High-speed rail hub transfer urban rail transit optimization method based on AnyLogic

Abstract

The invention provides an optimization method based on intelligent simulation and machine learning for the optimization problem of a transfer streamline of a high-speed rail hub and urban rail transit. Comprising the following steps: (1) estimating the passenger carrying rate of a high-speed train, extracting the arrival passenger flow of the high-speed train, and analyzing the distribution characteristics of the passenger flow; (2) establishing a whole-process transfer simulation model, and identifying a bottleneck in a transfer process by analyzing a current situation simulation result; and (3) according to the transfer bottleneck, determining the optimization measures for the transfer of the urban rail transit in the high-speed rail hub, identifying the importance of the measure parameters, determining the priority of the optimization measures, and proposing optimization suggestions. The invention provides a method for acquiring sample data based on a parameter change simulation experiment, establishes a whole-process simulation model of the transfer of the urban rail transit in the high-speed rail hub, designs and provides a priority judgment method of hub transfer optimization measures, provides a theoretical basis for a high-speed rail hub transfer organization, more effectively optimizes the transfer process, and improves the transfer efficiency. And the overall service level of the high-speed railway hub is improved.

Description

Technical Field

[0001] The present invention relates to the field of high-speed rail hub transfer organization and computer simulation, and in particular to an AnyLogic-based high-speed rail hub transfer urban rail transit optimization method. Background Art

[0002] As an important part of the comprehensive transportation system, the comprehensive transportation hub is the spatial carrier for realizing the conversion of urban and external traffic flows and is the key to improving passenger transport efficiency.

[0003] At present, there is still much room for improvement in the efficiency of transfer and transportation organization services at high-speed rail hubs. Due to the large passenger capacity of high-speed rail transportation and the uneven distribution of arrival times, the passenger flow arriving on high-speed rail is usually uneven in both space and time, with the characteristics of short-term multi-point high aggregation. Compared with the passenger flow input of various modes of transportation it connects, the passenger flow of high-speed rail will have a greater impact on the output of transportation modes within the city, which will greatly test the organization of transfer system equipment and streamlines within the hub, affecting the transfer efficiency and service experience of passengers. Summary of the invention

[0004] Purpose of the invention: In view of the shortcomings of the prior art, this paper proposes an optimization method for high-speed rail hub transfers to urban rail transit based on AnyLogic. By analyzing the simulation results of the entire transfer process of high-speed rail hub transfers to urban rail transit, an effective basis is provided for optimizing the transfer organization method of high-speed rail hubs.

[0005] Technical Solution: An AnyLogic-based method for optimizing high-speed rail hub transfers to urban rail transit, including the following steps:

[0006] (1) Estimate the passenger load factor of high-speed trains, extract the arrival passenger flow of high-speed trains, and analyze the distribution characteristics of passenger flow;

[0007] (2) Establish a full-process transfer simulation model and identify bottlenecks in the transfer process by analyzing the current simulation results;

[0008] (3) Based on the transfer bottleneck, determine the optimization measures for high-speed rail hub transfers to urban rail transit, complete the parameter change simulation experiment, and identify the importance of the measure parameters.

[0009] Furthermore, extracting the high-speed rail arrival passenger flow in step 1 comprises the following steps:

[0010] 1.1, Determine the train passenger capacity based on the economic connection potential model:

[0011]

[0012] Among them, G ab represents the potential energy of economic connection between regions a and b; it is stipulated that each region has n0 types of industries, xaθ , x bθ Respectively represent the proportion of industry θ in region a and region b in the region’s gross economic output; P a , P b Represents the population size of regions a and b, respectively, using the year-end population of each region; E a , E b Respectively represent the economic scale of regions a and b, expressed in terms of regional GDP; l ab It indicates the degree of traffic connection between regions a and b, and the operation time of the high-speed railway transportation system is used to characterize the time-space distance index; k is the economic conversion coefficient of regions a and b, k = E a / (E a +E b );

[0013] The high-speed rail arrival passenger flow is estimated based on the EMU train passenger capacity corresponding to different economic connection potentials.

[0014] 1.2, Calculate the high-speed rail arrival passenger flow based on the high-speed rail passenger transport capacity and train passenger load factor:

[0015]

[0016] Among them, V emu_q represents the passenger flow of EMU trains arriving at the hub, person-times; p represents the number of EMU trains arriving at the high-speed railway passenger hub, trains; q represents the number of directions in which the high-speed railway arrives at the line; represents the train capacity of the i0th EMU; It represents the average passenger load factor of trains arriving in direction j0.

