among them, the maximum available dwell time is set as the sum of the standard value and the actual dwell times of all stations that this shift has left. The trigger for the stop dynamic adjustment instruction is set as the maximum available dwell time being less than or equal to 0.

[0040] S202. According to the maximum available dwell time, proportionally adjust the initial dwell time of each remaining station of this shift to obtain its new initial dwell time. Specifically, calculate the new initial dwell time of each remaining station of this shift according to the following formula:

[0041]

[0042] Among them, T initial ' is the new initial dwell time of each remaining station of this shift, T initial is the original initial dwell time of each remaining station of this shift, Ts is the sum of the original initial dwell times of all remaining stations, and T m is the maximum available dwell time of all remaining stations of this shift;

[0043] S203. Update the estimated arrival times of all remaining stations of this shift according to the new initial dwell time.

[0044] Preferably, after step S104, the method further includes:

[0045] S105. When the current shift of this subway line ends, generate the shift information and its corresponding actual dwell time sequence, and add this data to step B1 to update the dwell time prediction model corresponding to this work attribute.

[0046] This application also proposes a data management system for rail transit. The system includes: a data acquisition module, a data adjustment module, and a data update module;

[0047] The data acquisition module is used to obtain the information of each shift of each subway line. The shift information includes work attributes, subway line numbers, start and end time periods, and station sequences;

[0048] The data adjustment module is used to input according to the work attribute into the corresponding pre-trained residence time prediction model to obtain the initial residence time sequence corresponding to the station sequence of this shift; based on the next station that this shift is heading for, obtain the expected number of passengers boarding at this station; based on the expected number of passengers boarding and the initial residence time of this station, obtain the target residence time of this station according to the first algorithm, replace the original initial residence time, and update the expected arrival times of all the remaining stations of this shift.

[0049] The data update module is used to dynamically update the initial residence times of all the remaining stations of this shift according to the pre-set global residence time management mechanism, generate a new initial residence time sequence, and repeatedly execute the content of the data adjustment module until the next station that this shift is heading for is the terminal station.

[0050] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0051] By constructing a residence time prediction model by differentiating different work attributes, making a preliminary adjustment to the traditional fixed residence time based on historical regular data to obtain the differentiated preliminary residence times of each station of the corresponding shift, and further, in order to overcome the variability of the rail transit passenger flow in the actual scenario, making a secondary adaptive adjustment to the initial residence time based on the real-time monitored expected number of passengers boarding and the expected boarding speed based on historical rules, it can effectively solve problems such as unreasonable traditional residence time adjustment and lack of scientific basis, differentiate the ideal residence times of different stations under different time characteristics, provide a reasonable and accurate residence time allocation for the operation route of the corresponding shift, and improve the operation efficiency of the subway rail transit and the passenger experience.

[0052] By adjusting the initial residence times of the remaining stations in proportion, it is ensured that the shift can end the operation according to the expected planned time period, avoiding operation delays caused by improper adjustment of the residence time; the proportional adjustment mechanism takes into account the original residence time ratio of each station, making the allocation of the residence time more reasonable and reducing the dissatisfaction of passengers caused by too long or too short waiting time; by introducing the global residence time management mechanism, the management of the residence time of the shift is made more refined and scientific, which helps to improve the overall operation level of the subway rail transit. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flowchart of the data management method for rail transit according to an embodiment of the present invention;

[0054] Figure 2 is a structural block diagram of the data management system for rail transit according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] To facilitate the understanding of the present invention, the present application will be described more comprehensively with reference to the relevant drawings; the preferred embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0056] It should be noted that the terms "vertical", "horizontal", "upper", "lower", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only embodiments.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs; the terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0058] Embodiment 1: Figure 1 It is a schematic flowchart of a data management method for rail transit according to an embodiment of the present invention.

[0059] As Figure 1 shown, a data management method for rail transit includes the following steps:

[0060] S101, based on the information of each train trip on each subway line, where the trip information includes work attribute, subway line number, start and end time periods, and station sequence, input according to the work attribute into the corresponding pre-trained residence time prediction model to obtain the initial residence time sequence corresponding to the station sequence of the trip, that is, each station corresponds to an initial residence time.

[0061] Among them, the work attribute is set to weekdays, weekends, and holidays.

