Public transport capacity monitoring and scheduling method and system

Abstract

The invention belongs to the technical field of public transport scheduling, and discloses a public transport capacity monitoring and scheduling method and system, and the method comprises the steps: collecting the real-time passenger flow volume, historical passenger flow volume and environmental factor data of a target bus line, and carrying out the standardization and abnormality detection; constructing a hybrid prediction model comprising a time sequence decomposition model and a random fluctuation model, and predicting the passenger flow in the future period; the vehicles are divided into fixed shifts, maneuvering shifts and standby shifts, and vehicle configuration is dynamically adjusted according to the predicted load factor; through multi-source data fusion and an intelligent prediction algorithm, accurate monitoring and dynamic scheduling of the public transport capacity are realized, the vehicle utilization rate and the service quality are improved, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of public transportation scheduling, and particularly relates to a method and system for monitoring and scheduling public transportation capacity. Background Art

[0002] Traditional public transportation scheduling methods mainly rely on static departure schedules and empirical capacity arrangements, lacking the ability to respond to real-time passenger flow changes.

[0003] Most existing bus capacity scheduling systems lack accurate passenger flow prediction capabilities. Traditional scheduling mainly relies on historical experience data or simple statistical models, and cannot effectively capture the impact of external factors such as weather changes, holidays, and large-scale events on passenger flow. As a result, the response is slow and the scheduling is untimely when passenger flow suddenly changes. Especially during morning and evening rush hours, special weather conditions, or large-scale events, there are often situations of insufficient or excessive vehicle configurations, resulting in resource waste or a decline in service quality. Vehicles mostly operate according to fixed schedules and it is difficult to make dynamic adjustments according to the actual passenger flow situation. This one-size-fits-all scheduling method is inefficient in urban environments with large passenger flow fluctuations. It can neither meet the passenger demand during peak hours nor cause resource waste during off-peak hours. All the above deficiencies lead to low efficiency in the allocation of public transportation resources, poor passenger experience, increased operating costs, and hinder the sustainable development of the urban public transportation system. Patent CN114091757B discloses a method for intelligent scheduling of buses and subways based on passenger flow analysis and prediction, which evenly divides the current day T into multiple time periods, and statistically analyzes the historical passenger flow data of public transportation for each time period of the current day T; inputs the historical passenger flow data of public transportation for each time period of the current day T into a prediction model to obtain the predicted passenger flow data of public transportation for each time period of the T+1 day; obtains the total predicted passenger flow N1 of public transportation at station N within the time period (t1, t2), and the time period (t1, t2) includes one or more of the above time periods; obtains the capacity N2 of station N within the time period (t1, t2); S5, performs capacity matching; increases or decreases the capacity of station N according to the p value; this method only relies on historical passenger flow data for prediction, ignoring the significant impact of environmental factors on passenger flow, resulting in limited prediction accuracy, especially large prediction deviations during weather changes, holidays, or special events.

[0004] Therefore, there is an urgent need for an intelligent method and system for monitoring and scheduling public transportation capacity that can achieve accurate passenger flow prediction, multi-source data fusion, and dynamic capacity scheduling. Summary of the Invention

[0005] The present invention provides a method and system for monitoring and scheduling public transportation capacity, which is used to solve the problems of lack of real-time response ability and accurate passenger flow prediction in traditional bus scheduling, and achieve the purpose of intelligent and dynamic allocation and scheduling of public transportation capacity resources.

[0006] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for monitoring and scheduling public transportation capacity, the method comprising: Step S1, standardize and perform anomaly detection on the collected data of the target bus line to generate preprocessed data, where the collected data includes the real-time passenger flow, historical passenger flow and environmental factor data of the target bus line; After standardizing and performing anomaly detection on the collected data, an operation dataset RT, a historical dataset HS and an environmental factor dataset EF are formed; the operation dataset RT records the real-time passenger flow, vehicle operation frequency and load factor; the historical dataset HS contains the passenger flow statistical data of the most recent 30 days; the environmental factor dataset EF contains weather data and holiday information; The standardization adopts the maximum-minimum normalization method, , is the original data, and are the maximum and minimum values of the data respectively, is the standardized data; The anomaly detection adopts the interval range method to eliminate the data outside the range, where is the data mean, is the data standard deviation; Step S2, based on the preprocessed data, use the constructed hybrid prediction model MM to predict the passenger flow in the future period of the target bus line to obtain the predicted passenger flow, where the hybrid prediction model MM includes a time series decomposition model TM and a stochastic volatility model RM; the time series decomposition model TM decomposes the passenger flow into a basic passenger flow, a periodic passenger flow and a trend passenger flow; the stochastic volatility model RM captures short-term stochastic fluctuations.