[0017] Furthermore, the establishment of the whole-process transfer simulation model in step 2 includes the following steps:

[0018] 2.1, Physical model construction

[0019] Collect detailed layout information of high-speed rail hubs and establish environmental models of train operation and passenger traffic behavior, mainly involving the spatial marking modules of pedestrian database and track database;

[0020] 2.2, Construction of Pedestrian Logical Model

[0021] On the basis of physical modeling, logical modeling is carried out according to the pedestrian transfer flow line. The core of pedestrian logical modeling is to divide the activities of pedestrians in the simulation area into multiple different movement processes, and finally form a complete pedestrian activity chain;

[0022] 2.3, Train logic model construction

[0023] Construct the process of high-speed rail trains and urban rail transit trains entering, stopping, and leaving stations. The implementation process of high-speed rail trains and urban rail transit trains in the simulation model: train generation combined with train timetable → select the track to enter the hub → stop to wait for passengers to drop off / board → drive out of the hub → train disappears.

[0024] Furthermore, the settings of the parameter selection and optimization experiment in step 3 are as follows:

[0025] The optimization measures are divided into facilities, layout guidance and demand categories. The facilities category includes channel category (number of stairs, elevator speed) and service category (number of ticket checking equipment, service time), the layout guidance category includes passenger flow guidance of channel facilities and layout guidance of service facilities, and the demand category refers to the passenger flow demand of urban rail transit passengers;

[0026] Without changing the main transfer process, the conventional service distribution in various service modules is set as a variable parameter. Combined with the changes in the number of facility services and the organizational layout, a variety of combination optimization schemes are generated, and based on this, a simulation experiment of parameter changes is carried out. The main process of the parameter change experiment: First, modify the parameter value type in the original simulation model and create a new parameter change sub-experiment. Then enter different parameter combinations and define the parameter change step size. Then set a fixed number of repetitions, run the experiment and output the data until the parameter combinations are exhausted and the experiment stops. The evaluation indicators are selected as follows:

[0027] (1) Hub transfer capacity

[0028] Hub transfer capacity refers to the maximum demand for transfers of a certain mode of transportation that a hub can meet under a specific operating scenario, and is an important indicator reflecting the hub's service level. The maximum passenger flow that successfully transfers to rail transit per unit time during the peak passenger flow period at the high-speed rail hub is selected as the quantitative characteristic value of transfer capacity;

[0029] (2) Average transfer time

[0030] The average transfer time refers to the average time it takes for passengers to transfer from one mode of transportation to another within a unit time period. The passenger transfer time is divided into three parts: the time to walk in the channel, the time to queue and receive service at the facility, and the waiting time. The formula for calculating the average transfer time is as follows:

[0031]

[0032] Among them, T trans represents the average transfer time; N represents the transfer passenger flow per unit time at the hub; t walk represents the time it takes passenger i to travel through the corridor within the hub; t serv represents the time passenger i spends queuing and receiving services at the facility; twait Indicates the waiting time for passenger i to arrive at the subway.

[0033] Furthermore, the optimization parameter importance identification method in step 3 is a lightweight gradient boosting machine algorithm.

[0034] Beneficial effects: The present invention estimates the passenger load of high-speed trains based on the economic connection potential, extracts the high-speed rail arrival passenger flow, establishes a full-process transfer simulation model, identifies the bottlenecks in the transfer process by analyzing the current simulation results, determines the optimization measures for high-speed rail hub transfer to urban rail transit, completes the parameter change simulation experiment, and identifies the importance of the measure parameters. The present invention can effectively identify the importance of various optimization measures and provide effective decision-making support for the operation organization of high-speed rail hub transfer to urban rail transit. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of an AnyLogic-based method for optimizing high-speed rail hub transfers to urban rail transit according to the present invention;

[0036] Figure 2 A physical model of a city's south railway station built using AnyLogic software;

[0037] Figure 3 A pedestrian logic model for transferring between some high-speed rail hubs and urban rail transit built using AnyLogic software;

[0038] Figure 4 This is a heat map of pedestrian density generated using Anylogic software to simulate the current status of high-speed rail hub transfer to urban rail transit. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further described below in conjunction with the accompanying drawings. Figure 1 The figure shows a flow chart of completing the simulation and optimization of high-speed rail hub transfer to urban rail transit of the present invention. The present invention optimizes the whole process of transfer between high-speed rail hub and urban rail transit based on intelligent simulation and machine learning, and mainly includes the following steps:

[0040] Step (1), estimating the passenger load factor of high-speed trains, extracting the high-speed train arrival passenger flow, and analyzing the distribution characteristics of the passenger flow;

[0041] Determine the train passenger load factor based on the economic connection potential model:

[0042]

[0043] Among them, G ab represents the potential energy of economic connection between regions a and b; it is stipulated that each region has n0 types of industries, x aθ , x bθRespectively represent the proportion of industry θ in region a and region b in the region’s gross economic output; P a , P b Represents the population size of regions a and b, respectively, using the year-end population of each region; E a , E b Respectively represent the economic scale of regions a and b, expressed in terms of regional GDP; l ab It indicates the degree of traffic connection between regions a and b, and the operation time of the high-speed railway transportation system is used to characterize the time-space distance index; k is the economic conversion coefficient of regions a and b, k = E a / (E a +E b ).

[0044] Based on this, the range of passenger load factor of EMU trains under different economic connection potentials is obtained. Since EMU trains usually run across different regions, when calculating the passenger load factor of the line, it is determined as the average value of the economic connection potential between each passing region and the target region.

[0045] Table 1 Passenger load factor of EMU trains

[0046]

[0047]

[0048] According to the economic connection potential model, the economic connection potential values ​​of some regions under the high-speed railway model are calculated as shown in Table 2. The shortest travel time is calculated and compared by train timetable, which refers to the shortest travel time of railway transportation between two places under the conditions of existing flights. Since there are generally many trains connecting the two places, and theoretically the two places can be connected by transfer, it is assumed that passengers always choose the train with the shortest travel time when traveling, and use it as the basis for determining the degree of transportation connection between the two regions. According to the train timetable passing through the high-speed railway hub, the provinces or municipalities passed by each high-speed railway line are determined to obtain the average economic connection potential of the line, and further obtain the passenger load factor of some trains with the destination as the high-speed railway hub.

[0049] Table 2 Calculation results of regional economic development indicators and economic connection potential

[0050]

[0051] The arrival passenger flow of high-speed rail can be calculated by the railway passenger transport capacity, which represents the maximum passenger transport volume that the railway can complete within a certain period of time. The calculation formula for the arrival passenger flow of high-speed rail is

[0052]

[0053] Among them, V emu_qrepresents the passenger flow of EMU trains arriving at the hub, person-times; p represents the number of EMU trains arriving at the high-speed railway passenger hub, trains; q represents the number of directions in which the high-speed railway arrives at the line; represents the train capacity of the i0th EMU; It represents the average passenger load factor of trains arriving in direction j0.

[0054] Table 3 Train schedule and passenger load factor passing through a certain city's South Station

[0055]

[0056] Step (2), establishing a full-process transfer simulation model, and identifying bottlenecks in the transfer process by analyzing the current simulation results;

[0057] 2.1 Physical model construction

[0058] AnyLogic software was selected to establish a micro-simulation model for the transfer between the high-speed rail hub and urban rail transit. AnyLogic physical modeling is an environmental model of train operation and passenger traffic behavior established after collecting detailed layout information. The establishment of the physical model mainly involves the spatial marking modules of the pedestrian library and the track library. The research focuses on the entire process of transferring from the high-speed rail hub to the urban rail transit. Therefore, the drawing of the high-speed rail waiting floor, the transfer area for other modes except urban rail transit, and other service facilities (such as information desks, convenience stores, toilets, etc.) were simplified during the modeling. By consulting the relevant information of a certain city’s south station and combining it with field survey data, the size, relative position and other information of different areas of a certain city’s south station were obtained, and a 3D model was constructed. Figure 3 shown.

[0059] 2.2 Construction of pedestrian logic model

[0060] On the basis of physical modeling, logical modeling is carried out according to the pedestrian transfer flow line. The core of pedestrian logical modeling is to divide the activities of pedestrians in the simulation area into multiple different movement processes, and finally form a complete pedestrian activity chain.

[0061] 2.3 Construction of train logic model

[0062] Construct the process of high-speed rail trains and urban rail transit trains entering, stopping, and leaving stations. The implementation process of high-speed rail trains and urban rail transit trains in the simulation model: train generation combined with train timetable → select the track to enter the hub → stop to wait for passengers to drop off / board → drive out of the hub → train disappears.