[0062] For example, for a certain subway line S1, there are a number of fixed train trips every day. If the day is Monday and the work attribute is a weekday, the start and end time periods of train trip 1 are from 8:00 to 8:30, and it successfully runs from the starting station to the terminal station in half an hour and passes through stations A, B, C, and D in sequence according to the start and end order. The trip information includes: [weekday, subway line S1, 8:00 - 8:30, (A, B, C, D)].

[0063] Specifically, each work attribute corresponds to a residence duration prediction model, and the residence duration prediction model for each work attribute is trained based on the ideal residence durations corresponding to a large number of different shifts in history. It should be noted that for each shift, the start and end times and the train running speed are fixed. Since the passenger flow conditions at different stations are different, the residence duration at each station can be adaptively adjusted. However, considering the rigor of the overall operation time, it is necessary to improve the passenger experience at the station as much as possible under the condition of not affecting the overall operation of the shift. Therefore, the sum of the residence durations of all stations in each shift needs to be maintained within a preset standard value (i.e., the fixed total residence duration originally set for this shift). This standard value can be set according to the experience of railway management personnel or experts, that is, the sum of the residence durations needs to be less than or equal to the standard value, so as to avoid affecting the rail transit operation regulations due to the disorder of shift information and prevent the shift from exceeding the predetermined operation time due to excessive adjustment of the residence duration.

[0064] S102, based on the next station that this shift is heading to, obtain the predicted number of boarding passengers at this station. The predicted number of boarding passengers is set as: the number of passengers who have successfully purchased tickets (swiped the code or card successfully) within the time period formed by the departure of the previous shift from this station and the predicted arrival time of this shift at this station.

[0065] It should be noted that the number of boarding passengers can be obtained in real time through the ticket management system.

[0066] S103, based on the predicted number of boarding passengers and the initial residence duration of this station, obtain the target residence duration of this station according to the first algorithm, replace the original initial residence duration, and update the predicted arrival times of all the remaining stations in this shift.

[0067] Specifically, the first algorithm is set as:

[0068]

[0069] Among them, T o is the target residence duration of this station, T inital is the initial residence duration of this station, N is the predicted number of boarding passengers at this station, N ideal is the ideal number of boarding passengers at this station under the initial residence duration, and k is a preset adjustment factor, which is set according to the actual situation to control the adjustment amplitude.

[0070] Specifically, the ideal number of boarding passengers at this station under the initial residence duration can be calculated according to the following formula:

[0071] N ideal = v × T initial

[0072] Among them, N idealis the ideal passenger volume, T initial is the initial dwell time, and v is the predicted passenger boarding speed of this station. The acquisition method of the predicted passenger boarding speed of this station includes:

[0073] A1. Based on the passenger boarding video streams of each carriage of this train at this station in history collected by the monitoring equipment in the rail transit;

[0074] A2. Using the image frame-by-frame analysis technology, obtain the successful passenger boarding volume of each carriage during the historical dwell time of this train, so as to calculate the total historical passenger boarding volume of this train during the corresponding historical dwell time;

[0075] Specifically, in the passenger boarding video stream, all passenger images in the boarding direction are identified through frame-by-frame analysis, and the passenger images in each boarding direction are continuously tracked. If they disappear, it means the passenger has successfully boarded, otherwise, the boarding fails. The statistics of the successful passenger boarding volume can be realized through existing mature image processing technologies, and the present invention will not elaborate and limit this.

[0076] A3. Based on the historical dwell time and the total historical passenger boarding volume, calculate the corresponding historical passenger boarding speed (total passenger boarding volume divided by historical dwell time);

[0077] A4. Statistically analyze the large number of historical passenger boarding speeds of this train at this station, and take the average to obtain the predicted passenger boarding speed of this station.

[0078] S104. According to the pre-set global dwell time management mechanism, dynamically update the initial dwell times of all remaining stations of this train, generate a new sequence of initial dwell times, and repeat steps S102 to S104 until the next station that this train is heading to is the terminal station.

[0079] The technical solutions in the embodiments of the present application above have at least the following technical effects or advantages:

[0080] By constructing a dwell time prediction model through differentiating different work attributes, based on historical regular data, the traditional fixed dwell time is initially adjusted to obtain the differentiated initial dwell times of each station of the corresponding train. Further, in order to overcome the variability of the rail transit passenger flow in the actual scenario, based on the predicted passenger volume monitored in real time and the predicted passenger boarding speed based on historical rules, the initial dwell time is secondarily adaptively adjusted, which can effectively solve problems such as unreasonable adjustment of the traditional dwell time and lack of scientific basis, differentiate the ideal dwell times of different stations under different time characteristics, provide a reasonable and accurate dwell time allocation for the operation route of the corresponding train, and improve the operation efficiency and passenger experience of the subway rail transit.

