Transportation situation deduction method and device for rail transit, electronic equipment and storage medium

Abstract

The invention relates to a rail transit transportation situation deduction method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the multi-step deduction of a column flow feature matrix of a station at a current statistical moment through a column flow situation deduction network for any station in a target rail transit network, and obtaining a column flow feature matrix of the station; determining a train delay condition of the station at a target statistical moment after the current statistical moment; for any station in the target rail transit network, performing multi-step deduction on the passenger flow characteristic matrix of the station at the current statistical moment by using a passenger flow situation deduction network, and determining the passenger retention condition of the station at the target statistical moment; and according to the train delay condition and the passenger retention condition of each station in the target rail transit network at the target statistical moment, carrying out transportation risk assessment, and determining the global transportation capacity risk of the target rail transit network at the target statistical moment. According to the embodiment of the invention, the accuracy and reliability of transportation situation deduction can be improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of data analysis, and particularly to a method and device for inferring the transportation situation of rail transit, an electronic device, and a storage medium. Background Art

[0002] During the operation of urban rail transit, train arrival delays may occur due to equipment failures or natural disaster factors, resulting in cascading delays of subsequent trains. At the same time, due to the complex coupling relationship of the rail transit network, delays on a single branch line will not only cause passenger congestion on that line, but also affect other lines, and even cause the paralysis of the entire rail transit network. Therefore, it is necessary to combine the train delay situation to predict the transportation situation and evaluate the operation capacity risk of the rail transit network, so as to assist the decision-making department to make reasonable decisions in a timely manner and ensure the normal operation of the rail transit. Summary of the Invention

[0003] In view of this, the present disclosure provides a technical solution for a method and device for inferring the transportation situation of rail transit, an electronic device, and a storage medium.

[0004] According to one aspect of the present disclosure, there is provided a method for inferring the transportation situation of rail transit, including: for any station in a target rail transit network, using a train flow situation inference network to perform multi-step inference on the train flow feature matrix of the station at the current statistical moment, and determining the train delay situation of the station at a target statistical moment after the current statistical moment, where the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station; for any station in the target rail transit network, using a passenger flow situation inference network to perform multi-step inference on the passenger flow feature matrix of the station at the current statistical moment, and determining the passenger congestion situation of the station at the target statistical moment, where the passenger flow feature matrix of any station at the current statistical moment includes: the congestion passenger situations of each station in the target rail transit network at the multiple historical statistical moments; performing a transportation risk assessment based on the train delay situation and the passenger congestion situation of each station in the target rail transit network at the target statistical moment, and determining the global operation capacity risk of the target rail transit network at the target statistical moment, where the global operation capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transportation demand at the target statistical moment.

[0005] In a possible implementation, the column flow situation deduction network is a recurrent neural network, including a long short-term memory network layer and a fully connected layer; for any station in the target rail transit network, using the column flow situation deduction network to perform multi-step deduction on the column flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment, including: initializing the long short-term memory network layer to determine the initial hidden state corresponding to the long short-term memory network layer; using the long short-term memory network layer to perform iterative processing on the column flow feature matrix corresponding to the station and updating the initial hidden state until the number of iterations is equal to the preset number of times to determine the target hidden state corresponding to the long short-term memory network layer; converting the target hidden state according to the fully connected layer to determine the train delay situation of the station at the target statistical moment.

[0006] In a possible implementation, for any station in the target rail transit network, using the passenger flow situation deduction network to perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the target statistical moment, including: according to the passenger flow situation deduction network, performing single-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the next statistical moment after the current statistical moment and before the target statistical moment; based on the sliding window method, updating the passenger flow feature matrix of the station at the current statistical moment according to the passenger retention situation of the station at the next statistical moment to determine the updated passenger flow feature matrix; repeating the above steps according to the updated passenger flow feature matrix until the number of single-step deductions is equal to the preset number of times to determine the passenger retention situation of the station at the target statistical moment.

[0007] In a possible implementation, the passenger flow situation deduction network is a directed graph spatio-temporal convolutional network based on multi-head attention, including: a spatial multi-head attention module, a temporal multi-head attention module, a directed graph spatio-temporal convolutional module, and a fully connected layer.

[0008] In a possible implementation, the global operation capacity risk of the target rail transit network at the target statistical moment includes: the station operation capacity risk of each station in the target rail transit network at the target statistical moment, and the section operation capacity risk of any section formed by two stations in the target rail transit network at the target statistical moment; the transportation risk assessment is carried out according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment to determine the global operation capacity risk of the target rail transit network at the target statistical moment, including: for any one station, determining the station operation capacity risk of this station at the target statistical moment according to the train delay situation and passenger retention situation of this station at the target statistical moment; for any one section, determining the section operation capacity risk of this section at the target statistical moment according to the train delay situations of the two stations corresponding to this section at the target statistical moment.

[0009] In a possible implementation, the station operation capacity risk of any one station at the target statistical moment includes: the station passenger flow saturation risk and platform passenger retention risk of this station at the target statistical moment; for any one station, determining the station operation capacity risk of this station at the target statistical moment according to the train delay situation and passenger retention situation of this station at the target statistical moment, including: determining the in-station passenger flow volume of this station at the target statistical moment; determining the passenger throughput and platform evacuation capacity of this station at the target statistical moment according to the train delay situation of this station at the target statistical moment; determining the station passenger flow saturation risk of this station at the target statistical moment according to the in-station passenger flow volume and passenger throughput of this station at the target statistical moment, and the corresponding passenger flow saturation risk consequences of this station; determining the platform passenger retention risk of this station at the target statistical moment according to the passenger retention situation and platform evacuation capacity of this station at the target statistical moment, and the corresponding passenger retention risk consequences of this station.

[0010] In a possible implementation, for any one section, determining the section operation capacity risk of this section at the target statistical moment according to the train delay situations of the two stations corresponding to this section at the target statistical moment, including: determining the section passenger flow volume of this section at the target statistical moment; determining the section transportation capacity of this section at the target statistical moment according to the train delay situations of the two stations corresponding to this section at the target statistical moment; determining the section operation capacity risk of this section at the target statistical moment according to the section passenger flow volume and section transportation capacity of this section at the target statistical moment, and the corresponding passenger flow saturation risk consequences of this section.

[0011] According to another aspect of the present disclosure, there is provided a device for inferring the transportation situation of rail transit, including: a train flow situation inference module, configured to, for any station in a target rail transit network, use a train flow situation inference network to perform multi-step inference on the train flow feature matrix of the station at the current statistical moment, and determine the train delay situation of the station at a target statistical moment after the current statistical moment. Among them, the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station; a passenger flow situation inference module, configured to, for any station in the target rail transit network, use a passenger flow situation inference network to perform multi-step inference on the passenger flow feature matrix of the station at the current statistical moment, and determine the passenger retention situation of the station at the target statistical moment. Among them, the passenger flow feature matrix of any station at the current statistical moment includes: the passenger retention situations of each station in the target rail transit network at the multiple historical statistical moments; a transportation risk assessment module, configured to perform transportation risk assessment according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment, and determine the global transport capacity risk of the target rail transit network at the target statistical moment. Among them, the global transport capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transport demand at the target statistical moment.

[0012] According to another aspect of the present disclosure, there is provided an electronic device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the above method.

[0013] According to another aspect of the present disclosure, there is provided a non-volatile computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0014] In the embodiments of the present disclosure, for any station in the target rail transit network, a train flow situation deduction network can be used to perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment, so as to determine the train delay situation of the station at the target statistical moment after the current statistical moment, thereby fully considering the impact of train delays on the uncertainty of train flow during the operation of urban rail transit, and through multi-step deduction, improving the short-term dependence of train flow situation deduction, increasing the adaptability to train delays, and improving the accuracy and reliability of train flow situation deduction; wherein, the train flow feature matrix of any station at the current statistical moment includes the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station. For any station in the target rail transit network, a passenger flow situation deduction network can be used to perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment, so as to determine the passenger retention situation of the station at the target statistical moment. By separately deducing the train flow situation and the passenger flow situation, the detailed attention to the transportation situation can be improved, and the advantages of multi-step deduction can also be effectively utilized during the passenger flow situation deduction process to improve the accuracy and reliability of the passenger flow situation deduction; wherein, the passenger flow feature matrix of any station at the current statistical moment includes the passenger retention situations of each station in the target rail transit network at multiple historical statistical moments. After completing the separate deductions of the train flow situation and the passenger flow situation, the transportation risk assessment can be carried out according to the train delay situation and the passenger retention situation of each station in the target rail transit network at the target statistical moment, so as to determine the global transport capacity risk of the target rail transit network at the target statistical moment, in order to indicate whether the target rail transit network can meet the passenger demand at the target statistical moment, realize a complete transportation situation deduction process, and ensure that the global transport capacity risk of the target rail transit network at the target statistical moment has high prediction accuracy and stability.

[0015] Other features and aspects of the present disclosure will become clear from the following detailed description of the exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are included in and constitute a part of this specification, illustrate exemplary embodiments, features, and aspects of the present disclosure and are used to explain the principles of the present disclosure together with the specification.

[0017] Figure 1 The flowchart showing a method for deducing the transportation situation of a rail transit according to an embodiment of the present disclosure;

[0018] Figure 2 The residual distribution diagram showing the prediction of train delay situations under different parameter settings according to an embodiment of the present disclosure;

[0019] Figure 3Shows a schematic process diagram of a single-step prediction according to the prior art;

[0020] Figure 4 Shows a schematic process diagram of a multi-step deduction according to an embodiment of the present disclosure;

[0021] Figure 5 Shows a schematic diagram of a road network transportation situation assessment system according to the prior art;

[0022] Figure 6 Shows a schematic diagram of the prediction effect of a train flow situation deduction network according to an embodiment of the present disclosure;

[0023] Figure 7 Shows a schematic structural diagram of a passenger flow situation deduction network according to an embodiment of the present disclosure;

[0024] Figure 8 Shows a prediction effect diagram of a transport capacity risk according to an embodiment of the present disclosure;

[0025] Figure 9 Shows a block diagram of a transport situation deduction device for rail transit according to an embodiment of the present disclosure;

[0026] Figure 10 Shows a block diagram of an electronic device according to an embodiment of the present disclosure. Detailed implementation manners

[0027] The following will describe various exemplary embodiments, features, and aspects of the present disclosure in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements with the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings do not have to be drawn to scale unless otherwise specified.