[0007] Step S3, divide the vehicles to be scheduled into fixed shifts FT, flexible shifts AT and standby shifts BT, and obtain the predicted load factor of the target bus line according to the predicted passenger flow.

[0008] Based on the predicted load factor, dynamically adjust the vehicle configuration by using the capacity scheduling rules.

[0009] Further, the expression of the time series decomposition model TM is: ; where, is the total passenger flow at the predicted time t; the basic passenger flow represents a fixed minimum passenger flow level, and its value is the mean of the lowest passenger flows in the same historical period; the periodic passenger flow represents the periodic cyclic passenger flow changes obtained by daily and weekly statistics: , where and are the daily cycle amplitude and phase respectively, and are the weekly cycle amplitude and phase respectively; Trend passenger flow represents the long-term change trend and is calculated by the moving average method; represents the short-term random fluctuation.

[0010] Furthermore, the random fluctuation model RM is a model established based on environmental factor data, and the environmental factor data includes: weather impact factor and holiday impact factor .

[0011] Weather impact factor takes the value of 1.0 for non-rainy and snowy weather, 0.8 for heavy rain or rainstorm, 0.7 for snowy days, and 1.2 for other periods; Holiday impact factor , takes the value of 1.0 for weekdays, 1.1 - 1.3 for ordinary weekends, and 1.3 - 1.5 for long holidays; The calculation formula of the random fluctuation model RM is: , where is the standard deviation of the passenger flow in the same historical period, is a random variable subject to the standard normal distribution.

[0012] Furthermore, the calculation formula for predicting the full load rate is: , where is the number of operating vehicles at time t, is the rated passenger capacity of a single vehicle.

[0013] Furthermore, the vehicles are divided into fixed shifts FT, flexible shifts AT, and standby shifts BT, and 60% of the fixed shifts FT and 40% of the flexible shifts AT are used for vehicle scheduling; The capacity dispatching is based on the following decision rules: When predicting the full load rate , reduce the number of flexible shifts AT, and the reduction ratio of flexible shifts AT is , but the maximum reduction rate of flexible shifts AT does not exceed 40%; When predicting the full load rate , maintain the current capacity configuration; When predicting the full load rate , increase the flexible shifts AT, and the increase ratio of flexible shifts AT is ; When predicting the full load rate , fully activate the standby shifts BT and dispatch vehicles from nearby lines for support.

[0014] Further, the evaluation of the hybrid prediction model MM is carried out through the mean absolute percentage error , root mean square error, and coefficient of determination for evaluation: Mean absolute percentage error: ; Root mean square error ; Coefficient of determination ; When or , trigger the model parameter adjustment mechanism, recalculate the base passenger flow and periodic passenger flow , and update the hybrid prediction model MM.

[0015] Further, the obtained predicted passenger flow is used for passenger flow prediction by a multi-time scale sliding window. The multi-time scale sliding window includes: a short-term prediction window, a medium-term prediction window, and a long-term prediction window. The short-term prediction window SW is used to predict the passenger flow changes in the next 30 - 60 minutes, with a sampling frequency of 5 minutes; the medium-term prediction window MW is used to predict the passenger flow changes in the next 3 - 6 hours, with a sampling frequency of 15 minutes; the long-term prediction window LW is used to predict the passenger flow changes in the next 24 hours, with a sampling frequency of 30 minutes.

[0016] In a second aspect, the present invention provides a public transportation capacity monitoring and dispatching system for implementing the public transportation capacity monitoring and dispatching method in the first aspect. The system includes: a data acquisition module, a passenger flow prediction module, and a capacity dispatching module.

[0017] The data acquisition module is used to collect real-time passenger flow, historical passenger flow, and environmental factor data of the target bus line; used to standardize and detect anomalies in the collected data; the passenger flow prediction module is used to establish a hybrid prediction model MM to predict the passenger flow in different future time periods; the capacity dispatching module is used to dynamically adjust vehicle configuration according to the predicted passenger flow.

[0018] Further, the data acquisition module includes: an on-vehicle passenger flow statistics unit, which collects boarding and alighting passenger flow data through an infrared sensor or a weight sensor; a station passenger flow monitoring unit, which monitors the number of waiting passengers at the station through a camera and an intelligent video analysis system; a vehicle GPS positioning unit, which collects real-time vehicle position, speed, and operation status information; an environmental information acquisition unit, which obtains weather information; and the units are all connected in real time to the data processing system of the data acquisition module through a wireless communication network.