[0081] Embodiment 2: In Embodiment 1, the residence duration prediction model for each working attribute is trained based on a large number of ideal residence durations corresponding to different shifts in history. It can be understood that currently, most rail transit systems have a fixed residence duration at each station. For example, it is set to 1 minute. However, the fixed residence duration is not necessarily the ideal residence duration of each station at different time periods. At different stations during the operation of an entire shift, due to the differentiated passenger flow and the special nature of the stations (such as railway stations, transfer points, etc.), the corresponding ideal residence durations will inevitably vary. In order to preliminarily adjust the residence duration of each station for different shifts based on historical law big data and provide reliable data support for subsequent real-time monitoring and secondary adjustment, the training method of the residence duration prediction model is extremely crucial.

[0082] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0083] In some embodiments, the acquisition method of the pre-trained residence duration prediction model includes:

[0084] B1. For each working attribute, collect a large amount of historical shift information and its corresponding actual residence duration sequence of different subway lines under this working attribute.

[0085] B2. For each historical shift information, obtain the monitoring video stream at each station (the video stream within the time window before and after the train closes the door).

[0086] B3. Based on the monitoring video stream of each station, intercept the video stream after the train closes the door, determine whether there are associated passengers. If there are, identify the associated passengers of this historical shift information. Based on the actual boarding volume and actual residence duration of this station, obtain the corresponding boarding speed, determine the upward adjustment amount as the ratio of the number of associated passengers to the boarding speed, and determine the sum of the corresponding actual residence duration and the upward adjustment amount as the ideal residence duration of the corresponding station; if not, execute step B4.

[0087] Among them, the actual boarding volume of this station can refer to the specific implementation method of step A2, and the present invention will not elaborate on this.

[0088] Among them, the associated passengers are determined as:

[0089] Passengers whose moving speed exceeds a preset speed threshold (passengers who want to catch up with this train, and the speed threshold is set according to the actual situation to define the urgency of the passengers) and are moving towards the waiting area of this train (draw the movement trajectories of all passengers in the monitoring video stream to generate their movement trajectory curves, extend the tangent of the movement trajectory curve in the direction of their movement, and intersect with the straight line where this train is located) are determined as associated passengers.

[0090] B4. In the monitoring video stream of each station, intercept the video stream before the train closes its doors, identify the duration during which the bound passengers exist, determine the reduction amount as the difference between the actual residence duration and the duration of the bound passengers, and determine the ideal residence duration of the corresponding station as the difference between the corresponding actual residence duration and the reduction amount.

[0091] Among them, the bound passengers are determined as: passengers facing the waiting area direction of this train.

[0092] It should be noted that the moment when the train closes its doors can be obtained by collecting train operation data or by image recognition in the monitoring video of this station. The present invention will not elaborate and limit this.

[0093] B5. Based on the ideal residence durations of all stations in the historical shift information, generate an ideal residence duration sequence of this historical shift information.

[0094] B6. Perform label annotation on all historical shift information, set the content of the label annotation as the corresponding ideal duration sequence, and form a training sample set.

[0095] B7. Sort each sample in the training sample set in the time dimension, and equally divide it into a short-time effect sample set, a medium-time effect sample set, and a long-time effect sample set (that is, evenly divide it into three time dimension intervals according to the number of sorted samples, and each time dimension interval corresponds to an effect sample set), and assign corresponding training weight factors A1, A2, and A3 respectively. A1 < A2 < A3 and A1 + A2 + A3 = 1. The closer to the current time, the more likely it is to be divided into the long-time effect sample set, indicating that the judgment value of its sample is higher.

[0096] B8. Use the training sample set to train a pre-determined neural network structure, train according to the training weight factors of different training samples during the training process, and optimize the model parameters to obtain the final residence duration prediction model.

[0097] In some embodiments, in step S101, based on the information of each shift of each subway line, input it into the corresponding pre-trained residence duration prediction model according to its working attributes, and determine the output ideal residence duration sequence as the initial residence duration sequence corresponding to the station sequence of this shift.