[0028] As used herein, the terms "including", "comprising", "having", or their variants are open-ended and include one or more stated features, wholes, elements, steps, components, or functions, but do not exclude the existence or addition of one or more other features, wholes, elements, steps, components, functions, or groups thereof.

[0029] When an element is referred to as being "connected", "coupled", "responsive" or a variant thereof to another element, it can be directly connected, coupled, or responsive to the other element, or there can be an intermediate element.

[0030] Although the terms first, second, third, etc. may be used herein to describe various elements / operations, these elements / operations should not be limited by these terms. These terms are only used to distinguish one element / operation from another. Thus, without departing from the teachings of the inventive concept, a first element / operation in some embodiments can be referred to as a second element / operation in other embodiments.

[0031] As used herein, the term "exemplary" means "serving as an example, embodiment, or illustration". Any embodiment described as "exemplary" herein is not necessarily to be construed as superior or better than other embodiments.

[0032] As used herein, the term "and / or" is merely a description of the associated relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the term "at least one" as used herein means any one of a plurality or any combination of at least two of a plurality. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set composed of A, B, and C.

[0033] In addition, for a better illustration of the present disclosure, numerous specific details are given in the following detailed description. Those skilled in the art should understand that the present disclosure can be implemented without some specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art are not described in detail so as to highlight the gist of the present disclosure.

[0034] In an urban rail transit system, the supply-demand matching of transport capacity and transport demand is a core element reflecting the operation quality and efficiency of urban rail transit, and is also the key to realizing the efficiency optimization and operation safety guarantee of urban rail transit.

[0035] The transport capacity and transport demand of an urban rail transit system can be respectively reflected by the train flow and passenger flow corresponding to the urban rail transit system. Among them, the train flow can be used to describe the operation of any train running line in the urban rail transit system, and the passenger flow can be used to describe the change in the number of passengers in any station in the urban rail transit system. Both the train flow and the passenger flow have strong dynamics and a certain degree of randomness, and have spatio-temporal distribution characteristics. In order to effectively express the supply-demand matching characteristics of the rail transit system at the system level, the dynamic matching relationship between the train flow and the passenger flow can be characterized by the concept of transport situation.

[0036] During the operation of urban rail transit, train arrival delays may occur due to equipment failures or natural disaster factors, also known as train lateness, and cause subsequent train delays. At the same time, due to the complex coupling relationship of the rail transit network, the delay of a single branch line will not only cause passenger congestion on that line, but also affect other lines, and even cause the paralysis of the entire rail transit network.

[0037] Therefore, it is necessary to combine the train delay situation to deduce the transportation situation of the rail transit network, so as to assist the decision-making department to make reasonable decisions in a timely manner, adjust the train operation, and ensure the normal operation of the rail transit. Specifically, the situation deduction means predicting the future development trend of the situation; in the field of urban rail transit, the transportation situation deduction can include steps such as the perception, understanding, evaluation, prediction, and safety decision-making of the transportation situation, and can be used to predict the changes in various transportation indicators of urban rail transit. However, in the existing technology, the common research on the transportation situation deduction of urban rail transit does not consider the train flow uncertainty caused by train delays.

[0038] In view of this, the present disclosure provides a method for deducing the transportation situation of rail transit, which can fully consider the impact of train flow uncertainty caused by train delays during the operation of urban rail transit, and can improve the short-term dependence of the transportation situation deduction and increase the adaptability to train delays through multi-step deduction; on this basis, by combining the separate deductions of the train flow situation and the passenger flow situation, it is also possible to improve the detailed attention to the transportation situation and further improve the accuracy and reliability of the transportation situation deduction. The following details the method for deducing the transportation situation of rail transit provided by the present disclosure.

[0039] Figure 1 The flowchart of a method for deducing the transportation situation of rail transit according to an embodiment of the present disclosure is shown. This method for deducing the transportation situation of rail transit can be executed by an electronic device such as a terminal device or a server. The terminal device can be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. This method for deducing the transportation situation of rail transit can be implemented by the processor calling the computer-readable instructions stored in the memory. Alternatively, the method for deducing the transportation situation of rail transit can be executed by the server. As Figure 1 shown, this method for deducing the transportation situation of rail transit includes:

[0040] In step S101, for any station in the target rail transit network, using the train flow situation deduction network, perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment, where the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station.

[0041] Among them, the target rail transit network can represent any complete urban rail transit road network, any area in the urban rail transit road network, or any train running line, and can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto. For the specific form of the target rail transit network, reference can be made to the implementation manners in related technologies. For example, it can be a road network composed of common rail transit tools such as subways and intercity trams. The present disclosure does not make specific limitations thereto.

[0042] The train flow situation deduction network can represent a neural network model for predicting the train delay situation of any station in the target rail transit network. The specific form and structure of the train flow situation deduction network can be flexibly set according to actual usage requirements. For example, it can be set as a Long Short Term Memory (LSTM), a Multi-Head Attention Graph Convolutional Network (MATGCN), etc. The present disclosure does not make specific limitations thereto.

[0043] The train flow feature matrix of any station at the current statistical moment can include the train delay situations of multiple stations before this station in multiple historical statistical moments before the current statistical moment in any train running line passing through this station, so as to describe the train flow uncertainty of this train running line at the current statistical moment.

[0044] Here, the multiple historical statistical moments are obtained by stepping forward according to a preset time step and a first step number with the current statistical moment as the reference. Among them, the specific value of the preset time step can be flexibly set according to actual usage requirements, and reference can be made to the preset departure interval time of this train running line, etc. The present disclosure does not make specific limitations thereto; the first step number, which can also be called the historical time step length, is used to indicate how many steps need to be taken forward with the preset time step, and its specific value can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto.

[0045] In an example, when the preset time step is set to 5 minutes (min), the value of the first step number is 6, and the current statistical moment is 12:00, the multiple historical statistical moments can include 11:30, 11:35, 11:40, 11:45, 11:50, 11:55, and 12:00.

[0046] The train delay situation of any station at any historical statistical moment is used to indicate whether there is a train delay within the preset time step after this historical statistical moment and the corresponding delay duration, and its specific content can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto.

[0047] In one example, when the preset time step is 5 minutes, the train delay situation of any station at the statistical moment of 12:00 can include the delay duration of all trains arriving at the station during the 5 minutes from 12:00 to 12:05.

[0048] The specific form of the train flow characteristic matrix can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations thereto.

[0049] In one example, since there is no coupling relationship in the operation of different train lines in the urban rail transit network, therefore, only the train delay situation of each train line itself can be considered. For any train line, assuming that the train line includes N stations, they can be respectively expressed as {s1, s2,..., s n}.

[0050] For any one of the stations s i , there can be a train delay time series with a length of L, which can be expressed as {D i1 , D i2 , ……, D iL}, where any one element can represent the delay situation of the trains arriving at the station s i at each historical statistical moment. The specific value of the length L here can be flexibly set according to actual usage requirements and depends on the specific value of the first step number. The present disclosure does not make specific limitations thereto.

[0051] For any train line passing through the station s i , the train delay situation of the station s i at the current statistical moment is related to the train delay situation of any station s i before the station s i-k , and the train delay situation of the station s i at the historical statistical moment. Therefore, the train flow characteristic matrix of the station s i at the current statistical moment can be expressed by formula (1):

[0052] D = {D i-k，t-1 , D i-k，t-2 ,..., D i-k，t-τ ,..., D i，t-1 , D i，t-2 ,..., D i，t-τ} (1)

[0053] Among them, k represents the station s on the train line iThe number of previous stations, and its specific value can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto; τ represents the number of first-step advancements, and its specific value can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto.

[0054] The number k of previous stations and the value of the number τ of first-step advancements determine the data volume of the column flow feature matrix of the station s i at the current statistical moment, thereby affecting the prediction of the station s i accuracy of the train delay situation at the target statistical moment.

[0055] Figure 2 Shows a residual distribution diagram of train delay situation prediction under different parameter settings according to an embodiment of the present disclosure. As Figure 2 shown, the horizontal axis represents the residual value of train delay situation prediction, and the vertical axis represents the probability of each residual value; the blue curve represents the residual distribution curve of train delay situation prediction when the number k of previous stations is 3 and the number τ of first-step advancements is 6; the green curve represents the residual distribution curve of train delay situation prediction when the number k of previous stations is 1 and the number τ of first-step advancements is 6; the red curve represents the residual distribution curve of train delay situation prediction when the number k of previous stations is 2 and the number τ of first-step advancements is 6; the yellow curve represents the residual distribution curve of train delay situation prediction when the number k of previous stations is 3 and the number τ of first-step advancements is 4; the purple curve represents the residual distribution curve of train delay situation prediction when the number k of previous stations is 3 and the number τ of first-step advancements is 3.

[0056] As Figure 2 shown, for any parameter combination, the probability of the residual being 0 is the highest, and the residual distribution satisfies the normal distribution. When the number k of previous stations is 3 and the number τ of first-step advancements is 6, the residual distribution curve of train delay situation prediction is the sharpest, indicating that the probability of small residuals is the highest, which also means that when the number k of previous stations is 3 and the number τ of first-step advancements is 6, the performance of the column flow trend deduction network is the best, and the accuracy of train delay situation prediction is the highest.

[0057] For any station in the target rail transit network, the column flow feature matrix of the station at the current statistical moment can be input into the column flow trend deduction network for multi-step deduction, so as to determine the train delay situation of the station at the target statistical moment after the current statistical moment, to indicate whether there is a train delay at the target statistical moment for the station and the corresponding delay duration, and its specific content can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations thereto.