[0063] Based on the time when each pedestrian gets off the bus, the maximum number of transfer passengers and the average transfer time of pedestrians are aggregated hourly to obtain the transfer capacity and average transfer time per hour during the hub peak period. Among them, the maximum number of transfer passengers during peak hours fluctuates between 4,000 and 4,500, with an average of 4,145; the average transfer time is concentrated between 25 and 30 minutes, with an average of 28.07 minutes. According to the real-time density and thermal distribution of pedestrians in the model, during the peak arrival period of passengers, the transfer bottlenecks are mainly concentrated in the building / escalator passage facilities and service facilities such as gates and ticket vending machines, so these areas will be optimized as key areas.

[0064] Step (3), according to the transfer bottleneck, determine the optimization measures for high-speed rail hub transfer to urban rail transit, complete the parameter change simulation experiment, and identify the importance of the measure parameters.

[0065] 3.1 Parameter selection and optimization experiment

[0066] The optimization measures are divided into facilities, layout guidance and demand categories. The facilities category includes channel category (number of stairs, elevator speed) and service category (number of ticket checking equipment, service time), the layout guidance category includes passenger flow guidance of channel facilities and layout guidance of service facilities, and the demand category refers to the passenger flow demand of urban rail transit passengers.

[0067] Table 4 Value ranges of measure parameters

[0068]

[0069] The specific evaluation indicators are as follows:

[0070] (1) Hub transfer capacity

[0071] Hub transfer capacity refers to the maximum demand for transfers of a certain mode of transportation that a hub can meet in a specific operating scenario, and is an important indicator reflecting the hub's service level. The maximum passenger flow that successfully transfers to rail transit per unit time during the peak passenger flow period of the railway arrival in the high-speed rail hub is selected as the quantitative characteristic value of transfer capacity.

[0072] (2) Average transfer time

[0073] The average transfer time refers to the average time it takes for passengers to transfer from one mode of transportation to another within a unit time period. The passenger transfer time is divided into three parts: the time to walk in the channel, the time to queue and receive service at the facility, and the waiting time. The formula for calculating the average transfer time is as follows:

[0074]

[0075] Among them, T trans represents the average transfer time; N represents the transfer passenger flow per unit time at the hub; twalk represents the time it takes passenger i to travel through the corridor within the hub; t serv represents the time passenger i spends queuing and receiving services at the facility; t wait Indicates the waiting time for passenger i to arrive at the subway.

[0076] Without changing the main transfer process, the conventional service distribution in various service modules is set as a variable parameter. Combined with the changes in the number of facility services and the organizational layout, a variety of combination optimization schemes are generated, and a simulation experiment of parameter changes is carried out based on this. The main process of the parameter change experiment: modify the parameter value type in the original simulation model, create a new parameter change sub-experiment, then enter different parameter combinations, define the parameter change step, and then set a fixed number of repetitions, run the experiment and output the data until the parameter combination is exhausted and the experiment stops.

[0077] 3.2 Identification measure parameter importance

[0078] Based on the LightGBM (Light Gradient Boosting Machine) algorithm, a priority determination method for improvement measures in the entire transfer process is designed. The optimization parameters are determined as elevator speed, service time of high-speed rail exit gates, service time of urban rail transit automatic ticket machines, service time of urban rail transit entrance gates, number of high-speed rail exit gates per group, number of urban rail transit automatic ticket machines per group, number of urban rail transit entrance gates per group, iron horse guardrails at channel facilities, iron horse guardrails at service facilities, and the proportion of passengers who need to buy tickets. The evaluation indicators are average transfer time and maximum number of transfer passengers. The whole process simulation model is used to output sample data under different parameter combinations. Based on the machine learning feature selection method, an importance identification model for improvement measure parameters is established.

[0079] (1) Collect relevant data from high-speed rail hubs and urban rail transit systems. The data includes eight main transfer influencing factors (the optimization measures are determined as elevator speed, service time of high-speed rail exit gates, service time of urban rail transit automatic ticket machines, service time of urban rail transit entrance gates, number of each group of high-speed rail exit gates, number of each group of urban rail transit automatic ticket machines, number of each group of urban rail transit entrance gates, installation of iron guardrails at passage facilities, installation of iron guardrails at service facilities, and proportion of passengers who need to buy tickets) and two evaluation indicators (average transfer time and maximum number of transfer passengers). The input data set is X = {x1, x2, ..., x n}, where x i is the i-th data point, and n is the total amount of data. For each data point, define the label or target variable as y = {y1, y2, ..., y n}. To ensure the integrity and consistency of the data, missing value filling and outlier processing were performed. Since LightGBM supports direct input of categorical features, categorical data is retained. The dataset is randomly divided into training and test sets, with the training set accounting for 80% and the test set accounting for 20%.