[0019] Further, the system further includes a visualization display module and an emergency handling module; The visualization display module is used to display the line operation status, vehicle distribution, heat map of passenger flow density, and full load rate warning information in real time; When the predicted full load rate exceeds 95%, the emergency handling module automatically triggers an emergency plan, including: temporarily adding shuttle buses, adjusting the departure interval, and linking with surrounding bus lines. At the same time, it pushes congestion tips and alternative travel suggestions to passengers through the public information platform.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: By establishing a hybrid prediction model and a multi-level scheduling strategy, the present invention realizes accurate prediction of public transportation passenger flow and intelligent scheduling of transport capacity, can significantly increase the vehicle utilization rate by 15 - 20%, reduce the passenger waiting time by more than 30%, and at the same time reduce the operation cost by about 10 - 15%, improving the reliability and resilience of the public transportation system and providing an intelligent solution for urban traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of a public transportation transport capacity monitoring and scheduling method of the present invention; Figure 2 is a schematic diagram of the composition of a public transportation transport capacity monitoring and scheduling system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be described clearly and completely below. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0023] Embodiment 1: As Figure 1 shown, it is a flowchart of a public transportation transport capacity monitoring and scheduling method of the present invention. From a program perspective, the execution subject of the process can be a program running on an application server or an application terminal. It can be understood that this method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities.

[0024] The method includes: Step S1, standardize and perform anomaly detection on the collected data of the target bus line to generate preprocessed data, where the collected data includes the real-time passenger flow, historical passenger flow, and environmental factor data of the target bus line.

[0025] After the collected data are standardized and anomaly detected, they are composed of an operation data set RT, a historical data set HS and an environmental factor data set EF; the operation data set RT records real-time passenger flow, vehicle operation frequency and full load rate; the historical data set HS contains passenger flow statistics for the past 30 days; the environmental factor data set EF contains weather data and holiday information.

[0026] In this embodiment, data collection adopts a multi-source heterogeneous data fusion strategy to ensure the comprehensiveness and accuracy of the data; real-time passenger flow data is collected through both on-board sensors and site cameras and summarized every 5 minutes; historical passenger flow data includes a 30-day rolling data window, organized in three time dimensions: hour, day, and week; environmental factor data includes real-time weather conditions, weather forecast for the next 48 hours, and pre-coded holiday information.

[0027] Taking the No. 301 bus line in a certain city as an example, the system records the number of people getting on and off at each stop every 5 minutes during the morning rush hour of 6:30-8:30 on weekdays, and compares the data of the same period in the previous 4 weeks. Combined with the weather conditions of the day (such as light rain), it calculates abnormal values ​​in real time and applies replacement strategies. On the eve of a large shopping festival, the system successfully identified passenger flow growth that was higher than the normal fluctuation range by comparing historical data for the same period, avoiding misjudging this reasonable growth as abnormal data.

[0028] The standardization adopts the maximum and minimum value normalization method. , is the original data, and are the maximum and minimum values ​​of the data, respectively. is the standardized data.

[0029] Standardization ensures that data of different dimensions can be effectively compared and integrated. In practical applications, passenger flow often fluctuates in the range of 0-2000 people / hour. By normalizing the maximum and minimum values, it is converted into a standard value between 0 and 1. For example, the historical maximum passenger flow of a station is 1800 people / hour and the minimum is 50 people / hour. When 1000 people / hour is currently detected, the standardized value is (1000-50) / (1800-50)≈0.54.

[0030] The anomaly detection adopts the interval range method to eliminate the abnormal Range data, where is the data mean, It is the standard deviation of the data. This range can cover approximately 95% of the normal data, effectively identifying outliers without overly excluding fluctuating data. During implementation, outliers are not directly discarded but are replaced by the average value of the adjacent time periods to maintain data continuity. For continuous outliers, the system will trigger an alarm and require manual review to prevent misjudging normal large-scale passenger flow changes (such as large events) as abnormal data.

[0031] The outlier detection part uses a sliding window mechanism to implement dynamic threshold setting. The system automatically updates the mean μ and standard deviation σ of each time period every week to adapt to seasonal changes. For example, during the summer vacation, the average passenger flow of a certain route on weekdays drops to 800 people per hour, and the standard deviation is 100 people per hour. Then the outlier threshold range is [600, 1000]. When a passenger flow of 1100 people per hour is detected, the system does not directly exclude it but records it as an outlier and temporarily replaces it with the average value of 850 people per hour at the same time period in the previous three days, and at the same time triggers a manual review notice.