[0058] The target statistical moment can be determined by stepping backward from the current statistical moment based on a preset time step and the number of second steps. Among them, the number of second steps, which can also be referred to as the future time step length, is used to indicate how many steps need to be taken backward with the preset time step. Its specific value can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0059] In one example, the preset time step is set to 5 minutes, and the value of the number of second steps is 3. Then the target statistical moment is the moment 15 minutes after the current statistical moment.

[0060] The multi-step deduction in the embodiments of the present disclosure is relative to single-step prediction. Single-step prediction usually means directly predicting the train delay situation of any station at the target statistical moment based on the train flow characteristic matrix of the station at the current statistical moment, that is, only one prediction is made.

[0061] Figure 3 The process schematic diagram of a single-step prediction according to the prior art is shown. As Figure 3 shown, T6 represents the current statistical moment, T0 to T5 represent 6 historical statistical moments before the current statistical moment, T7 to T9 represent three future statistical moments after the current statistical moment, and T9 is taken as the target statistical moment; 1 square represents 1 preset time step, there is an interval of 1 preset time step between T1 and T0, an interval of 2 preset time steps between T2 and T0, and so on.

[0062] Single-step prediction takes the train delay situation of any station from T0 to T5 as input and directly determines the train delay situation of the station at T9, without predicting the train delay situation of the station at T7 and T8.

[0063] Single-step prediction usually cannot analyze in detail the dependency relationship of each preset time step between the current statistical moment and the target statistical moment, lacking short-term dependence; while in the train delay scenario, there will be sudden changes in the target rail transit network, with strong short-term dependence; therefore, in the train delay scenario, the accuracy of using the single-step prediction method for train flow trend deduction is relatively low.

[0064] The multi-step deduction provided by the present disclosure can mean that based on the train flow characteristic matrix of any station at the current statistical moment, using the sliding window method, multiple iterative predictions are made according to the preset time step until the train delay situation of the station at the target statistical moment is determined.

[0065] Figure 4 The process schematic diagram of a multi-step deduction according to the embodiments of the present disclosure is shown. As Figure 4As shown, T6 represents the current statistical moment, T0 to T5 represent the 6 historical statistical moments before the current statistical moment, T7 to T9 represent the three future statistical moments after the current statistical moment, and T9 is the target statistical moment; 1 square represents 1 preset time step. There is an interval of 1 preset time step between T1 and T0, an interval of 2 preset time steps between T2 and T0, and so on.

[0066] For multi-step deduction, the train delay situation of any station from T0 to T5 is used to construct the train flow characteristic matrix of this station at T6. First, the train flow situation deduction network is used to predict the train delay situation of this station at T7, and with the sliding window method, the train flow characteristic matrix is updated according to the train delay situation of this station at T7, removing the train delay situation of this station at T1, and obtaining the updated train flow characteristic matrix constructed by the train delay situation of this station from T2 to T7.

[0067] According to the updated train flow characteristic matrix, the train flow situation deduction network is used again for prediction to determine the train delay situation of this station at T8, and the above process is repeated until the train delay situation of this station at T9 is determined.

[0068] Through multi-step deduction, the dependence relationship of each preset time step between the current statistical moment and the target statistical moment can be analyzed in more detail, with high short-term dependence, strong adaptability to the train delay scenario, and can improve the accuracy of train flow situation deduction, making the predicted train delay situation of any station at the target statistical moment have high accuracy and reliability.

[0069] In an example, in the train flow characteristic matrix of any station corresponding to the current statistical moment, the number of previous stations k = 3, and the number of first-step advances τ = 6; based on the same train flow situation deduction network, single-step prediction and multi-step deduction are respectively used to predict the train delay situation of this station at the target statistical moment, and the prediction accuracy is evaluated through three indicators: root mean square error (RMSE), mean absolute error (MAE), and weighted mean average percentage error (WMAPE).

[0070] Specifically, RMSE can be expressed as formula (2):

[0071]

[0072] Among them, y i represents the predicted train delay situation of any station at the target statistical moment; y i *It represents the actual train delay situation of the station at the target statistical moment; n represents the number of samples used. MAE can be expressed as formula (3):

[0073]

[0074] WMAPE can be expressed as formula (4):

[0075]

[0076] Referring to Table 1, as shown in Table 1, based on the column flow situation deduction network, when predicting the train delay situation of the station at the target statistical moment using the multi-step deduction method, the corresponding RMSE, MAE, and WMAPE are all smaller than those when predicting the train delay situation of the station at the target statistical moment using the single-step prediction method based on the same column flow situation deduction network; that is to say, compared with the single-step prediction, the multi-step deduction method provided by the embodiments of the present disclosure has higher accuracy and reliability.

[0077] Table 1

[0078]

[0079] Later, in combination with the possible implementation manners of the present disclosure, the process of using the column flow situation deduction network to perform multi-step deduction on the column flow feature matrix of any station in the target rail transit network at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment will be described in detail, and details will not be elaborated here.

[0080] In step S102, for any station in the target rail transit network, use the passenger flow situation deduction network to perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the target statistical moment, where the passenger flow feature matrix of any station at the current statistical moment includes: the passenger retention situations of each station in the target rail transit network at multiple historical statistical moments.

[0081] The passenger flow situation deduction network can represent a neural network model used to predict the passenger retention situation of any station in the target rail transit network; the specific form and result of the passenger flow situation deduction network can be flexibly set according to actual usage requirements. For example, it can be set as an interactive graph network model (IG-Net), a spatio-temporal graph attention convolutional neural network model (TGACN), etc., and the present disclosure does not make specific limitations on this.

[0082] The passenger flow feature matrix of any station at the current statistical moment can include the situation of stranded passengers at each station in the target rail transit network at multiple historical statistical moments, so as to reflect the spatio-temporal characteristics of the change in the number of passengers at the station.

[0083] The situation of stranded passengers at any station at any historical statistical moment is used to indicate whether there are stranded passengers at that historical statistical moment and the number of corresponding stranded passengers. Its specific content can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0084] Among them, the specific form of the passenger flow feature matrix can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0085] In an example, the target rail transit network can be defined as a directed graph G=(V, E, A), where V represents nodes, that is, stations in the target rail transit network, and the number of nodes N, that is, the number of stations in the target rail transit network; E represents the set of edges of the directed graph, and the edge represents the connection relationship between nodes, that is, the line section between stations. For the train running routes in the upward direction and the downward direction, there are different line sections, which can be respectively represented as the in-edge and out-edge of the node; A represents the adjacency matrix. For the target rail transit network including N stations, the dimension of the adjacency matrix is N×N, and the element a i,j represents the connection relationship between station i and station j. If station j is the downstream station of station i, then a i,j =1. If station j is the upstream station of station i, then a i,j =0.

[0086] On this basis, the passenger flow feature matrix of any station at the current statistical moment can be expressed as formula (5):

[0087] X=(X t , X t-1 ,…X t-τ )∈R N×M×τ (5) where τ represents the number of steps in the first step; X t ∈R N×M represents the characteristics of N stations at time t. N is the number of stations in the target rail transit network, and M is the number of input features. Its specific value can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0088] In an example, for the problem of passenger flow trend deduction, the value of M is usually set to 2, that is, X t includes the number of stranded passengers Stay t ∈R N at each station at time t, and the time code Time at each station at time tt ∈R N For two features, since the number of stranded passengers usually has periodic characteristics. For example, it usually shows the periodicity of morning and evening rush hours on weekdays. Time coding can be used to better capture such periodic characteristics. Therefore, X t can be expressed as formula (6):

[0089] X t =(Stay t , Time t ) (6)

[0090] For any station in the target rail transit network, the passenger flow feature matrix of the station at the current statistical moment can be input into the passenger flow situation deduction network for multi-step deduction to determine the passenger stranded situation of the station at the target statistical moment, so as to indicate whether there are stranded passengers at the station at the target statistical moment and the corresponding number of stranded passengers, and ensure its high accuracy and reliability. The specific content of the passenger stranded situation can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations thereon.

[0091] Subsequently, in combination with possible implementation manners of the present disclosure, the process of using the passenger flow situation deduction network to perform multi-step deduction on the passenger flow feature matrix of any station in the target rail transit network at the current statistical moment to determine the passenger stranded situation of the station at the target statistical moment will be described in detail, and will not be elaborated here.

[0092] In step S103, according to the train delay situation and passenger stranded situation of each station in the target rail transit network at the target statistical moment, a transportation risk assessment is performed to determine the global transport capacity risk of the target rail transit network at the target statistical moment, where the global transport capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transport demand at the target statistical moment.

[0093] The existing methods for urban rail transit transportation situation deduction usually use the regional rail transit transportation situation assessment system to perform an overall deduction on the current passenger flow situation and current train flow situation of the target rail transit network, and directly obtain the future transportation situation of the target rail transit network. This overall deduction method is relatively rough and pays insufficient attention to the details that change in the transportation situation, resulting in poor prediction accuracy.

[0094] Therefore, the transportation situation deduction method for rail transit provided by the present disclosure first conducts separate deductions on the train flow situation and the passenger flow situation through the foregoing process to improve the attention to the details of the transportation situation, and then conducts a transportation risk assessment based on the train delay situation and the passenger retention situation of each station in the target rail transit network at the target statistical moment, determines the global transport capacity risk of the target rail transit network at the target statistical moment, and realizes the deduction of the transportation situation of the urban rail transit network, which has high accuracy and reliability.

