Method, device and electronic equipment for predicting passenger flow in airport transfer area

Abstract

The present application provides a method, device and electronic device for predicting the flow of arriving passengers in an airport transfer area. The method comprises: obtaining the actual diversion ratio data corresponding to each diversion point of arriving passengers heading to the transfer area at a first moment, and the flight plan for a preset time period starting at the first moment; inputting the flight plan and the actual diversion ratio data into a preset simulation model to obtain the flow prediction result of arriving passengers in the airport transfer area at a second moment, and the second moment is the end moment of the preset time period. Through this method, the problem that the existing technology cannot accurately predict the flow of arriving passengers in all transfer areas of the airport can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device and electronic equipment for predicting the flow of arriving passengers in an airport transfer area. Background Art

[0002] At present, the research on the prediction of arriving passenger flow is divided into two types: long-term and short-term prediction. Among them, the long-term prediction method is to predict the passenger throughput in future years based on the historical data of annual passenger throughput; and combine the peak hour passenger flow ratio to obtain the peak hour arrival passenger flow; convert it into the passenger flow in the transfer area of ​​various transportation tools through the passenger travel sharing ratio, and use the simulation method to plan the layout of the airport's comprehensive transportation area. The short-term prediction method uses the method of intelligent prediction model to collect the passenger flow of each transfer area, and build a prediction model through machine learning methods to realize the prediction of passenger flow in the transfer area. However, both the long-term prediction method and the short-term prediction method have certain defects:

[0003] a. Long-term prediction method

[0004] The long-term prediction method cannot be used for short-term flow prediction due to the real-time changes in traffic sharing ratio. At the same time, the simulation method used is a static simulation in which the parameters do not change with time, so the long-term prediction method has defects.

[0005] b. Short-term prediction method

[0006] Due to the current status of airport information construction, the short-term prediction method has limited ways to collect data on arriving passengers, and usually can only achieve predictions for a single transfer area. In addition, in order to achieve traffic predictions for all transfer areas, some researchers use the proportion of WIFI connection traffic as the real-time sharing ratio of passengers in each transfer area. Obviously, the proportion of WIFI connections in each transfer area cannot be used as the real-time sharing ratio of passengers, so the short-term prediction method also has defects. Summary of the invention

[0007] The purpose of the embodiments of the present application is to provide a method, device and electronic equipment for predicting the flow of arriving passengers in an airport transfer area, so as to improve the problem that "the existing technology cannot accurately predict the flow of arriving passengers in an airport transfer area".

[0008] The present invention is achieved in that:

[0009] In a first aspect, an embodiment of the present application provides a method for predicting the flow of arriving passengers in an airport transfer area, the method comprising: obtaining actual diversion ratio data corresponding to each diversion point for arriving passengers to go to a transfer area at a first moment, and a flight plan for a preset time period with the first moment as the starting moment; inputting the flight plan and the actual diversion ratio data into a pre-set simulation model to obtain a flow prediction result of arriving passengers in the airport transfer area at a second moment, the second moment being the end moment of the preset time period.

[0010] In the embodiment of the present application, by taking the actual diversion ratio data corresponding to each diversion point of the arriving passengers to the transfer area at the first moment and the flight plan of the preset time period starting at the first moment as input data, and inputting the preset simulation model, the flow prediction results of the arriving passengers at all airport transfer areas at the second moment can be accurately obtained. Compared with the prior art, short-term flow prediction can be achieved without using the WIFI connection flow ratio as the real-time sharing ratio of passengers in each transfer area.

[0011] In combination with the technical solution provided in the first aspect above, in some possible implementations, the simulation model includes an arrival passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module, and the inputting of the flight plan and the actual diversion ratio data into a preset simulation model to obtain a flow prediction result of arrival passengers in the airport transfer area at a second moment includes: processing the flight plan through the arrival passenger generation module to obtain an arrival passenger flow generation result at a baggage carousel exit at the second moment; processing the actual diversion ratio data through the diversion ratio prediction module to obtain a diversion ratio prediction result corresponding to each diversion point at the second moment; and processing the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module to obtain the flow prediction result.