[0032] Step S2, based on the preprocessed data, use the constructed hybrid prediction model MM to predict the passenger flow in the future time period of the target bus route, and obtain the predicted passenger flow. The hybrid prediction model MM includes a time series decomposition model TM and a stochastic volatility model RM; the time series decomposition model TM decomposes the passenger flow into a basic passenger flow, a periodic passenger flow, and a trend passenger flow; the stochastic volatility model RM captures short-term stochastic fluctuations.

[0033] The core advantage of the hybrid prediction model MM is to separate and process the deterministic component and the stochastic component. The time series decomposition model TM is responsible for capturing regular changes, while the stochastic volatility model RM processes short-term fluctuations. In practical applications, the system determines the daily cycle amplitude Ad and phase φd, as well as the weekly cycle amplitude Aw and phase φw through Fourier analysis. Taking a bus route on a main road in a certain city as an example, through the analysis of three months of data, it is obtained that: the daily cycle amplitude Ad = 400 (people per hour), the phase φd = π / 2 (corresponding to the morning rush hour), the weekly cycle amplitude Aw = 200 (people per hour), and the phase φw = 0 (the passenger flow is the highest on Monday). This means that the passenger flow during the morning rush hour on weekdays will be about 400 people per hour higher than the daily average, and the overall passenger flow on Monday is about 200 people per hour higher than on weekends. The system automatically updates these parameters every month to ensure that the model always reflects the latest passenger flow change rules.

[0034] The expression of the time series decomposition model TM is: ; where is the total passenger flow at the predicted time t; the basic passenger flow represents the fixed minimum passenger flow level, and its value is the mean of the lowest passenger flow volume in the same historical period; the periodic passenger flow Indicates the cyclic passenger flow changes obtained by daily and weekly statistics: ,in and are the daily cycle amplitude and phase, and are the cycle amplitude and phase respectively.

[0035] Trending passenger flow Indicates the long-term trend, calculated by the moving average method; Indicates short-term random fluctuations.

[0036] The random fluctuation model RM is a model established based on environmental factor data, and the environmental factor data includes: weather influencing factors and holiday impact factors , weather factors The value is 1.0 for non-rainy or snowy weather, 0.8 for heavy rain or rainstorm, 0.7 for snowy days, and 1.2 for other periods; Holiday impact factor , the value for weekdays is 1.0, the value for ordinary weekends is 1.1-1.3, and the value for long holidays is 1.3-1.5; the formula for calculating random fluctuations is: ,in, is the standard deviation of the historical passenger flow in the same period, is a random variable that follows a standard normal distribution.

[0037] The setting of the weather impact factor Wf is based on the comparison of historical passenger flow during the same period: in heavy rain or rainstorm weather, the passenger flow decreases by an average of 20% (so the value is 0.8); in snowy days, the passenger flow decreases more, averaging 30% (so the value is 0.7); and in light rainy or light haze days, the public transportation passenger flow increases by about 20% (so the value is 1.2), which is mainly because private car travel is restricted and passengers turn to public transportation. The holiday impact factor Ef is finely divided according to the type of holiday: the value range of ordinary weekends is 1.1-1.3, of which the upper limit of commercial area lines is 1.3 and the lower limit of residential area lines is 1.1; during long holidays, the scenic area lines take a value of 1.5, the commercial area lines take a value of 1.4, and the residential area lines take a value of 1.3. For example, during the May Day holiday, the passenger flow of the bus line leading to the Happy Valley scenic area is expected to increase by 50% compared with weekdays.

[0038] Other periods mainly refer to special weather conditions such as light rain, fog and haze. At this time, passenger flow will increase by about 20%, mainly because private car travel has decreased and people have chosen public transportation instead. The range setting of holiday impact factors allows fine-tuning according to specific holiday types and local characteristics, such as shopping festivals and traditional holidays. The random variable ε introduces randomness based on statistical characteristics, enabling the model to simulate uncertainty in the real world.

[0039] Step S3: Divide the vehicles to be scheduled into fixed shifts FT, flexible shifts AT, and standby shifts BT, and obtain the predicted load factor of the target bus route according to the predicted passenger flow volume.

[0040] Based on the predicted load factor, dynamically adjust the vehicle configuration by using the vehicle capacity scheduling rules.

[0041] Predicted load factor The calculation formula is: , where is the number of operating vehicles at time t, is the rated passenger capacity of a single vehicle.