[0095] Specifically, the distribution characteristics of high-load passenger flow are mainly reflected in the number of passengers staying at the station, while the train flow characteristics are mainly reflected in the train departure interval. Therefore, the global transport capacity risk of the target rail transit network under the condition of high-load passenger flow is mainly affected by the number of passengers staying at each station and the train departure interval time. Among them, when the train operation plan is determined, the train departure interval time can be expressed as a function of the train delay situation. To sum up, the global transport capacity risk of the target rail transit network at the target statistical moment can be expressed by formula (7):

[0096] RN(t) = F(Stay(t), D(t)) (7)

[0097] Wherein, RN(t) represents the global transport capacity risk of the target rail transit network at the target statistical moment; Stay(t) represents the passenger retention situation of each station in the target rail transit network at the target statistical moment; D(t) represents the train delay situation of each station in the target rail transit network at the target statistical moment; F(·) represents the function that maps the passenger retention situation and the train delay situation to the global transport capacity risk, that is, the road network transportation situation evaluation system constructed with the passenger retention situation and the train delay situation as the core. Its specific form can refer to the implementation methods in related technologies, and the present disclosure does not make specific limitations on this.

[0098] Figure 5 Fig. shows a schematic diagram of a road network transportation situation evaluation system according to the prior art. As Figure 5 shown, the global transport capacity risk of the target rail transit network specifically includes the transport capacity risks corresponding to each train running line in the target rail transit network. The transport capacity risk corresponding to any one train running line includes: the passenger saturation risk of each station on this train running line, the passenger retention risk on the upper platform of each station, the passenger retention risk on the lower platform of each station, the passenger flow saturation risk in the upper section of each station, and the passenger flow saturation risk in the lower section of each station.

[0099] Based on the above-mentioned road network transportation situation evaluation system, the train delay situation and passenger detention situation of each station in the target rail transit network at the target statistical moment can be used to conduct transportation risk assessment and determine the global transport capacity risk of the target rail transit network at the target statistical moment. Among them, for the specific methods of calculating each risk in this road network transportation situation evaluation system, reference can be made to the implementation manners in related technologies, and the present disclosure does not make specific limitations thereon.

[0100] The process of conducting transportation risk assessment based on the train delay situation and passenger detention situation of each station in the target rail transit network at the target statistical moment to determine the global transport capacity risk of the target rail transit network at the target statistical moment will be described in detail later in combination with possible implementation manners of the present disclosure, and will not be elaborated here.

[0101] In the embodiment of the present disclosure, for any station in the target rail transit network, a train flow situation deduction network can be used to perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment, so as to fully consider the influence of train delay on the uncertainty of train flow during the operation of urban rail transit, and through multi-step deduction, improve the short-term dependence of train flow situation deduction, increase the adaptability to train delay, and improve the accuracy and reliability of train flow situation deduction; among them, the train flow feature matrix of any station at the current statistical moment includes the train delay situations of multiple stations before the station in multiple historical statistical moments before the current statistical moment in any train running line passing through the station. For any station in the target rail transit network, a passenger flow situation deduction network can be used to perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger detention situation of the station at the target statistical moment. By separately deducing the train flow situation and the passenger flow situation, the detailed attention to the transportation situation can be improved, and the advantages of multi-step deduction can also be effectively utilized during the passenger flow situation deduction process to improve the accuracy and reliability of passenger flow situation deduction; among them, the passenger flow feature matrix of any station at the current statistical moment includes: the detained passenger situations of each station in the target rail transit network in multiple historical statistical moments. After completing the separate deductions of the train flow situation and the passenger flow situation, the transportation risk assessment can be conducted based on the train delay situation and passenger detention situation of each station in the target rail transit network at the target statistical moment to determine the global transport capacity risk of the target rail transit network at the target statistical moment, so as to indicate whether the target rail transit network can meet the passenger transport demand at the target statistical moment, realize a complete transportation situation deduction process, and ensure that the global transport capacity risk of the target rail transit network at the target statistical moment has high prediction accuracy and stability.

[0102] In a possible implementation, the train flow situation deduction network is a recurrent neural network, including a long short-term memory network layer and a fully connected layer; for any station in the target rail transit network, using the train flow situation deduction network, perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment, including: initializing the long short-term memory network layer to determine the initial hidden state corresponding to the long short-term memory network layer; using the long short-term memory network layer to perform iterative processing on the train flow feature matrix corresponding to the station and update the initial hidden state until the number of iterations is equal to the preset number of times to determine the target hidden state corresponding to the long short-term memory network layer; determine the train delay situation of the station at the target statistical moment according to the conversion of the fully connected layer to the target hidden state.

[0103] The train flow situation deduction problem can be transformed into a time series prediction problem. Therefore, the train flow situation deduction network can be set as a recurrent neural network to predict the train delay situation of any station at the target statistical moment. Among them, the recurrent neural network includes a long short-term memory (LSTM) network and a fully connected layer.

[0104] The specific number of LTSMs can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this. It should be noted that by stacking multiple LTSMs, the learning ability of the recurrent neural network for complex time series data (that is, the train flow feature matrix of any station at the current statistical moment) can be improved. However, when the number of LSTMs is too large, it may cause overfitting of the recurrent neural network. Preferably, the number of LSTMs can be set to 2 layers.

[0105] For the train flow situation deduction problem of predicting the train delay situation of any station at the target statistical moment, it not only depends on the train delay situations of the station itself at multiple historical statistical moments, but also depends on the train delay situations of the previous stations on the train running line where the station is located at multiple historical statistical moments.

[0106] Therefore, through the recurrent neural network including LSTM, the three gating units of the input gate, forget gate, and output gate in the LSTM can be used to realize the memory of long-term dependence relationships and avoid the problem of gradient disappearance, so that the predicted train delay situation of any station at the target statistical moment has high accuracy and reliability.

[0107] Specifically, before conducting the column flow situation deduction, it is necessary to initialize the LSTM and set the initial state for the LSTM, which specifically includes the initial hidden state h0 and the initial cell state C0. Among them, the specific content of the initial hidden state h0 and the initial cell state C0 can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this. Preferably, both the initial hidden state h0 and the initial cell state C0 are initialized as zero vectors.

[0108] After the initialization is completed, the LSTM can be used to iteratively process the column flow feature matrix corresponding to the station according to the preset time steps. The specific process of iteratively processing the column flow feature matrix corresponding to any station can refer to the description of the above multi-step deduction, and will not be elaborated here.

[0109] In each iteration process, the initial hidden state h t and the initial cell state C0 are also updated to obtain the updated hidden state h t and the cell state C t . Specifically, the LSTM can receive the column flow feature matrix corresponding to the station input in the current iteration process, as well as the hidden state h t-1 of the previous iteration process, and based on the gating mechanism, respectively determine the update values of the input gate, forget gate, and output gate; according to the update values of the forget gate and the input gate, obtain the cell state C t of the current iteration process; then, in combination with the update value of the output gate and the cell state C t , calculate the hidden state h t of the current iteration process as the input for the next iteration. Through the above update process, information that has little impact on the train delay situation of the station at the target time, such as the train delay situations of previous stations far away from the station at multiple historical statistical moments, will be forgotten in the forget gate of the LSTM, while newly input information, such as the impact of the determined weather change on the train delay situation of the station at the target time during the iteration process, can be saved in the cell state C t through the input gate of the LSTM, increasing the short-term dependence of the column flow situation deduction.

[0110] After the number of iterations meets the preset number, the hidden state obtained in the last iteration can be determined as the target hidden state; and according to the fully connected layer, the target hidden state is transformed to determine the train delay situation of the station at the target statistical moment. Among them, the specific value of the preset number depends on the time interval between the target statistical moment and the current statistical moment, and the quantitative relationship with the preset time steps can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0111] In one example, when the preset time step is 5 minutes and the target statistical moment is the moment 15 minutes after the current statistical moment, the preset number of times can be determined to be 3.

[0112] Figure 6 The figure shows a schematic diagram of the prediction effect of a column flow trend deduction network according to an embodiment of the present disclosure. As Figure 6 shown, in the column flow feature matrix corresponding to any station at the current statistical moment, when the number k of the previous stations is 3 and the number τ of the first-step progressions is 6, the train delay situation of this station at the target statistical moment is predicted by the recurrent neural network including at least one LSTM and one fully connected layer provided by the embodiment of the present disclosure. The vertical axis represents the predicted train delay situation, expressed by the delay duration; the horizontal axis represents the actual train delay situation, also expressed by the delay duration; any point represents the distribution law of the predicted train delay situation relative to the actual train delay situation.

[0113] As Figure 6 shown, most points are distributed along the 45° diagonal, indicating that the deviation of the predicted train delay situation relative to the actual train delay situation is small, the predicted train delay situation has high accuracy, and moreover, as the delay duration increases, the deviation of the predicted train delay situation still remains in a small fluctuation state, indicating that the prediction performance of the recurrent neural network is high and stable.

[0114] In a possible implementation manner, for any station in the target rail transit network, using the passenger flow trend deduction network to perform multi-step deduction on the passenger flow feature matrix of this station at the current statistical moment to determine the passenger retention situation of this station at the target statistical moment, including: according to the passenger flow trend deduction network, performing single-step deduction on the passenger flow feature matrix of this station at the current statistical moment to determine the passenger retention situation of this station at the next statistical moment after the current statistical moment and before the target statistical moment; based on the sliding window method, updating the passenger flow feature matrix of this station at the current statistical moment according to the passenger retention situation of this station at the next statistical moment to determine the updated passenger flow feature matrix; repeating the above steps according to the updated passenger flow feature matrix until the number of single-step deductions is equal to the preset number of times to determine the passenger retention situation of this station at the target statistical moment.

[0115] According to the passenger flow trend deduction network, single-step deduction can be performed on the passenger flow feature matrix of this station at the current statistical moment to determine the passenger retention situation of this station at the next statistical moment after the current statistical moment and before the target statistical moment. Here, the time interval between the next statistical moment and the current statistical moment is equal to the preset time step.

[0116] The specific method for implementing single-step deduction using the passenger flow situation deduction network can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations thereto. The specific form of the passenger flow situation deduction network can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations thereto.

[0117] In a possible implementation manner, the passenger flow situation deduction network is a directed graph spatio-temporal convolutional network based on multi-head attention, including: a spatial multi-head attention module, a temporal multi-head attention module, a directed graph spatio-temporal convolutional module, and a fully connected layer.