[0012] In the embodiment of the present application, the baggage carousel exit is a location that all arriving passengers will pass through before entering the airport transfer area. By obtaining the arrival passenger flow generation result at the baggage carousel exit, the total number of arriving passengers at the second moment can be obtained. In addition, the arrival passenger flow generation result and the diversion ratio prediction result are processed by the transfer area passenger flow prediction module, and the flow prediction result of the arrival passengers in the airport transfer area at the second moment can be accurately obtained.

[0013] In combination with the technical solution provided in the first aspect above, in some possible implementations, the diversion ratio prediction module includes a spatiotemporal graph convolutional neural network and a gated recurrent unit, and the actual diversion ratio data is processed by the diversion ratio prediction module to obtain the diversion ratio prediction results corresponding to the each diversion point at the second moment, including: processing the actual diversion ratio data by the spatiotemporal graph convolutional neural network to obtain the spatial correlation characteristics of the passenger flow at the second moment; processing the actual diversion ratio data by the gated recurrent unit to obtain the time dynamic correlation characteristics of the passenger flow at the second moment; processing the spatial correlation characteristics and the time dynamic correlation characteristics by the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result.

[0014] In the embodiment of the present application, since the arriving passengers from different boarding gates to each diversion point are in a topological structure in space, the spatial dependency relationship between different boarding gates and each diversion point can be obtained by using a spatiotemporal graph convolutional neural network, and the fixed spatial correlation characteristics can be obtained through the adjacency matrix in the spatiotemporal graph convolutional neural network. In addition, the gated recurrent unit can be used to obtain the time characteristics of the passenger boarding gate of the terminal to each diversion node, and capture the time lag effect caused by the spatial structure of the terminal, thereby avoiding the problem of affecting the accuracy caused by the late arrival of passengers at the diversion node at a long distance boarding gate within the same time scale, and then accurately obtain the diversion ratio prediction result at the second moment.

[0015] In combination with the technical solution provided in the first aspect above, in some possible implementation methods, obtaining actual diversion ratio data corresponding to each diversion point for arriving passengers heading to the transfer area at the first moment includes: obtaining surveillance videos of each diversion point at the first moment; and obtaining the actual diversion ratio data based on the surveillance videos.

[0016] In the embodiment of the present application, through the above method, the actual diversion ratio data at the first moment can be obtained more quickly and accurately, thereby improving the accuracy of the flow prediction results of arriving passengers in the airport transfer area at the second moment.

[0017] In combination with the technical solution provided in the first aspect above, in some possible implementations, the method further includes: when the time reaches the second moment, obtaining the actual passenger flow at the baggage carousel exit and the actual diversion ratio data corresponding to each diversion point; optimizing the simulation model according to the arrival passenger flow generation result and the actual passenger flow at the second moment; optimizing the simulation model according to the diversion ratio prediction result and the actual diversion ratio data at the second moment.

[0018] In an embodiment of the present application, the simulation model is optimized based on the arrival passenger flow generation result generated at the first moment and the actual passenger flow at the second moment; and the simulation model is optimized based on the diversion ratio prediction result predicted at the first moment and the actual diversion ratio data at the second moment. This enables the simulation model to be continuously adjusted during use, thereby improving the prediction accuracy of the simulation model.

[0019] In combination with the technical solution provided in the first aspect above, in some possible implementation methods, the simulation model is obtained through the following steps: constructing an initial simulation model according to the flight plan, the walking process of arriving passengers, multiple preset diversion points and the spatial layout of the airport, and initializing the parameter values ​​of the parameters characterizing the passenger characteristics and luggage characteristics in the initial simulation model, and setting the initial value of the diversion ratio; training the initial simulation model according to the historical flight plan and the historical diversion ratio data corresponding to the various diversion points to obtain the simulation model.

[0020] In combination with the technical solution provided in the first aspect above, in some possible implementations, the initial simulation model is trained according to the historical flight plan and the historical diversion ratios corresponding to the various diversion points, including: inputting the historical flight plan into the arrival passenger generation module of the initial simulation model to obtain the simulated passenger flow at the baggage carousel exit; judging whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range according to the feedback incentive function, and if the error is not within the preset range, correcting the parameter values ​​characterizing the passenger characteristics and the baggage characteristics according to the error; repeating the above steps until the error is within the preset range to obtain the final parameter value; inputting the historical diversion ratio data into the diversion ratio prediction module of the initial simulation model to obtain the diversion ratio prediction result; training the diversion ratio prediction module according to the diversion ratio prediction result and the error between the historical diversion ratio data to obtain the spatial correlation characteristics and time dynamic correlation characteristics of the passenger flow corresponding to the various diversion points.