[0042] Vehicles are divided into fixed shifts FT, flexible shifts AT, and standby shifts BT. In the public transportation vehicle capacity scheduling system, vehicles are divided into three types of shifts.

[0043] Fixed shifts FT are the basic vehicles operating according to a fixed schedule and frequency, accounting for 60% of the total vehicles, ensuring the stability of the basic service level of the route and not being adjusted due to short-term passenger flow fluctuations. Flexible shifts AT are vehicles that can be flexibly adjusted according to passenger flow predictions, accounting for about 40%, and the increase or decrease ratio is based on a specific calculation formula. Standby shifts BT are additional vehicles enabled in special or emergency situations, serving as the last line of defense for the bus system to cope with passenger flow peaks or emergencies. This three-level vehicle classification scheme can flexibly respond to passenger flow changes while maintaining service stability, improving vehicle utilization rate and reducing operating costs.

[0044] Use 60% fixed shifts FT and 40% flexible shifts AT for vehicle scheduling; the ratio of 60% fixed shifts to 40% flexible shifts is the optimal balance point obtained through simulation tests of multiple routes and various passenger flow patterns, ensuring both service stability and sufficient scheduling flexibility.

[0045] The setting of the load factor threshold takes into account the balance between passenger comfort and operating efficiency. 60% as the lower limit is determined based on the study of carriage comfort, ensuring that passengers have enough personal space; 80% as the upper adjustment threshold is considered for the reasonable utilization of vehicle capacity during peak periods; 90% as the emergency threshold is to avoid a sudden drop in service quality and an increase in safety risks caused by extreme congestion.

[0046] The vehicle capacity scheduling is based on the following decision rules: When the predicted load factor is, reduce the number of flexible shifts AT, and the reduction ratio is , but the maximum reduction rate does not exceed 40%; When the predicted load factor is, maintain the current vehicle capacity configuration; When the predicted load factor When increasing the flexible shift AT, the increase ratio is ; When the predicted full load rate is reached, all standby shifts BT are fully activated, and vehicles on nearby routes are dispatched for support at the same time.

[0047] The design of the dispatching ratio calculation formula follows the progressive principle. For example, when the predicted full load rate is 85%, the increase ratio of the flexible shift is (85% - 80%) / 0.1 = 50%, that is, half of the current flexible shift is increased. This progressive adjustment avoids overreaction and ensures that resource input matches demand growth.

[0048] The prediction model is evaluated using the mean absolute percentage error , root mean square error, and coefficient of determination: Mean absolute percentage error: ; Root mean square error ; Coefficient of determination ; When or is reached, the model parameter adjustment mechanism is triggered to recalculate the base passenger flow and periodic passenger flow .

[0049] The obtained predicted passenger flow is used for passenger flow prediction by a multi - time - scale sliding window. The multi - time - scale sliding window includes: a short - term prediction window, a medium - term prediction window, and a long - term prediction window. The short - term prediction window SW is used to predict the passenger flow changes in the next 30 - 60 minutes, with a sampling frequency of 5 minutes; the medium - term prediction window MW is used to predict the passenger flow changes in the next 3 - 6 hours, with a sampling frequency of 15 minutes; the long - term prediction window LW is used to predict the passenger flow changes in the next 24 hours, with a sampling frequency of 30 minutes.

[0050] The short - term prediction window SW is mainly used for real - time dispatching decisions. The 5 - minute sampling frequency can capture passenger flow fluctuations in a timely manner; the medium - term prediction window MW supports shift arrangement adjustment and resource allocation, and the 15 - minute sampling frequency balances prediction accuracy and computational efficiency; the long - term prediction window LW is used for strategic planning and resource preparation, and the 30 - minute sampling frequency is sufficient to reflect the passenger flow change trend within a day.

[0051] The multi - time - scale strategy enables the system to simultaneously meet decision - making requirements for different time ranges. The short - term prediction window SW (5 - minute sampling frequency) is mainly used for real - time dispatching decisions, such as temporarily adding vehicles; the medium - term prediction window MW (15 - minute sampling frequency) guides the shift arrangement adjustment of on - duty personnel and vehicles; the long - term prediction window LW (30 - minute sampling frequency) is used for vehicle allocation and personnel arrangement planning for the next day.

[0052] The characteristic variables and computing resources used by each prediction window are also different: the short-term window SW mainly relies on the real-time data of the last 30 minutes and a simplified prediction model, and the computing delay is controlled within 10 seconds; the medium-term window MW combines the cumulative data of the current day and the historical data of the same period, and the model complexity is moderate; the long-term window LW makes full use of all available information such as historical data, weather forecasts, and event calendars, and adopts a complete hybrid prediction model.