[0098] It should be noted that step S101 further includes: if the sum of all elements in the initial residence duration sequence of this shift is greater than the standard value, adaptively adjust the initial residence duration of each station according to the following formula, and update the initial residence duration sequence:

[0099] The new initial residence duration of this station = (the original initial residence duration of this station / the sum of the original initial residence durations of all stations) × the standard value.

[0100] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages:

[0101] By combining a large amount of historical shift information, monitoring video stream analysis, and weight assignment in the time dimension, the boarding pattern of each shift at different stations and different times is mined. The dwell time prediction model can more accurately predict the ideal dwell time of different work attributes and different stations at different time periods, and reasonably preliminarily adjust the fixed dwell time; it can consider the dynamic change law of passenger flow and the special nature of the station, provide differential initial dwell time allocation for different stations of different shifts, and improve the operation efficiency and passenger experience of the subway rail transit.

[0102] Embodiment 3: In Embodiment 1, for each dynamic adjustment of the initial dwell time of the station, the overall dwell time needs to be globally allocated and adjusted through the global dwell time management mechanism. In this embodiment, the global dwell time management mechanism needs to be further defined.

[0103] Therefore, the embodiments of the present application are optimized to a certain extent on the basis of the above embodiments.

[0104] In some embodiments, the global dwell time management mechanism specifically includes:

[0105] S201, obtain the maximum available dwell time of all remaining stations of this shift (the remaining stations are set as the stations that have not performed step S103), and determine whether to trigger the stop dynamic adjustment instruction. If so, return to step S103 to switch the target dwell time determined for the next station that this shift is going to to the original initial dwell time; otherwise, execute step S202.

[0106] Among them, the maximum available dwell time is set as: the sum of the standard value and the actual dwell time of all stations that this shift has left; the trigger for the stop dynamic adjustment instruction is set as: the maximum available dwell time is less than or equal to 0.

[0107] S202, according to the maximum available dwell time, proportionally adjust the initial dwell time of each remaining station of this shift to obtain its new initial dwell time.

[0108] The adjustment principle is to ensure that the sum of the dwell times of all stations does not exceed the standard value, and at the same time, as much as possible, keep the dwell time of each station consistent with its original ratio, focusing on differentiating the importance of different stations in the overall situation.

[0109] Specifically, in step S202, the new initial dwell time of each remaining station of this shift is specifically calculated according to the following formula:

[0110]

[0111] Among them, T initial ' is the new initial dwell time for each remaining stop of this shift, and T initial is the original initial dwell time for each remaining stop of this shift. Ts is the sum of the original initial dwell times of all remaining stops, and T m is the maximum available dwell time for all remaining stops of this shift.

[0112] S203. Update the estimated arrival times of all remaining stops of this shift according to the new initial dwell time.

[0113] The technical solutions in the embodiments of the present application described above have at least the following technical effects or advantages:

[0114] By adjusting the initial dwell times of the remaining stops in equal proportion, it is ensured that the shift can end its operation according to the expected planned time period, avoiding operation delays caused by improper adjustment of the dwell time; the equal-proportion adjustment mechanism takes into account the original dwell time ratio of each stop, making the distribution of the dwell time more reasonable and reducing the dissatisfaction of passengers due to too long or too short waiting times; by introducing the global dwell time management mechanism, the management of the shift dwell time is made more refined and scientific, which helps to improve the overall operation level of the subway rail transit.

[0115] Embodiment 4: As the operating environment and passenger demands change, the dwell time prediction model needs to be able to continuously learn and optimize to adapt to the new operating conditions.

[0116] Therefore, the embodiments of the present application are optimized to a certain extent on the basis of the above embodiments.

[0117] In some embodiments, after step S104, the method further includes:

[0118] S105. When the current shift of this subway line ends its operation, generate the shift information and its corresponding actual dwell time sequence, and add this data to step B1 to update the dwell time prediction model corresponding to this work attribute.

[0119] Thus, through the dwell time prediction model, the ideal dwell time is obtained by correcting the currently generated shift information and its actual dwell time sequence based on the monitoring video stream, and the dwell time strategies for different shifts under different work attributes of the railway are continuously optimized by combining real-time data.