[0021] In combination with the technical solution provided in the first aspect above, in some possible implementations, the feedback excitation function is: Among them, γ is the discount rate of reinforcement learning, y i For the simulated passenger flow, is the actual passenger flow, and i is the preset number of simulations.

[0022] In combination with the technical solution provided in the first aspect above, in some possible implementations, the parameters characterizing the passenger characteristics and luggage characteristics in the initial simulation model include passenger arrival distribution law parameters, passenger average walking speed, first piece of luggage on the carousel time and luggage arrival distribution law parameters.

[0023] In a second aspect, an embodiment of the present application provides a method for constructing a simulation model, the method comprising: constructing an initial simulation model according to flight plans, the walking flow of arriving passengers, a plurality of preset diversion points and the spatial layout of the airport, initializing the parameter values ​​of the parameters characterizing the passenger characteristics and Li characteristics in the initial simulation model, and setting an initial value of the diversion ratio; training the initial simulation model according to historical flight plans and historical diversion ratio data corresponding to each diversion point to obtain the simulation model.

[0024] In the embodiment of the present application, through the above method, a simulation model for predicting the flow of arriving passengers in the airport transfer area at future times can be quickly and accurately constructed.

[0025] In a third aspect, an embodiment of the present application provides an airport transfer area arrival passenger flow prediction device, the device comprising: an acquisition module, used to obtain actual diversion ratio data corresponding to each diversion point of arriving passengers heading to the transfer area at a first moment, and a flight plan for a preset time period with the first moment as the starting moment; a prediction module, used to input the flight plan and the actual diversion ratio data into a pre-set simulation model, to obtain the airport transfer area arrival passenger flow prediction result at a second moment, the second moment being the end moment of the preset time period.

[0026] In a fourth aspect, an embodiment of the present application provides a device for constructing a simulation model, the device comprising: a processing module, for constructing an initial simulation model according to flight plans, the walking process of arriving passengers, a plurality of preset diversion points and the spatial layout of the airport, and for initializing the parameter values ​​of the parameters characterizing the passenger characteristics and the luggage characteristics in the initial simulation model, and for setting an initial value of the diversion ratio; a construction module, for training the initial simulation model according to historical flight plans and historical diversion ratio data corresponding to each diversion point, to obtain the simulation model.

[0027] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory, the processor and the memory being connected; the memory being used to store programs; the processor being used to call the programs stored in the memory, and executing methods provided in the above-mentioned first aspect embodiment and / or in combination with some possible implementations of the above-mentioned first aspect embodiment, and / or executing methods provided in some possible implementations of the above-mentioned second aspect embodiment.

[0028] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, executes a method provided in the embodiment of the first aspect described above and / or in combination with some possible implementations of the embodiment of the first aspect described above, and / or executes a method provided in some possible implementations of the embodiment of the second aspect described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 A schematic diagram of the structure of the first simulation model provided in an embodiment of the present application.

[0031] Figure 2 A schematic diagram of the structure of the second simulation model provided in an embodiment of the present application.

[0032] Figure 3 A flowchart of the steps of a method for constructing a simulation model provided in an embodiment of the present application.

[0033] Figure 4 A schematic diagram of the overall framework of a spatiotemporal graph convolutional neural network provided in an embodiment of the present application.

[0034] Figure 5 A flowchart of the steps of a method for predicting the flow of arriving passengers in an airport transfer area provided in an embodiment of the present application.

[0035] Figure 6 A schematic diagram showing a comparison between a predicted result of the passenger flow inbound at a parking lot area and the actual passenger flow inbound at a parking lot area provided in an embodiment of the present application.

[0036] Figure 7 A schematic diagram of a comparison between a predicted result of the passenger flow inbound to the port and the actual passenger flow inbound to the port in a subway area provided in an embodiment of the present application.

[0037] Figure 8 A schematic diagram of a comparison between a predicted result of the arrival passenger flow in an airport bus area and the actual arrival passenger flow provided in an embodiment of the present application.