[0118] Generally, the passenger flow distribution of each station in the target rail transit network is a network structure. Therefore, when deducing the passenger flow situation of any station in the target rail transit network, not only the temporal correlation between different historical statistical moments is required, but also the spatial correlation between different stations needs to be captured. For this reason, the spatio-temporal attention graph convolutional network (ASTGCN) is usually used in the prior art to realize the passenger flow situation deduction in the urban rail transit network. However, ASTGCN represents the urban rail transit network as an undirected graph, ignoring the directionality of passenger flow, which has upward and downward directions.

[0119] In view of this, the transportation situation deduction method for rail transit provided in the embodiments of the present disclosure sets the passenger flow situation deduction network as a directed graph spatio-temporal convolutional network based on multi-head attention (Multi-Head Attention Based Spatial-Temporal Directed Graph Convolutional Networks, MA-STDGCN), which can make full use of the directed graph to learn the directionality of the passenger flow in the target rail transit network and can capture more information by using the multi-head attention mechanism.

[0120] Figure 7 Fig. shows a schematic structural diagram of a passenger flow situation deduction network according to an embodiment of the present disclosure. As Figure 7 shown, the directed graph spatio-temporal convolutional network based on multi-head attention includes: a spatial multi-head attention module, a temporal multi-head attention module, a directed graph spatio-temporal convolutional module, and a fully connected layer.

[0121] After the passenger flow feature matrix of any station at the current statistical moment is input into the passenger flow situation deduction network, corresponding output structures are obtained through the spatial multi-head attention module and the temporal multi-head attention module respectively. The respective output results of the spatial multi-head attention module and the temporal multi-head attention module are input into the directed graph spatio-temporal convolutional module after being weighted by the spatio-temporal attention matrix, and then the passenger retention situation of the station at the next statistical moment is obtained through mapping and output by the fully connected layer.

[0122] Among them, the spatial multi-head attention module is used to extract the spatial correlation between different stations in the passenger flow feature matrix of the station at the current statistical moment, and can capture the long and short distance dependencies between stations through the multi-head attention mechanism; the temporal multi-head attention model is used to extract the temporal correlation between different historical statistical moments in the passenger flow feature matrix of the station at the current statistical moment, and can obtain multi-scale dependencies through the multi-head attention mechanism.

[0123] For the specific implementation methods of the spatial multi-head attention module, the temporal multi-head attention module, and the directed graph spatio-temporal convolution module, reference can be made to the implementation manners in the related technologies, and the present disclosure does not make specific limitations thereon.

[0124] In an example, the spatial multi-head attention module includes H1 attention heads, and the feature dimension corresponding to any one attention head is d h = M / H1, where M is the number of input features.

[0125] The passenger flow feature matrix X ∈ R of any station at the current statistical moment N×M×τ , is first divided according to the batch size B to obtain the input X ∈ R of the spatial multi-head attention module B×N×M×τ ; then it is divided according to the number of attention heads H1 to obtain the input of each attention head For the subspace divided by any one attention head, the calculation of attention can be performed, and the specific process can be expressed by formulas (8) to (10):

[0126]

[0127] Among them, and both represent learnable weight matrices for fusing information in different dimensions of the input; (lhs h ) b,n,t represents the left end term of the attention matrix; (rhs h ) b,n,t represents the right end term of the attention matrix; σ represents the sigmoid activation function; S h represents the station relationship matrix learned by one attention head.

[0128] By aggregating the station relationship matrices learned by all attention heads, the spatial attention matrix corresponding to the passenger flow feature matrix of the station at the current statistical moment can be obtained. The spatial attention matrix can be expressed by formula (11):

[0129]

[0130] Among them, S ∈ R B×N×NRepresents the spatial attention matrix, where each element in N×N represents the association relationship between station i and station j.

[0131] Any element S in the spatial attention matrix S i,j , can represent the correlation strength between station i and station j, and the sum of all attention scores of other stations to station i should be 1, that is, ∑ j∈N S i，j = 1. The spatial attention matrix S can be normalized by the softmax function, and the processing result can be expressed as formula (12):

[0132]

[0133] where S′ i，j represents the correlation strength after normalization; N represents the number of stations.

[0134] Performing the above processing using the spatial multi-head attention module can obtain the correlation between station nodes. Compared with the single-head attention mechanism, the spatial multi-head attention module can respectively focus on the relationship between short-distance stations and long-distance stations in the target rail transit network and capture different sub-spaces.

[0135] In an example, the temporal multi-head attention module includes H2 attention heads, and the feature dimension corresponding to any one attention head is d h = M / H2, where M is the number of input features.

[0136] The passenger flow feature matrix X ∈ R of any station at the current statistical moment N×M×τ , is first divided according to the batch size B to obtain the input X ∈ R of the spatial multi-head attention module B×N×M×τ ; then it is divided according to the number of attention heads H2 to obtain the input X ∈ R of each attention head B×N×H2×M×τ . For the sub-space divided by any one attention head, the calculation of attention can be performed, and the specific process can be expressed as formulas (13) to (15):

[0137]

[0138] where and both represent learnable weight matrices for fusing information in different dimensions of the input; σ represents the sigmoid activation function; E h represents the historical statistical moment relationship matrix learned by one attention head.

[0139] Aggregating the historical statistical moment relationship matrices learned by all attention heads, the time attention matrix corresponding to the passenger flow feature matrix of the station at the current statistical moment can be obtained. The time attention matrix can be expressed as formula (16):

[0140]

[0141] where, E ∈ R B×T×T represents the time attention matrix, and each element in T×T represents the historical statistical moment t i and the historical statistical moment t j correlation relationship.

[0142] Any element E i,j in the time attention matrix E can represent the correlation strength between the historical statistical moment t i and the historical statistical moment t j . Through the softmax function, the time attention matrix E can be normalized, and the processing result can be expressed as formula (17):

[0143]

[0144] where, E′ i，j represents the correlation strength after normalization; T represents the total number of historical statistical moments.

[0145] Using the spatial multi-head attention module for the above processing, the correlation between the passenger retention situation of any station at the current statistical moment and the passenger retention situations of the station at multiple historical statistical moments can be obtained, and the long-term and short-term correlations can be respectively focused on.

[0146] In an example, the directed graph spatio-temporal convolution module realizes graph convolution by using a linear operator diagonalized in the Fourier domain to replace the classical convolution operator, and introduces the Chebyshev polynomial for the directed graph to perform eigenvalue decomposition on the Laplacian matrix.

[0147] Specifically, for any directed graph G, the signal x in the directed graph G can be filtered by the kernel g θ , and the filtering process can be expressed as formula (18):

[0148] g θ *Gx = g θ (L)x = g θ (UΛU T )X = Ug θ (Λ)U T x (18)

[0149] Where *G represents the graph convolution operation; L represents the Laplacian matrix corresponding to the directed graph G, which can be decomposed into L = UΛU T , Λ=diag([λ0,…λ n-1 ])∈R N×M is a diagonal matrix, and U is the Fourier basis.

[0150] When the scale of the directed graph G is large, it takes a long time to directly perform eigenvalue decomposition on the Laplace matrix. Therefore, Chebyshev polynomials are usually used to approximate the solution. The recursive definition of Chebyshev polynomials can be expressed as formula (19):

[0151] T k (x) = 2xT k-1 (x)-T k-2 (x) (19)

[0152] Among them, T k Represents the information of the k-order neighborhood around any node in the directed graph G.

[0153] However, the Chebyshev polynomials assume by default that the neighborhood relationship is symmetric. In the actual urban rail transit network, the flow of up and down passengers has obvious directionality. Directly approximating the adjacency matrix A of the directed graph G as an undirected graph will lead to the loss of directionality, making the passenger flow situation deduction results inaccurate.

[0154] In order to solve the above problem, the embodiment of the present disclosure introduces Chebyshev polynomials for directed graphs. According to the incoming and outgoing edges of each node in the directed graph G, the adjacency matrix A of the directed graph G can be decomposed into the incoming edge adjacency matrix A in and the outgoing edge adjacency matrix A out , and the adjacency matrix A, the inbound edge adjacency matrix A in and the outgoing edge adjacency matrix A out The dimensions of are 187 × 187. The input edge adjacency matrix A in and the outgoing edge adjacency matrix A out They can be expressed as formulas (20) and (21) respectively:

[0155]

[0156] For the incoming edge adjacency matrix A in and the outgoing edge adjacency matrix A out Calculate the normalized directed Laplacian matrix, which can be expressed as formulas (22) and (23):

[0157]

[0158] Among them, D in Represents the incoming edge adjacency matrix A in degree; Dout Represent the out-edge adjacency matrix A out degrees.

[0159] Through the above decomposition method, the in-edges and out-edges of each node in the directed graph G can be modeled respectively, so as to better capture the directional information of the up-and-down passenger flow.

[0160] Furthermore, for the directed Laplacian matrix and Chebyshev polynomial expansions are performed respectively, and the results are fused to determine the k-order neighborhood information corresponding to the directed graph G. The k-order neighborhood information of any node in the directed graph G can be expressed by formula (24):

[0161]

[0162] where, represents the k-order neighborhood information of any node in the directed graph G; represents the k-order in-neighborhood information of any node in the directed graph G; represents the k-order out-neighborhood information of any node in the directed graph G.

[0163] By accumulating the information of each node itself and the information of the 1st to K-1th order neighborhoods around the node, the output result of the graph convolution can be obtained, which can be expressed by formula (25):

[0164]

[0165] On this basis, using the result of the spatial attention matrix to adjust the correlation between nodes, calculate the Hadamard product of the k-order neighborhood information of any node in the directed graph G and the spatial attention matrix S′ i，j Then the output result of the graph convolution can be expressed by formula (26):

[0166]

[0167] Furthermore, after the graph convolution operation has captured the neighborhood information of each node in the directed graph G in the spatial dimension, the directed graph spatio-temporal convolution module can further stack standard convolution layers in the temporal dimension to update the node signals by merging information at adjacent time slices.