[0120] The technical solutions in the embodiments of the present application described above have at least the following technical effects or advantages:

[0121] By continuously feeding the actual operation data back into the model training process, the dwell time prediction model can continuously learn and optimize, more accurately reflecting the actual operation situation and passenger demands; the addition of the actual dwell time sequence enriches the training sample set, making the model more accurate in predicting the dwell time of future train trips; as the operation environment and passenger demands change, the model can timely adjust the prediction strategy to maintain a high prediction accuracy; more accurate dwell time prediction helps reduce the unreasonable waiting or departure time of trains, improving the overall operation efficiency of the subway rail transit.

[0122] Furthermore, an embodiment of the present invention also provides a data management system for rail transit.

[0123] Figure 2 It is a structural block diagram of the data management system for rail transit according to an embodiment of the present invention.

[0124] As Figure 2 shown, a data management system for rail transit includes: a data acquisition module, a data adjustment module, and a data update module.

[0125] Among them, the data acquisition module is used to obtain the information of each trip of each subway line, and the trip information includes working attributes, subway line number, start and end time periods, and station sequence; the data adjustment module is used to input the working attributes into the corresponding pre-trained dwell time prediction model to obtain the initial dwell time sequence corresponding to the station sequence of this trip; based on the next station that this trip is heading to, obtain the expected number of boarding passengers at this station; based on the expected number of boarding passengers and the initial dwell time of this station, obtain the target dwell time of this station according to the first algorithm, replace the original initial dwell time, and update the expected arrival time of all the remaining stations of this trip; the data update module is used to dynamically update the initial dwell time of all the remaining stations of this trip according to the pre-set global dwell time management mechanism, generate a new initial dwell time sequence, and repeat the content of the data adjustment module until the next station that this trip is heading to is the terminal station.

[0126] It should be noted that other specific implementation contents of the embodiments of the present invention can refer to the above-mentioned data management method for rail transit.

[0127] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A data management method for rail transit, characterized in that: include: S101, based on the information of each trip of each subway line, the trip information includes work attributes, subway line number, start and end time periods, and station sequence, and inputs the work attributes into the corresponding pre-trained dwell time prediction model to obtain an initial dwell time sequence corresponding to the station sequence of the trip; the method for obtaining the dwell time prediction model includes: For each work attribute, collect a large amount of historical schedule information of different subway lines and the corresponding actual stay time series, as well as the surveillance video stream of each station; Based on the surveillance video stream of each station, the video stream after the train door is closed is intercepted, and the associated passengers of the historical train information are identified. Based on the actual passenger volume and actual stay time of the station, the boarding speed is obtained, and the ratio of the number of associated passengers to the boarding speed is determined as the upward adjustment amount, and the actual stay time is adjusted to the ideal stay time of the corresponding station by using the upward adjustment amount; if there is no associated passenger, the video stream before the train door is closed is intercepted, and the maintenance time of the bound passenger is identified, and the difference between the actual stay time and the maintenance time of the bound passenger is determined as the downward adjustment amount, and the actual stay time is adjusted to the ideal stay time of the corresponding station by using the downward adjustment amount; the associated passengers are determined as: in the video stream after the train door is closed, the passenger's moving speed exceeds the preset speed threshold and is facing the train waiting area; the bound passengers are determined as: in the video stream before the train door is closed, the passenger is facing the train waiting area; Generate an ideal stay time sequence based on the ideal stay time of all stations in the historical shift information; use the ideal stay time sequence to label the historical shift information to form a training sample set, assign corresponding training weight factors to the training samples based on the time dimension, and perform training based on the training weight factors of different training samples; S102, based on the next stop to which the bus is heading, obtaining the estimated number of passengers boarding the stop; S103, based on the estimated number of passengers and the initial stay time at the stop, a target stay time at the stop is obtained according to the first algorithm, the target stay time at the stop is replaced by the original initial stay time, and the estimated arrival time at all remaining stops of the flight is updated; S104, dynamically update the initial dwell time of all remaining stops of the shift according to the pre-set global dwell time management mechanism, generate a new initial dwell time sequence, and repeat steps S102 to S104 until the next stop to which the shift is heading is the terminal.

2. The data management method for rail transit according to claim 1, characterized in that: The work attributes are set as weekdays, weekends, and holidays. Each work attribute corresponds to a stay duration prediction model. The stay duration prediction model for each work attribute is trained based on ideal stay durations corresponding to a large amount of historical information on different shifts.