[0038] Fig. 9 A schematic diagram of a comparison between a predicted result of the passenger flow inbound at a taxi area and the actual passenger flow inbound at a taxi area provided in an embodiment of the present application.

[0039] Fig.10 A module block diagram of a device for predicting the flow of arriving passengers in an airport transfer area provided in an embodiment of the present application.

[0040] Fig.11 A module block diagram of a simulation model construction device provided in an embodiment of the present application.

[0041] Fig.12 A module block diagram of an electronic device provided in an embodiment of the present application.

[0042] Icons: 100-simulation model; 10-data input module; 20-arrival passenger generation module; 21-passenger generation submodule; 22-baggage carousel extraction submodule; 23-passenger behavior submodule; 30-diversion ratio prediction module; 31-space-time graph convolutional neural network; 32-gated recurrent unit; 33-diversion ratio correction submodule; 40-transfer area passenger flow prediction module; 50-parameter correction module; 200-airport transfer area arrival passenger prediction device; 201-acquisition module; 202-prediction module; 203-optimization module; 204-training module; 300-simulation model construction device; 301-processing module; 302-construction module; 400-electronic device; 410-processor; 420-memory. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0044] In view of the fact that the prior art cannot accurately predict the passenger flow in the airport transfer area, the inventor of the present application has proposed the following embodiments to solve the above problem after research and exploration.

[0045] The present application embodiment provides a simulation model 100 for predicting the passenger flow in the airport transfer area at a future time. Figure 1 and Figure 2 As shown, the simulation model 100 includes: a data input module 10, an arriving passenger generation module 20, a diversion ratio prediction module 30, a transfer area passenger flow prediction module 40 and a parameter correction module 50.

[0046] Among them, the data input module 10 is used to obtain input data, which includes the flight plan of a preset time period with a first moment as the starting moment, the actual diversion ratio data corresponding to each diversion point of the arriving passengers going to the transfer area at the first moment, and the actual passenger flow at the baggage carousel exit at the second moment and the actual diversion ratio data corresponding to each diversion point, wherein the above-mentioned second moment is the end moment of the above-mentioned preset time period, and the above-mentioned baggage carousel exit is a place that all arriving passengers will pass before entering the airport transfer area.

[0047] The arriving passenger generation module 20 is used to process the flight plan of a preset time period starting at the first moment, and obtain the arriving passenger flow generation result at the baggage carousel exit at the second moment.

[0048] In addition, the above-mentioned arriving passenger generation module 20 includes: a passenger generation submodule 21, a baggage carousel extraction submodule 22 and a passenger behavior submodule 23, wherein the above-mentioned passenger generation submodule 21 is used to process the above-mentioned flight plan and generate corresponding passengers in the simulation model 100 in the order of time points. The above-mentioned baggage carousel extraction submodule 22 is used to process the above-mentioned flight plan, and in combination with the passengers generated in the simulation model 100 and the movement of the passengers in the simulation model 100, obtain the passenger flow at the baggage carousel exit at the second moment. The above-mentioned passenger behavior submodule 23 is used to construct the passenger walking behavior according to the social force model, and according to the shortest path principle, set the passenger path finding algorithm to ensure that the passengers reach the airport transfer area according to the shortest path in the terminal.

[0049] The diversion ratio prediction module 30 is used to process the actual diversion ratio data corresponding to each diversion point for arriving passengers to go to the transfer area at a first moment, and obtain the diversion ratio prediction result corresponding to each diversion point at a second moment; and when the time reaches the second moment, according to the error between the actual diversion ratio data corresponding to each diversion point at the current moment and the diversion ratio prediction result, correct its parameters.

[0050] In addition, the above-mentioned diversion ratio prediction module 30 includes: a spatiotemporal graph convolutional neural network 31, a gated loop unit 32 and a diversion ratio correction submodule 33, wherein the above-mentioned spatiotemporal graph convolutional neural network 31 is used to process the input actual diversion ratio to obtain the spatial correlation characteristics of the passenger flow at the second moment. The above-mentioned gated loop unit 32 is used to process the input actual diversion ratio to obtain the temporal dynamic correlation characteristics of the passenger flow at the second moment; and then the above-mentioned spatial correlation characteristics and temporal dynamic correlation characteristics are processed through the fully connected layer in the spatiotemporal graph convolutional neural network 31 to obtain the above-mentioned diversion ratio prediction result. The above-mentioned diversion ratio correction submodule 33 is used to correct the various parameters in the spatiotemporal graph convolutional neural network 31 and the gated loop unit 32 according to the error between the actual diversion ratio data corresponding to each diversion point at the current moment and the diversion ratio prediction result when the time reaches the second moment.