[0168] Using the directed graph spatio-temporal convolution module for the above processing can fully capture the temporal and spatial features in the column flow feature matrix and passenger flow feature matrix of each station at the current statistical moment.

[0169] The specific quantities of the spatial multi-head attention module, the temporal multi-head attention module, and the directed graph spatio-temporal convolution module can be flexibly set according to actual usage requirements. For example, a spatio-temporal block can be formed by one spatial multi-head attention module, one temporal multi-head attention module, and one directed graph spatio-temporal convolution module, and multiple spatio-temporal blocks can be stacked to further extract dynamic spatio-temporal correlations in a larger range. The present disclosure does not make specific limitations on this.

[0170] Finally, the passenger flow situation deduction network can transform the output result of the directed graph spatio-temporal convolution module through a fully connected layer to obtain the passenger retention situation.

[0171] For any station, after completing a single-step deduction using the passenger flow situation deduction network, based on the sliding window method, according to the passenger retention situation of the station at the next statistical moment, the passenger flow feature matrix of the station at the current statistical moment can be updated to determine the updated passenger flow feature matrix; repeat the above steps according to the updated passenger flow feature matrix until the number of single-step deductions is equal to the preset number, and determine the passenger retention situation of the station at the target statistical moment. The iterative process here can refer to the foregoing description of multi-step deductions and will not be elaborated here.

[0172] Among them, the specific value of the preset number depends on the time interval between the target statistical moment and the current statistical moment, and the quantitative relationship with the preset time step can be flexibly set according to actual usage requirements. The present disclosure does not make specific limitations on this.

[0173] In an example, when the preset time step is 5 minutes and the target statistical moment is the moment 15 minutes after the current statistical moment, the preset number can be determined to be 3.

[0174] In an example, in the passenger flow feature matrix corresponding to any station at the current statistical moment, the number of previous stations k = 3, and the number of first-step entries τ = 6; the random forest model (RF), LSTM, the LSTM model combined with the graph convolutional neural network (GCN), the spatio-temporal graph convolutional network (STGCN), ASTGCN, and the passenger flow situation deduction network in the form of MA-STDGCN provided by the embodiments of the present disclosure are respectively used for passenger flow situation deduction to predict the passenger retention situation of the station at the target statistical moment, and the prediction accuracy is evaluated through three indicators: RMSE, MAE, and WMAPE.

[0175] Referring to Table 2, as shown in Table 2, the directed graph spatio-temporal convolution network based on multi-head attention provided by the embodiments of the present disclosure performs better than other network models in all three indicators, and has high accuracy and stability.

[0176] Table 2

[0177]

[0178] In a possible implementation, the global operation capacity risk of the target rail transit network at the target statistical moment includes: the station operation capacity risk of each station in the target rail transit network at the target statistical moment, and the section operation capacity risk of any section formed by any two stations in the target rail transit network at the target statistical moment; according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment, a transportation risk assessment is carried out to determine the global operation capacity risk of the target rail transit network at the target statistical moment, including: for any station, determining the station operation capacity risk of this station at the target statistical moment according to the train delay situation and passenger retention situation of this station at the target statistical moment; for any section, determining the section operation capacity risk of this section at the target statistical moment according to the train delay situations of the two stations corresponding to this section at the target statistical moment.

[0179] The global operation capacity risk of the target rail transit network at the target statistical moment can be expressed as the cumulative result of the station operation capacity risk of each station in the target rail transit network at the target statistical moment and the section operation capacity risk of any section formed by any two stations in the target rail transit network at the target statistical moment. Specifically, the global operation capacity risk of the target rail transit network at the target statistical moment can be expressed by formula (27):

[0180] RN(t) = ∑ i∈N RS i (t) + ∑ j∈N RI j (t) (27)

[0181] Wherein, RN(t) represents the global operation capacity risk of the target rail transit network at the target statistical moment; RS i (t) represents the station operation capacity risk of any station i in the target rail transit network at the target statistical moment; RI j (t) represents the section operation capacity risk of any section j formed by any two stations in the target rail transit network at the target statistical moment.

[0182] On this basis, independent analysis can be carried out for each station in the dimension of station operation capacity, and the station operation capacity risk RS i (t) of this station at the target statistical moment is determined according to the train delay situation and passenger retention situation of this station at the target statistical moment. Among them, for the specific method of determining the station operation capacity risk RS i (t) of any station at the target statistical moment, reference can be made to the implementation manners in the related technologies, and the present disclosure does not make specific limitations thereto.

[0183] In a possible implementation, the station operation capacity risk of any station at the target statistical moment includes: the station passenger flow saturation risk and the platform passenger retention risk of the station at the target statistical moment; for any station, according to the train delay situation and passenger retention situation of the station at the target statistical moment, determine the station operation capacity risk of the station at the target statistical moment, including: determining the in-station passenger flow of the station at the target statistical moment; according to the train delay situation of the station at the target statistical moment, determine the passenger throughput and platform evacuation capacity of the station at the target statistical moment; according to the in-station passenger flow and passenger throughput of the station at the target statistical moment, and the corresponding passenger flow saturation risk consequence of the station, determine the station passenger flow saturation risk of the station at the target statistical moment; according to the passenger retention situation and platform evacuation capacity of the station at the target statistical moment, and the corresponding passenger retention risk consequence of the station, determine the platform passenger retention risk of the station at the target statistical moment.

[0184] Specifically, the station operation capacity risk of any station at the target statistical moment can be composed of two parts: the station passenger flow saturation risk and the platform passenger retention risk of the station at the target statistical moment, and can be specifically expressed as formula (28):

[0185] RS i (t) = RS is (t) + RS iw (t) (28)

[0186] Wherein, RS is (t) represents the station passenger flow saturation risk of any station at the target statistical moment; RS iw (t) represents the platform passenger retention risk of the station at the target statistical moment.

[0187] The station passenger flow saturation risk of any station at the target statistical moment can reflect the risk value that the passenger flow of the station reaches the station saturation at the target statistical moment. Its specific form and determination method can refer to the implementation in the related technology, and the present disclosure does not make specific limitations thereto.

[0188] In an example, based on the risk assessment principle, the risk value can be expressed as the product of the risk probability and the risk consequence. Therefore, the station saturation risk of any station at the target statistical moment can be determined according to the station saturation risk probability of the station at the target statistical moment and the corresponding passenger flow saturation risk consequence of the station.

[0189] The station saturation risk probability of any station at the target statistical moment can be expressed as a function of the station saturation of the station at the target statistical moment; while the station saturation of any station at the target statistical moment can be expressed by the ratio of the in-station passenger flow and the passenger throughput of the station at the target statistical moment. Among them, the specific content of the passenger flow saturation risk consequence corresponding to any station can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations thereon. The specific method for determining the in-station passenger flow of any station at the target statistical moment can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations thereon.

[0190] Therefore, for any station, the passenger throughput of the station at the target statistical moment can be determined according to the train delay situation of the station at the target statistical moment; and then, according to the in-station passenger flow and the passenger throughput of the station at the target statistical moment, and the passenger flow saturation risk consequence corresponding to the station, the station passenger flow saturation risk at the target statistical moment can be determined. Among them, the specific method for determining the passenger throughput of the station at the target statistical moment according to the train delay situation of the station at the target statistical moment can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations thereon.

[0191] Based on the above principle, the station saturation risk of any station at the target statistical moment can be expressed as formula (29):

[0192]

[0193] where, w i1 (t) represents the passenger flow saturation risk consequence corresponding to any station; flow i (t) represents the in-station passenger flow of the station at the target statistical moment; c i (t) represents the passenger throughput of the station at the target statistical moment; represents the station saturation of the station at the target statistical moment; f(·) represents the risk probability function, and its specific form can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations thereon.

[0194] In an example, the risk probability function can be expressed as formula (30):

[0195]

[0196] where, x represents the input of the risk probability function.

[0197] The platform passenger retention risk of any station at the target statistical moment can reflect the risk value that the number of passengers retained on each platform in the station reaches the upper limit of the platform evacuation capacity. Its specific form and determination method can refer to the implementation methods in related technologies, and the present disclosure does not make specific limitations on this.

[0198] Taking the above example based on the risk assessment principle that the risk value is expressed as the product of the risk probability and the risk consequence, the platform passenger retention risk of any station at the target statistical moment can be determined according to the passenger retention risk probability of the station at the target statistical moment and the passenger retention risk consequence corresponding to the station. Among them, the passenger retention risk probability of any station at the target statistical moment can be expressed as a function of the passenger evacuation time of the station at the target statistical moment; and the passenger evacuation time of any station at the target statistical moment can be expressed as the ratio of the passenger retention situation of the station at the target statistical moment to the platform evacuation capacity. The specific content of the passenger retention risk consequence corresponding to any station can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations on this.

[0199] Therefore, for any station, the platform evacuation capacity of the station at the target statistical moment can be determined according to the train delay situation of the station at the target statistical moment; and then, based on the passenger retention situation and the platform evacuation capacity of the station at the target statistical moment, and the passenger retention risk consequence corresponding to the station, the platform passenger retention risk of the station at the target statistical moment can be determined. Among them, the specific method for determining the platform evacuation capacity of the station at the target statistical moment according to the train delay situation of the station at the target statistical moment can refer to the implementation methods in related technologies, and the present disclosure does not make specific limitations on this.

[0200] Based on the above principle, the platform passenger retention risk of any station at the target statistical moment can be expressed as formula (31):

[0201]

[0202] where, w i2 (t) represents the passenger retention risk consequence corresponding to any station; stay i (t) represents the passenger retention situation of the station at the target statistical moment; z i (t) represents the platform evacuation capacity of the station at the target statistical moment; represents the passenger evacuation time of the station at the target statistical moment; f(·) represents the risk probability function.