[0080] (2) Build a LightGBM dataset. Convert the data into a data format supported by LightGBM. The LightGBM Dataset class automatically manages categorical features and data types. The training set is denoted as D train =(X train ,y train ), the test set is denoted as D test =(X test ,y test ).

[0081] (3) Set parameters. The performance of LightGBM is affected by a variety of hyperparameters. In order to optimize the performance and stability of the model, the following main parameters are set. The boosting type selects the gradient boosting decision tree (GBDT). The objective function selects the regression or classification loss function according to the task type; for regression tasks, the loss function is the root mean square error (RMSE):

[0082]

[0083] Among them, y i is the actual value, is the predicted value, n is the number of samples. The model complexity is controlled by setting the number of leaf nodes. The regularization parameter λ L1 and λ L2 Used to control the complexity of the model to prevent overfitting. The feature sampling ratio ranges from [0,1], which controls the feature ratio used by each tree, and the sample sampling ratio ranges from [0,1], which controls the proportion of randomly selected samples in each iteration.

[0084] (4) Use LightGBM's gradient boosting framework to train the model. The goal of training is to minimize the loss function:

[0085]

[0086] In each iteration, the model adjusts the leaf node weights according to the current negative gradient (i.e. the gradient of the loss function) to reduce the error. The update formula for each round of iteration is:

[0087]

[0088] Among them, F m(X) is the prediction result of the mth tree, η is the learning rate, j is the number of leaf nodes, J is the total number of all leaf nodes in the tree, R j is the leaf node area, w j is the weight of the leaf node, I(X∈R j ) is the indicator function. When the input X is located in the region R of the jth leaf node j When internal, I(X∈R j ) is 1, otherwise I(X∈R j ) is 0.

[0089] (5) Make predictions on the test set to evaluate the model performance. The root mean square error (RMSE) is used to evaluate the performance of the regression task to quantify the prediction error:

[0090]

[0091] (6) Feature importance analysis. LightGBM provides feature importance assessment to identify factors that have a greater impact on transfer evaluation indicators. Feature importance is obtained by calculating the cumulative contribution of each feature to the reduction of model loss:

[0092]

[0093] Among them, FI represents the feature importance, M is the total number of trees, ΔL m is the loss reduction brought by the tree, I(f j ∈T m ) is the indicator function, when the characteristic f j Appears in the mth tree T m In the middle, I(f j ∈T m ) is 1, otherwise I(f j ∈T m ) is 0.

[0094] (7) Parameter tuning. Adjust the parameters through five-fold cross validation. Define the average root mean square error (RMSE) of cross validation as the optimization objective:

[0095]

[0096] Table 5 Results of relative importance determination of optimization parameters

[0097]

[0098] The LightGBM algorithm is used to calculate the relative importance of different optimization measure parameters to the two evaluation indicators, and the priority of factor improvement in the optimization scheme design is clarified. In the scenario of a certain city's South Station, the focus is on guiding and improving the demand for urban rail transit ticket purchases, considering the number and service time of urban rail transit station hall service facilities, and appropriately considering the number of high-speed rail exit gates. The high-speed rail exit service time and the optimization of passenger flow layout at the service facilities are not considered for the time being. It is recommended that high-speed rail hubs improve hub transfer efficiency by reducing offline ticket purchase demand, shortening facility service time, and adjusting the number of facilities.

Claims

1. An optimization method for high-speed rail hub transfer to urban rail transit based on AnyLogic, characterized by: The following steps are involved: (1) Estimate the passenger load factor of high-speed trains, extract the arrival passenger flow of high-speed trains, and analyze the distribution characteristics of passenger flow; (2) Establish a full-process transfer simulation model and identify transfer bottlenecks by analyzing the current simulation results; (3) Based on the transfer bottleneck, determine the optimization measures for high-speed rail hub transfers to urban rail transit, complete the parameter change simulation experiment, and identify the importance of the measure parameters.