3. The data management method for rail transit according to claim 1, characterized in that: The estimated number of passengers is set as: the number of passengers who successfully purchased tickets within the time period formed by the time when the previous bus left the station and the time when the current bus is expected to arrive at the station.

4. The data management method for rail transit according to claim 1, characterized in that: The first algorithm is set as: Among them, T o is the target stay time of the site, T initial is the initial stay time at the station, N is the expected number of passengers boarding the station, and N ideal is the ideal passenger volume at the station under the initial stay time, and k is the preset adjustment amplitude factor.

5. The data management method for rail transit according to claim 4, characterized in that: The ideal passenger volume of the station under the initial stay time is calculated according to the following formula: N ideal =v×T initial Among them, N ideal For ideal passenger load, T initial is the initial stay duration, v is the predicted passenger boarding speed at the station; The site's predicted pickup speed is obtained in the following ways: A1. The monitoring equipment in the rail transit collects the video stream of passengers boarding each carriage of the train at the station in history; A2. Utilize the image frame-by-frame analysis technology to obtain the number of passengers successfully boarded in each carriage during the historical length of time the train stays at the station, thereby calculating the total number of passengers in history during the corresponding historical length of time the train stays; A3. Based on the historical length of stay and the total number of passengers in history, calculate the corresponding historical passenger speed; A4. Count a large number of historical boarding speeds of the bus at the station and take the average to get the predicted boarding speed of the station.

6. The data management method for rail transit according to claim 5, characterized in that: The S101 further includes: based on the information of each trip of each subway line, inputting it into the corresponding pre-trained dwell time prediction model according to its working attributes, determining the output ideal dwell time sequence as the initial dwell time sequence corresponding to the station sequence of the trip; if the sum of all elements in the initial dwell time sequence of the trip is greater than the preset standard value, adaptively adjusting the initial dwell time of each station according to the following formula, and updating the initial dwell time sequence: The new initial stay time of the site = (the original initial stay time of the site / the sum of the original initial stay times of all sites) × standard value.

7. The data management method for rail transit according to claim 1, characterized in that: The global stay duration management mechanism specifically includes: S201, obtaining the maximum available stay time of all remaining stations of the shift, and determining whether to trigger the stop dynamic adjustment instruction. If so, returning to step S103 to switch the target stay time determined at the next station to which the shift is heading to to the original initial stay time, otherwise, executing step S202; wherein the maximum available stay time is set to: the sum of the standard value and the actual stay time of all stations that the shift has left; and the triggering of the stop dynamic adjustment instruction is set to: the maximum available stay time is less than or equal to 0; S202, according to the maximum available dwell time, the initial dwell time of each remaining stop of the shift is adjusted in proportion to obtain its new initial dwell time, and the new initial dwell time of each remaining stop of the shift is calculated according to the following formula: Among them, T initial ' is the new initial dwell time of each remaining stop of the shift, T initial is the original initial stay time of each remaining stop of the shift, Ts is the sum of the original initial stay time of all remaining stops, T m The maximum available dwell time at all remaining stops of the flight; S203: Update the estimated arrival time of all remaining stops of the flight according to the new initial stay duration.

8. The data management method for rail transit according to claim 1, characterized in that: After step S104, the method further includes: S105, when the subway line ends the current shift, generate the shift information and its corresponding actual stay time sequence, add the data to the training sample set, and update the stay time prediction model corresponding to the work attribute.

9. A data management system for rail transit, applied to a data management method for rail transit as claimed in any one of claims 1 to 8, characterized in that: The system includes: a data acquisition module, a data adjustment module, and a data update module; The data collection module is used to obtain the information of each trip of each subway line. The trip information includes working attributes, subway line number, start and end time periods, and station sequence; The data adjustment module is used to input the corresponding pre-trained dwell time prediction model according to the work attributes, and obtain the initial dwell time sequence corresponding to the station sequence of the shift; based on the next station to which the shift is heading, obtain the estimated number of passengers boarding the station; based on the estimated number of passengers boarding and the initial dwell time of the station, obtain the target dwell time of the station according to the first algorithm, replace the original initial dwell time, and update the estimated arrival time of all remaining stations of the shift; The data update module is used to dynamically update the initial stay duration of all remaining stops of the shift according to the pre-set global stay duration management mechanism, generate a new initial stay duration sequence, and repeatedly execute the contents of the data adjustment module until the next stop to which the shift is heading is the terminal.

Images