[0051] The transfer area passenger flow prediction module 40 is used to process the arrival passenger flow generation result output by the arrival passenger generation module 20 and the diversion ratio prediction result output by the diversion ratio prediction module 30 according to the specific setting positions of each diversion point in the simulation model 100, and obtain the flow prediction result of the arrival passengers in the airport transfer area at the second moment.

[0052] The parameter correction module 50 is used to receive the actual passenger flow at the baggage carousel exit at the current moment sent by the data input module 10 when the time reaches the second moment, and correct the parameters in the passenger generation submodule 21 and the baggage carousel extraction submodule 22 according to the error between the actual passenger flow and the diversion ratio prediction result.

[0053] The present application also provides a method for constructing a simulation model, which is used to construct the above simulation model. Figure 3 The specific process and steps of a method for constructing a simulation model are described.

[0054] It should be noted that the construction method of the simulation model provided in the embodiment of the present application is not based on Figure 3 The order shown below is a limitation.

[0055] Step S101: construct an initial simulation model according to the flight plan, the walking flow of arriving passengers, multiple preset diversion points and the spatial layout of the airport, initialize the parameter values ​​of the parameters representing the passenger characteristics and luggage characteristics in the initial simulation model, and set the initial value of the diversion ratio.

[0056] Among them, the flight plan includes: flight number, flight arrival time, boarding gate number, number of passengers, number of luggage, baggage carousel number and aircraft stand identification; the preset multiple diversion points are various diversion points determined according to the actual situation of the airport; the walking process of arriving passengers is: passengers arrive at the terminal from the boarding gate, and reach the baggage claim area through the path inside the building. Among them, passengers with luggage need to wait for the conveyor belt to transport the luggage to the baggage claim hall, complete the luggage claim, and then go to the airport transfer area from the exit of the baggage claim hall, while passengers without luggage go directly to the airport transfer area from the exit of the baggage claim hall, that is, passengers without luggage do not need to go through the baggage claim link.

[0057] Take a certain airport as an example. The transfer area of ​​the airport includes: parking area, subway area, airport bus area and taxi area. In the process of all passengers going from the baggage claim hall exit to each airport transfer area, they need to pass through four key diversion points, which are: the diversion point between the second-floor parking building passage and the elevator to the first floor; if the passenger goes to the second-floor parking building passage, he will pass the diversion point between the subway and the parking transfer area; if the passenger goes to the first floor, he will pass the diversion point between the airport bus and the taxi transfer area; if the passenger goes to the taxi transfer area, he will pass the diversion point between Taxi Area 1 and Taxi Area 2. It should be noted that the specific location and number of each diversion point are related to the spatial layout of the airport, and the specific location and number of the diversion points can be set according to the actual situation of the airport.

[0058] Specifically, a simulation scene is constructed according to the spatial layout of the airport. For example, a simulation scene can be constructed according to a CAD drawing of the airport, or a simulation scene can be constructed according to a three-dimensional model of the airport.

[0059] A passenger behavior module is established so that passengers walk in the terminal according to the social force model and reach the airport transfer area according to the shortest path.

[0060] Establish a parameter correction module composed of reinforcement learning algorithm.

[0061] Establish an arrival passenger generation module, which includes a baggage carousel retrieval submodule and a passenger generation module. Specifically, the passenger generation submodule is established according to the flight plan, so that passengers appear in the terminal of the simulation model according to the flight arrival time, and set the passenger arrival distribution law parameter λ person / minute and the average passenger walking speed V meter / second, wherein the above passenger arrival distribution law parameters and the average passenger walking speed represent the passenger characteristics. Establish a baggage carousel retrieval submodule, according to the flight plan, set the number of passengers carrying luggage, so that passengers carrying luggage go to the airport transfer area after the luggage retrieval link; and passengers without luggage go directly to the airport transfer area through the luggage retrieval exit, and set the first piece of luggage on the carousel time T minutes and the luggage arrival distribution law parameter τ piece / minute, wherein the above first piece of luggage on the carousel time and the luggage arrival distribution law parameter represent the luggage characteristics.