[0203] Furthermore, the interval operation capacity risk RI of the interval composed of any two stations at the target statistical moment can also be determined according to the train delay situation and the passenger retention situation of the two stations at the target statistical moment.i (t). Among them, any two of these stations can be flexibly selected according to actual usage requirements, and it should be ensured that the two stations are on the same train running line. The present disclosure does not make specific limitations in this regard; preferably, any two stations are two adjacent stations on the same train running line. Determine the station operation capacity risk RI of any one section at the target statistical moment i (t). The specific method can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations in this regard.

[0204] In a possible implementation manner, for any one section, according to the train delay conditions of the two stations corresponding to this section at the target statistical moment, determine the section operation capacity risk of this section at the target statistical moment, including: determining the section passenger flow volume of this section at the target statistical moment; according to the train delay conditions of the two stations corresponding to this section at the target statistical moment, determine the section transportation capacity of this section at the target statistical moment; according to the section passenger flow volume and section transportation capacity of this section at the target statistical moment, and the passenger flow saturation risk consequence corresponding to this section, determine the section operation capacity risk of this section at the target statistical moment.

[0205] Among them, the specific method for determining the section passenger flow volume of any one section at the target statistical moment can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations in this regard.

[0206] Generally, the section transportation capacity of any one section at the target statistical moment can be expressed as a mapping function of the train delay conditions of the two stations corresponding to this section at the target statistical moment. Therefore, for any one section, according to the train delay conditions of the two stations corresponding to this section at the target statistical moment, the section transportation capacity of this section at the target statistical moment can be determined; the specific determination method can refer to the implementation manners in related technologies, and the present disclosure does not make specific limitations in this regard.

[0207] According to the section passenger flow volume and section transportation capacity of this section at the target statistical moment, and the passenger flow saturation risk consequence corresponding to this section, the section operation capacity risk of this section at the target statistical moment can be determined. Among them, the specific content of the passenger flow saturation risk consequence corresponding to any one section can be flexibly set according to actual usage requirements, and the present disclosure does not make specific limitations in this regard.

[0208] Taking the above example where the risk value is expressed as the product of the risk probability and the risk consequence based on the risk assessment principle, the interval operation risk of any interval at the target statistical moment can be determined according to the interval saturation risk probability of the interval at the target statistical moment and the interval saturation risk consequence corresponding to the interval. The interval saturation risk probability of any interval at the target statistical moment can be expressed as a function of the interval saturation degree of the interval at the target statistical moment; while the interval saturation degree of any interval at the target statistical moment can be expressed by the ratio of the interval passenger flow and the interval transportation capacity at the target statistical moment. To sum up, the interval operation risk of any interval at the target statistical moment can be expressed by formula (32):

[0209]

[0210] Among them, w j (t) represents the passenger flow saturation risk consequence corresponding to any interval; y j (t) represents the interval passenger flow of the interval at the target statistical moment; d j (t) represents the interval transportation capacity of the interval at the target statistical moment.

[0211] Through the above process, the transportation situation deduction method of rail transit provided by the embodiments of the present disclosure splits the global operation risk of the target rail transit network at the target statistical moment into the station operation risk of each station in the target rail transit network and the interval operation risk of each interval at the target statistical moment, constructs a hierarchical global operation risk system, can introduce the uncertainty of train operation, and fully considers the impact of train delays on each station and each interval. Compared with the commonly used overall deduction method in the prior art, it has higher accuracy and reliability.

[0212] Figure 8 Shows a prediction effect diagram of an operation risk according to an embodiment of the present disclosure. As Figure 8 shown, the horizontal axis represents the statistical moment, and the vertical axis represents the regional operation risk of any train running line in the urban rail transit network. The red dots represent the real operation risk of the train running line at each statistical moment, and the blue curve represents the predicted global operation risk of the train running line at each statistical moment.

[0213] As Figure 8 shown, through the transportation situation deduction method of rail transit provided by the embodiments of the present disclosure, the predicted regional operation risk of the train running line at each statistical moment has high accuracy, and the deviation from the real operation risk of the train running line at each statistical moment is small.

[0214] In one example, the commonly used overall deduction methods in the prior art, namely, the AutoRegression (AR) method, the Support Vector Regression (SVR) method, and the LSTM method, and the transportation situation deduction method of the present disclosure are respectively used to deduce the transportation situation of the target rail transit network and the train running lines 1 to 10 included in the target rail transit network, determine the global transport capacity risk of the target rail transit network at the target statistical moment, and the regional transport capacity risks of the train running lines 1 to 10 at the target statistical moment; and the deduction accuracy of each method is evaluated by the WMAPE.

[0215] Referring to Table 3, as shown in Table 3, whether it is the global transport capacity risk of the target rail transit network at the target statistical moment or the regional transport capacity risk of any one train running line at the target statistical moment, the prediction accuracy of the transportation situation deduction method of the present disclosure is higher than that of the three overall deduction methods; moreover, the error of the global transport capacity risk of the target rail transit network at the target statistical moment is only 3.2%, with relatively high accuracy.

[0216] Table 3

[0217]

[0218]

[0219] In the embodiments of the present disclosure, for any station in the target rail transit network, the train flow situation deduction network can be used to perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment, so as to determine the train delay situation of the station at the target statistical moment after the current statistical moment, thereby fully considering the impact of train delays on the uncertainty of train flow during the operation of urban rail transit, and through multi-step deduction, improving the short-term dependence of train flow situation deduction, increasing the adaptability to train delays, and improving the accuracy and reliability of train flow situation deduction; wherein, the train flow feature matrix of any station at the current statistical moment includes the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train operation line passing through the station. For any station in the target rail transit network, the passenger flow situation deduction network can be used to perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment, so as to determine the passenger retention situation of the station at the target statistical moment. By separately deducing the train flow situation and the passenger flow situation, the detailed attention to the transportation situation can be improved, and the advantages of multi-step deduction can also be effectively utilized during the passenger flow situation deduction process to improve the accuracy and reliability of the passenger flow situation deduction; wherein, the passenger flow feature matrix of any station at the current statistical moment includes: the passenger retention situations of each station in the target rail transit network at multiple historical statistical moments. On the other hand, the passenger flow situation deduction network in the present disclosure adopts a directed graph spatio-temporal convolutional network based on multi-head attention. Compared with the existing passenger flow situation deduction methods, the influence of passenger flow directionality is introduced, which can further improve the accuracy of passenger flow situation deduction. After completing the separate deductions of the train flow situation and the passenger flow situation, according to the train delay situation and the passenger retention situation of each station in the target rail transit network at the target statistical moment, a hierarchical global transport capacity risk system can be adopted to fully consider the impact of train delays on each station and each section, perform transportation risk assessment, and determine the global transport capacity risk of the target rail transit network at the target statistical moment, so as to indicate whether the target rail transit network can meet the passenger demand at the target statistical moment, realize a complete transportation situation deduction process, and ensure that the global transport capacity risk of the target rail transit network at the target statistical moment has high prediction accuracy and stability.

[0220] It can be understood that the above-mentioned various method embodiments mentioned in the present disclosure can be combined with each other to form combined embodiments without violating the principle logic. Due to space limitations, the present disclosure will not elaborate further. Those skilled in the art can understand that in the above methods of the specific implementation manner, the specific execution order of each step should be determined according to its function and possible internal logic.

[0221] In addition, the present disclosure also provides a device for inferring the transportation situation of rail transit, an electronic device, and a non-volatile storage medium. Any of the above can be used to implement any of the methods for inferring the transportation situation of rail transit provided by the present disclosure. The corresponding technical solutions and descriptions are referred to the corresponding records in the method section and will not be repeated here.

[0222] Figure 9 The block diagram of a device for inferring the transportation situation of rail transit according to an embodiment of the present disclosure is shown. As shown in the figure, the device 900 includes:

[0223] A train flow situation inference module 901, configured to, for any station in the target rail transit network, use the train flow situation inference network to perform multi-step inference on the train flow feature matrix of the station at the current statistical moment, and determine the train delay situation of the station at the target statistical moment after the current statistical moment. Among them, the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station;

[0224] A passenger flow situation inference module 902, configured to, for any station in the target rail transit network, use the passenger flow situation inference network to perform multi-step inference on the passenger flow feature matrix of the station at the current statistical moment, and determine the passenger retention situation of the station at the target statistical moment. Among them, the passenger flow feature matrix of any station at the current statistical moment includes: the retention passenger situations of each station in the target rail transit network at multiple historical statistical moments;

[0225] A transportation risk assessment module 903, configured to perform transportation risk assessment according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment, and determine the global transportation capacity risk of the target rail transit network at the target statistical moment. Among them, the global transportation capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transportation demand at the target statistical moment.

[0226] In a possible implementation manner, the train flow situation inference network is a recurrent neural network, including a long short-term memory network layer and a fully connected layer; the train flow situation inference module 901 is specifically configured to: initialize the long short-term memory network layer to determine the initial hidden state corresponding to the long short-term memory network layer; use the long short-term memory network layer to perform iterative processing on the train flow feature matrix corresponding to the station, and update the initial hidden state until the number of iterations is equal to the preset number of times to determine the target hidden state corresponding to the long short-term memory network layer; and determine the train delay situation of the station at the target statistical moment according to the conversion of the target hidden state by the fully connected layer.

[0227] In a possible implementation, the passenger flow situation deduction module 902 is specifically configured to: according to the passenger flow situation deduction network, perform a single-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation at the next statistical moment after the current statistical moment and before the target statistical moment of the station; based on the sliding window method, update the passenger flow feature matrix of the station at the current statistical moment according to the passenger retention situation of the station at the next statistical moment to determine the updated passenger flow feature matrix; repeat the above steps according to the updated passenger flow feature matrix until the number of single-step deductions is equal to the preset number, and determine the passenger retention situation of the station at the target statistical moment.