[0053] When conducting passenger flow prediction based on a multi-time scale sliding window, the characteristic variables and computing resources used by each prediction window are different. The long-term prediction window adopts a complete hybrid prediction model, while the short-term and medium-term windows use a simplified version of the hybrid prediction model to balance prediction accuracy and computing efficiency.

[0054] The short-term prediction window SW mainly relies on the real-time data of the last 30 minutes and a simplified prediction model, and the computing delay is controlled within 10 seconds. The expression of the hybrid prediction model for the short-term prediction window is: ; where retains the original definition of the base passenger flow, representing the fixed minimum passenger flow level, and its value is the mean of the minimum passenger flow volumes in the same historical period; only considers the periodic pattern of the last 30 minutes, which is simplified to: ; is the real-time random fluctuation, which is simplified for calculation as: ; where is the standard deviation of the passenger flow in the last 30 minutes, is the weather impact factor, is a random variable subject to the standard normal distribution.

[0055] The medium-term prediction window MW combines the cumulative data of the current day and the historical data of the same period, and the model complexity is moderate. The expression of the hybrid prediction model for the medium-term prediction window is: ; where retains the original definition of the base passenger flow; considers the periodic pattern of the current day: ; is the simplified trend passenger flow, only considering the moving average of the last 7 days; is the medium-term random fluctuation: ; where is the standard deviation of the passenger flow at the same time of the current day, is the weather impact factor, is the holiday impact factor, is a random variable subject to the standard normal distribution.

[0056] The design of this multi-time-scale hybrid prediction model enables the system to simultaneously meet the decision-making requirements for different time ranges. The short-term prediction window is mainly used for real-time scheduling decisions, the medium-term prediction window guides the shift scheduling adjustment of on-duty personnel and vehicles, and the long-term prediction window is used for the vehicle allocation and personnel arrangement planning for the next day, thus realizing the comprehensive intelligent management of bus capacity.

[0057] This technology has been piloted on a main line in a certain city for 6 months and achieved remarkable results. The line is 32 kilometers long, with a total of 35 stations, and the one-way passenger flow during the peak period reaches 2,200 people per hour. After applying this method, the system performance is as follows: The accuracy of passenger flow prediction has been significantly improved, and the MAPE has dropped from the original 23.7% to 8.5%. Especially during bad weather and special events, the improvement in prediction accuracy is particularly obvious. For example, during a sudden heavy rain in February 2024, the system accurately predicted a 16.8% decrease in passenger flow and timely adjusted the capacity configuration, maintaining a vehicle utilization rate of over 85%.

[0058] The utilization efficiency of capacity resources has been greatly improved. The average full load rate has increased from the original 62% to 76%, and the crowded complaints during the peak period have decreased by 42%. By dynamically adjusting the flexible shifts, during the morning peak period from 7:00 to 8:00 on weekdays, the system added 5 additional motor vehicles at the peak of passenger flow, effectively alleviating congestion; while during the non-peak period from 10:00 to 15:00, the operation of 8 originally scheduled fixed-shift vehicles was reduced, reducing fuel consumption and vehicle wear.

[0059] The comprehensive economic benefits are significant. The average monthly operating cost of the pilot line has decreased by 14.3%, saving approximately 2,800 liters of fuel per month and reducing carbon emissions by approximately 6.5 tons. At the same time, the average waiting time of passengers has been shortened from the original 12 minutes to 8.5 minutes, and the service satisfaction score has increased from 75 points to 87 points. Especially during the National Day Golden Week in 2023, the system successfully coped with the challenge of a 35% increase in passenger flow compared with normal days through accurate prediction and early scheduling, maintaining good service quality and receiving unanimous praise from passengers and operators. This practical application case fully demonstrates the remarkable effect of the present invention in improving the utilization efficiency of public transportation resources, reducing operating costs, and enhancing service quality, providing a feasible technical solution for the intelligent management of urban public transportation.

[0060] Embodiment 2: As Figure 2 shown, the present invention provides a schematic diagram of the composition of a public transportation capacity monitoring and scheduling system for implementing the public transportation capacity monitoring and scheduling method in the first aspect. The system includes: a data collection module, a passenger flow prediction module, and a capacity scheduling module.