2. The AnyLogic-based high-speed rail hub transfer urban rail transit optimization method according to claim 1 is characterized in that: The step 1 of extracting the high-speed rail arrival passenger flow comprises the following steps: 1.1, Determine the train passenger capacity based on the economic connection potential model: Among them, G ab represents the potential energy of economic connection between regions a and b; it is stipulated that each region has n0 types of industries, x aθ , x bθ Respectively represent the proportion of industry θ in region a and region b in the region’s gross economic output; P a , P b Represents the population size of regions a and b, respectively, using the year-end population of each region; E a , E b Respectively represent the economic scale of regions a and b, expressed in terms of regional GDP; l ab It indicates the degree of traffic connection between regions a and b, and the operation time of the high-speed railway transportation system is used to characterize the time-space distance index; k is the economic conversion coefficient of regions a and b, k = E a / (E a +E b ); According to the passenger load factor of EMU trains corresponding to different economic connection potentials, the high-speed rail arrival passenger flow is estimated; 1.2, Calculate the high-speed rail arrival passenger flow based on the high-speed rail passenger transport capacity and train passenger load factor: Among them, V emu_q represents the passenger flow of EMU trains arriving at the hub; p represents the number of EMU trains arriving at the high-speed railway passenger hub; q represents the number of directions in which the high-speed railway arrives at the line; represents the train capacity of the i0th EMU; It represents the average passenger load factor of trains arriving in direction j0.

3. The AnyLogic-based high-speed rail hub transfer urban rail transit optimization method according to claim 1 is characterized in that: The step 2 of establishing the whole-process transfer simulation model includes the following steps: 2.1, build physical model: Collect detailed layout information of high-speed rail hubs, establish environmental models of train operation and passenger traffic behavior, and involve spatial marking modules for pedestrian and track libraries; 2.2, build a pedestrian logic model: Based on the physical modeling, logical modeling is carried out according to the pedestrian transfer flow line; The core of pedestrian logic modeling is to divide pedestrian activities in the simulation area into multiple different motion processes, and finally form a complete pedestrian activity chain; 2.3, build the train logic model: Construct the process of high-speed railway trains and urban rail transit trains entering, stopping and leaving stations; the implementation process of high-speed railway trains and urban rail transit trains in the simulation model: train generation combined with train timetable → select the track to enter the hub → stop to wait for passengers to get off / board → drive out of the hub → train disappears.

4. The AnyLogic-based high-speed rail hub transfer urban rail transit optimization method according to claim 1 is characterized in that: The settings of parameter selection and optimization experiment in step 3 are as follows: The optimization measures are divided into facility category, layout guidance category and demand category according to functional attributes; the facility category includes channel category and service category, the layout guidance category includes passenger flow guidance of channel facilities and layout guidance of service facilities, and the demand category refers to the passenger flow demand of urban rail transit passengers; Without changing the main transfer process, the conventional service distribution in various service modules is set as a variable parameter. Combined with the changes in the number of facility services and the organizational layout, a variety of combination optimization schemes are generated, and based on this, a simulation experiment of parameter changes is carried out; parameter change experiment process: first modify the parameter value type in the original simulation model, and create a new parameter change sub-experiment, then enter different parameter combinations, define the parameter change step, and then set a fixed number of repetitions, run the experiment and output the data until the parameter combination is exhausted and the experiment stops; the evaluation indicators are selected as follows: (1) Hub transfer capacity Hub transfer capacity refers to the maximum demand for transfers of a certain mode of transportation that a hub can meet under a specific operating scenario. It is an important indicator reflecting the hub service level. The maximum passenger flow that successfully transfers to rail transit per unit time during the peak passenger flow period of the high-speed rail hub is selected as the quantitative characteristic value of transfer capacity. (2) Average transfer time The average transfer time refers to the average time it takes for passengers to transfer from one mode of transportation to another within a unit time period. The passenger transfer time is divided into three parts: the time to walk in the channel, the time to queue for facilities and receive services, and the waiting time. The calculation formula for the average transfer time is as follows: Among them, T trans represents the average transfer time; N represents the transfer passenger flow per unit time at the hub; t walk represents the passage travel time of passenger i in the hub; t serv represents the time passenger i spends queuing and receiving services at the facility; t wait Indicates the waiting time for passenger i to arrive at the subway.

5. The AnyLogic-based high-speed rail hub transfer urban rail transit optimization method according to claim 1 is characterized in that: The optimization parameter importance identification method in step 3 is a lightweight gradient boosting machine algorithm.

Images