[0228] In a possible implementation, the passenger flow situation deduction network is a directed graph spatio-temporal convolutional network based on multi-head attention, including: a spatial multi-head attention module, a temporal multi-head attention module, a directed graph spatio-temporal convolutional module, and a fully connected layer.

[0229] In a possible implementation, the global operation capacity risk of the target rail transit network at the target statistical moment includes: the station operation capacity risk of each station in the target rail transit network at the target statistical moment, and the section operation capacity risk of any section formed by any two stations in the target rail transit network at the target statistical moment; the transportation risk assessment module 903 is specifically configured to: for any station, determine the station operation capacity risk of the station at the target statistical moment according to the train delay situation and the passenger retention situation of the station at the target statistical moment; for any section, determine the section operation capacity risk of the section at the target statistical moment according to the train delay situations of the two stations corresponding to the section at the target statistical moment.

[0230] In a possible implementation, the station operation capacity risk of any station at the target statistical moment includes: the station passenger flow saturation risk and the platform passenger retention risk of the station at the target statistical moment; the transportation risk assessment module 903 is further configured to: determine the in-station passenger flow volume of the station at the target statistical moment; determine the passenger throughput and the platform evacuation capacity of the station at the target statistical moment according to the train delay situation of the station at the target statistical moment; determine the station passenger flow saturation risk of the station at the target statistical moment according to the in-station passenger flow volume and the passenger throughput of the station at the target statistical moment, and the consequences of the passenger flow saturation risk corresponding to the station; determine the platform passenger retention risk of the station at the target statistical moment according to the passenger retention situation and the platform evacuation capacity of the station at the target statistical moment, and the consequences of the passenger retention risk corresponding to the station.

[0231] In a possible implementation, the transportation risk assessment module 903 is further configured to: determine the passenger flow volume of the section at the target statistical moment; determine the transportation capacity of the section at the target statistical moment according to the train delay conditions of the two stations corresponding to the section at the target statistical moment; determine the transportation capacity risk of the section at the target statistical moment according to the passenger flow volume and transportation capacity of the section at the target statistical moment, and the passenger flow saturation risk consequence corresponding to the section.

[0232] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the methods described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be elaborated here.

[0233] The embodiments of the present disclosure further provide an electronic device, including a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the above method.

[0234] The embodiments of the present disclosure further provide a non-volatile computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0235] Figure 10 The block diagram of an electronic device according to an embodiment of the present disclosure is shown. For example, the device 1900 can be provided as a server or a terminal device. Referring to Figure 10 , the device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by a memory 1932 for storing instructions executable by the processing component 1922, such as application programs. The application programs stored in the memory 1932 can include one or more modules each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0236] The device 1900 may further include a power supply component 1926 configured to perform power management of the device 1900, a wired or wireless network interface 1950 configured to connect the device 1900 to a network, and an input / output interface 1958 (I / O interface). The device 1900 can operate based on an operating system stored in the memory 1932, such as Windows Server TM , MacOS X TM , Unix TM , Linux TM , FreeBSD TM or the like.

[0237] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, and the computer program instructions can be executed by a processing component 1922 of the device 1900 to complete the above method.

[0238] A computer-readable storage medium may be a tangible device that can hold and store programs / instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device, such as a punched card or raised structures in a groove having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as an instantaneous signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0239] The computer programs (or computer-readable program instructions) described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include a copper transmission cable, an optical fiber transmission, a wireless transmission, a router, a firewall, a switch, a gateway computer, and / or an edge server. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0240] A computer program (or computer program instructions) for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state-setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer-readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0241] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0242] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that the instructions, when executed by the processor of the computer or other programmable data processing apparatus, create a means for implementing the functions / acts specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, which instructions cause a computer, a programmable data processing apparatus, and / or other devices to operate in a particular manner, such that the computer-readable medium storing the instructions comprises a manufacture, which includes instructions for implementing various aspects of the functions / acts specified in one or more blocks of the flowchart and / or block diagram.

[0243] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, causing a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process such that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in one or more boxes of the flowchart and / or block diagram.

[0244] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of code, or a portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block of the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or acts, or by a combination of dedicated hardware and computer instructions.

[0245] The embodiments of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the market, or to enable other ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for deducing the transportation situation of rail transit, characterized in that, Including: For any station in the target rail transit network, using the train flow situation deduction network, perform multi-step deduction on the train flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment. Among them, the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before the station at multiple historical statistical moments before the current statistical moment in any train running line passing through the station; For any station in the target rail transit network, using the passenger flow situation deduction network, perform multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the target statistical moment. Among them, the passenger flow feature matrix of any station at the current statistical moment includes: the situations of the retained passengers at each station in the target rail transit network at the multiple historical statistical moments; According to the train delay situations and passenger retention situations of each station in the target rail transit network at the target statistical moment, conduct transportation risk assessment to determine the global operation capacity risk of the target rail transit network at the target statistical moment. Among them, the global operation capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transport demand at the target statistical moment.

2. The method according to claim 1, wherein The train flow situation deduction network is a recurrent neural network, including a long short-term memory network layer and a fully connected layer; The step of, for any station in the target rail transit network, using the train flow situation deduction network, performing multi-step deduction on the train flow feature matrix of the station at the current statistical moment to determine the train delay situation of the station at the target statistical moment after the current statistical moment includes: Initialize the long short-term memory network layer to determine the corresponding initial hidden state of the long short-term memory network layer; Use the long short-term memory network layer to perform iterative processing on the train flow feature matrix corresponding to the station and update the initial hidden state until the number of iterations is equal to the preset number of times to determine the corresponding target hidden state of the long short-term memory network layer; According to the conversion of the target hidden state by the fully connected layer, determine the train delay situation of the station at the target statistical moment.

3. The method according to claim 1 or 2, characterized in that, The step of, for any station in the target rail transit network, using the passenger flow situation deduction network, performing multi-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the target statistical moment includes: According to the passenger flow situation deduction network, perform single-step deduction on the passenger flow feature matrix of the station at the current statistical moment to determine the passenger retention situation of the station at the next statistical moment after the current statistical moment and before the target statistical moment; Based on the sliding window method, update the passenger flow feature matrix of the station at the current statistical moment according to the passenger retention situation of the station at the next statistical moment to determine the updated passenger flow feature matrix; Repeat the above steps according to the updated passenger flow feature matrix until the number of single-step deductions equals the preset number, and determine the passenger retention situation at the target statistical moment of the station.

4. The method according to claim 3, wherein The passenger flow trend deduction network is a directed graph spatio-temporal convolutional network based on multi-head attention, including: a spatial multi-head attention module, a temporal multi-head attention module, a directed graph spatio-temporal convolutional module, and a fully connected layer.

5. The method according to claim 1, wherein The global operation capacity risk of the target rail transit network at the target statistical moment includes: the station operation capacity risk of each station in the target rail transit network at the target statistical moment, and the section operation capacity risk of any section formed by two stations in the target rail transit network at the target statistical moment; The transportation risk assessment is carried out according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment to determine the global operation capacity risk of the target rail transit network at the target statistical moment, including: For any station, determine the station operation capacity risk of the station at the target statistical moment according to the train delay situation and passenger retention situation of the station at the target statistical moment; For any section, determine the section operation capacity risk of the section at the target statistical moment according to the train delay situations of the two stations corresponding to the section at the target statistical moment.

6. The method according to claim 5, characterized in that The station operation capacity risk of any station at the target statistical moment includes: the station passenger flow saturation risk and platform passenger retention risk of the station at the target statistical moment; The determination of the station operation capacity risk of any station at the target statistical moment according to the train delay situation and passenger retention situation of the station at the target statistical moment includes: Determine the in-station passenger flow volume of the station at the target statistical moment; According to the train delay situation of the station at the target statistical moment, determine the passenger throughput and platform evacuation capacity of the station at the target statistical moment; Determine the station passenger flow saturation risk of the station at the target statistical moment according to the in-station passenger flow volume and passenger throughput of the station at the target statistical moment, and the consequences of the passenger flow saturation risk corresponding to the station; Determine the platform passenger retention risk of the station at the target statistical moment according to the passenger retention situation and platform evacuation capacity of the station at the target statistical moment, and the consequences of the passenger retention risk corresponding to the station.

7. The method according to claim 5 or 6, characterized in that, The determination of the section operation capacity risk of any section at the target statistical moment according to the train delay situations of the two stations corresponding to the section at the target statistical moment includes: Determine the section passenger flow volume of the section at the target statistical moment; According to the train delay situations of the two stations corresponding to the section at the target statistical moment, determine the section transportation capacity of the section at the target statistical moment; Determine the section operation capacity risk of the section at the target statistical moment according to the section passenger flow volume and section transportation capacity of the section at the target statistical moment, and the consequences of the passenger flow saturation risk corresponding to the section.

8. A transport situation deduction device for rail transit, characterized in that, Including: The train flow situation deduction module is used to perform multi-step deduction on the train flow feature matrix of any station in the target rail transit network at the current statistical moment by using the train flow situation deduction network, and determine the train delay situation of the station at the target statistical moment after the current statistical moment. Among them, the train flow feature matrix of any station at the current statistical moment includes: the train delay situations of multiple stations before this station at multiple historical statistical moments before the current statistical moment in any train running line passing through this station; The passenger flow situation deduction module is used to perform multi-step deduction on the passenger flow feature matrix of any station in the target rail transit network at the current statistical moment by using the passenger flow situation deduction network, and determine the passenger retention situation of the station at the target statistical moment. Among them, the passenger flow feature matrix of any station at the current statistical moment includes: the passenger retention situations of each station in the target rail transit network at the multiple historical statistical moments; The transportation risk assessment module is used to perform transportation risk assessment according to the train delay situation and passenger retention situation of each station in the target rail transit network at the target statistical moment, and determine the global transport capacity risk of the target rail transit network at the target statistical moment. Among them, the global transport capacity risk of the target rail transit network at the target statistical moment is used to indicate whether the target rail transit network can meet the passenger transport demand at the target statistical moment.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A non-volatile computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 7.

Images