[0061] This system adopts a modular design, and each functional module is connected through a standardized interface, which is convenient for maintenance and upgrade. The data acquisition module, as the basis of the system, provides multi-source heterogeneous data; the passenger flow prediction module realizes the scientific prediction of future passenger flow; the transport capacity scheduling module is the decision execution layer, which converts the prediction results into specific vehicle scheduling instructions. The three main modules form a closed loop of data acquisition - analysis and prediction - decision execution to achieve the intelligent management of bus transport capacity.

[0062] The data acquisition module is used to collect real-time passenger flow, historical passenger flow and environmental factor data of the target bus line; to standardize and perform anomaly detection on the collected data; the passenger flow prediction module is used to establish a hybrid prediction model MM to predict the passenger flow in different future time periods; the transport capacity scheduling module is used to dynamically adjust vehicle configuration according to the predicted passenger flow.

[0063] The data acquisition module includes: an on-vehicle passenger flow statistics unit that collects boarding and alighting passenger flow data through an infrared sensor or a weight sensor; a station passenger flow monitoring unit that monitors the number of passengers waiting at the station through a camera and an intelligent video analysis system; a vehicle GPS positioning unit that collects real-time vehicle position, speed and operating status information; an environmental information acquisition unit that obtains weather information; all the units are connected to the data processing system of the data acquisition module in real time through a wireless communication network.

[0064] The data acquisition system adopts a distributed architecture, and each acquisition unit works independently and transmits data in real time through a wireless network. The on-vehicle passenger flow statistics unit uses dual verification of an infrared sensor and a weight sensor, and the accuracy rate can reach more than 95%; the station passenger flow monitoring unit uses video analysis technology to automatically identify the number of waiting passengers and preprocesses the data through edge computing to reduce the transmission burden; the vehicle GPS positioning unit collects position data once every 10 seconds and can accurately track the vehicle operation trajectory; the environmental information acquisition unit obtains the weather forecast and measured data provided by the meteorological department in real time through an API interface. All acquisition units adopt a breakpoint resumption and data caching mechanism to ensure data integrity during network fluctuations.

[0065] The system also includes a visualization display module and an emergency handling module; The visualization display module is used to display the line operation status, vehicle distribution, passenger flow density heat map and full load rate warning information in real time; When the predicted full load rate exceeds 95%, the emergency handling module automatically triggers an emergency plan, including: temporarily adding shuttle buses, adjusting the departure interval and linking with surrounding bus lines, and at the same time pushing congestion tips and alternative travel suggestions to passengers through a public information platform.

[0066] The visualization display module adopts an interactive dashboard design, supporting multi-dimensional data display and drill-down analysis. The display of the line operation status includes the real-time location, running speed, and delay of vehicles; the passenger flow density heat map intuitively shows the passenger flow distribution of each station and section; the overloading rate warning information is displayed in a hierarchical manner. The sections with a rate below 60% are shown in green, 60%-80% in yellow, 80%-90% in orange, and above 90% in red, facilitating dispatchers to quickly identify the areas that need attention. The emergency handling module designs a multi-level response mechanism. The 95% overloading rate trigger point is the critical value determined based on the research of passenger comfort. When this value is exceeded, the passenger experience will deteriorate significantly. In the emergency plan, additional shuttle buses are preferentially arranged for the most crowded sections; the adjustment range of the departure interval is dynamically calculated according to the excess value of the overloading rate; the alternative routes are determined through road network analysis for the associated surrounding lines. The passenger information push system is connected to mainstream maps and travel APPs to ensure that information reaches users in a timely manner.

[0067] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A public transportation capacity monitoring and dispatching method, characterized in that, The method includes: Step S1: Standardize and perform anomaly detection on the collected data of the target bus line to generate preprocessed data, where the collected data includes the real-time passenger flow, historical passenger flow, and environmental factor data of the target bus line; Step S2: Based on the preprocessed data, use the constructed hybrid prediction model MM to predict the passenger flow in the future period of the target bus line to obtain the predicted passenger flow. The hybrid prediction model MM includes a time series decomposition model TM and a stochastic volatility model RM; the time series decomposition model TM decomposes the passenger flow into a basic passenger flow, a periodic passenger flow, and a trend passenger flow; the stochastic volatility model RM captures short-term stochastic fluctuations; Step S3: Divide the vehicles to be scheduled into fixed shifts FT, flexible shifts AT, and standby shifts BT, and obtain the predicted load factor of the target bus line according to the predicted passenger flow; Based on the predicted load factor, dynamically adjust the vehicle configuration using the vehicle capacity scheduling rules.

2. The public transport capacity monitoring and dispatching method according to claim 1, wherein The expression of the time series decomposition model TM is as follows: ; where is the total passenger flow at the predicted time t; the basic passenger flow represents the fixed minimum passenger flow level, and its value is the average of the minimum passenger flow volumes in the same historical period; the periodic passenger flow represents the periodic cyclic passenger flow changes obtained by daily and weekly statistics: , where and are the daily cycle amplitude and phase respectively, and are the weekly cycle amplitude and phase respectively; Trend passenger flow Indicates the long-term change trend, calculated by the moving average method; Indicates short-term random fluctuations.

3. The public transport capacity monitoring and dispatching method according to claim 2, wherein, The random fluctuation model RM is a model established based on environmental factor data, and the environmental factor data includes: weather impact factors and holiday impact factors . The calculation formula of the random fluctuation model RM is: , where is the standard deviation of passenger flow in the same historical period, is a random variable subject to the standard normal distribution.

4. The public transport capacity monitoring and dispatching method according to claim 2, characterized in that Predicted full load rate The calculation formula is as follows: , where is the number of operating vehicles at time t, is the rated passenger capacity per vehicle.

5. The public transportation capacity monitoring and dispatching method according to claim 4, wherein 60% of the fixed shifts FT and 40% of the flexible shifts AT are used for vehicle scheduling of the vehicles to be scheduled; the vehicle capacity scheduling rules include: When the predicted full load rate is reached, reduce the number of flexible shifts AT, and the reduction ratio of flexible shifts AT is , but the maximum reduction rate of flexible shifts AT does not exceed 40%; When predicting the full load rate maintain the current transport capacity configuration; When predicting the full load rate increase the flexible shift AT, and the proportion of increasing the flexible shift AT is ; When predicting the full load rate all standby shifts BT are enabled, and vehicles on nearby routes are dispatched.

6. The public transport capacity monitoring and dispatching method according to claim 5, characterized in that, The evaluation of the hybrid prediction model MM is carried out through the mean absolute percentage error , root mean square error, and coefficient of determination: Mean Absolute Percentage Error: ; Root Mean Square Error ; Coefficient of determination ; When or occurs, trigger the model parameter adjustment mechanism to recalculate the base passenger flow and the periodic passenger flow , and update the hybrid prediction model MM.

7. The public transport capacity monitoring and dispatching method according to claim 6, characterized in that, The predicted passenger flow is obtained by using a multi-time scale sliding window for passenger flow prediction. The multi-time scale sliding window includes a short-term prediction window, a medium-term prediction window, and a long-term prediction window. The short-term prediction window SW is used to predict the passenger flow change in the next 30 - 60 minutes, with a sampling frequency of 5 minutes; the medium-term prediction window MW is used to predict the passenger flow change in the next 3 - 6 hours, with a sampling frequency of 15 minutes; the long-term prediction window LW is used to predict the passenger flow change in the next 24 hours, with a sampling frequency of 30 minutes.

8. A public transportation capacity monitoring and dispatching system for implementing the public transportation capacity monitoring and dispatching method according to any one of claims 1-7, characterized in that, The system includes: a data collection module, a passenger flow prediction module, and a vehicle capacity scheduling module; The data collection module is used to collect the real-time passenger flow, historical passenger flow, and environmental factor data of the target bus line; and is used to standardize and perform anomaly detection on the collected data; the passenger flow prediction module is used to establish a hybrid prediction model MM to predict the passenger flow in different future time periods; the vehicle capacity scheduling module is used to dynamically adjust the vehicle configuration according to the predicted passenger flow.

9. The public transport capacity monitoring and dispatching system according to claim 8, characterized in that, The data collection module includes: An on-vehicle passenger flow statistics unit that collects the passenger flow data of getting on and off through an infrared sensor or a weight sensor; a station passenger flow monitoring unit that monitors the number of passengers waiting at the station through a camera and an intelligent video analysis system; a vehicle GPS positioning unit that collects the real-time position, speed, and operating status information of the vehicle; an environmental information collection unit that obtains weather information; all the units are connected to the data processing system of the data collection module in real time through a wireless communication network.

10. The public transportation capacity monitoring and dispatching system according to claim 8, characterized in that, The system further includes a visualization display module and an emergency handling module; The visualization display module is used to display the line operation status, vehicle distribution, passenger flow density heat map, and load factor warning information in real time; When the predicted load factor exceeds 95%, the emergency handling module automatically triggers an emergency plan, including: temporarily adding shuttle buses, adjusting the departure interval, and linking with surrounding bus lines, and at the same time pushing crowding tips and alternative travel suggestions to passengers through a public information platform.

Images