[0062] A diversion ratio prediction module is established, which includes a spatiotemporal graph convolutional neural network and a gated recurrent unit. The spatiotemporal graph convolutional neural network is used to obtain spatial correlation characteristics of passenger flow, and the gated recurrent unit is used to obtain temporal dynamic correlation characteristics of passenger flow.

[0063] Establish a passenger flow prediction module for the transfer area and set the initial value of the diversion ratio at the preset diversion point.

[0064] After the initial simulation model is constructed, the method may continue to execute step S102.

[0065] Step S102: training the initial simulation model according to the historical flight plans and the historical diversion ratio data corresponding to each diversion point to obtain a simulation model.

[0066] Specifically, step S102 is divided into two training processes, and the two training processes are described below respectively.

[0067] The first training process is: input the historical flight plan into the initial simulation model to simulate the arrival passenger generation module to obtain the simulated passenger flow at the baggage carousel exit; according to the feedback incentive function, determine whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within the preset range. If the error is not within the preset range, the parameter values ​​representing the passenger characteristics and the baggage characteristics are corrected according to the error; repeat the above steps until the error is within the preset range to obtain the final parameter value.

[0068] Specifically, historical flight plans are selected as simulation historical data, for example: you can select a flight plan for a certain time period as simulation historical data, or you can select a flight plan for a certain day in the past as simulation historical data; and use the passenger flow in the baggage claim hall according to the video statistics corresponding to the selected historical flight plan as the label data, for example: select the historical flight plan from 8:00 to 8:30 on June 10, 2021 as the simulation historical data, then the label data is the passenger flow in the baggage claim hall according to the video statistics at 8:30; for example: select the historical flight plan on June 12, 2021 as the simulation historical data, then the simulation data can be regarded as multiple continuous time periods, for example: set the preset time period to 30 minutes, that is, the simulation historical data can be divided into 48 continuous time periods of 00:00-00:30, 00:30-01:000,…, 23:30-24:00, then the label data is the passenger flow in the baggage claim hall according to the video statistics at 00:30, 01:00,…, 24:00.

[0069] The parameters characterizing the characteristics of passengers and luggage are set, wherein the parameters characterizing the characteristics of passengers and luggage include: passenger arrival distribution law parameter λ person / minute, average passenger walking speed V meters / second, first piece of luggage on the carousel time T minutes and luggage arrival distribution law parameter τ piece / minute. As shown in Table 1, the above parameters are set as follows.

[0070] Table 1

[0071] Parameter name Initial Value Fluctuation range unit Flight passenger arrival distribution parameter λ 5 2～10 Person / minute Average walking speed of passengers V 0.7 0.5～1.2 m / s Time for first piece of luggage to be put on carousel T 8 5～12 minute Baggage arrival distribution parameter τ 9 5～20 Pieces / minute

[0072] From Table 1 above, we can see that the action mechanism action i =[λ i ,V i ,T i ,τ i ] and i=1,2,...,100, the parameter selection strategy is λ i ∈[2,10], V i ∈[0.5,1.2], T i ∈[5,12], τ i∈[5,20], the initial state is action0=[5,0.7,8,9]. It should be noted that the initial values ​​of the above four parameters can be arbitrarily selected and combined within the given fluctuation range.

[0073] After the above parameters are set, the historical flight plan is input into the initial simulation model to obtain the simulated passenger flow at the baggage carousel exit.

[0074] After obtaining the simulated passenger flow, the parameter correction module can obtain the error between the simulated passenger flow and the passenger flow of the baggage claim hall in the corresponding video statistics (i.e., the corresponding tag data). And according to the feedback incentive function, it is determined whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within the preset range. If the error is not within the preset range, the parameter values ​​representing the passenger characteristics and the baggage characteristics are corrected according to the error.

[0075] Repeat the above steps until the error is within the preset range to obtain the final parameter value.

[0076] It should be noted that the above feedback incentive function is:

[0077]

[0078] Among them, γ is the discount rate of reinforcement learning, y i For the simulated passenger flow, is the actual passenger flow, and i is the preset number of simulations.

[0079] In addition, it should be noted that in the simulation model, the passenger flow at the baggage carousel exit is the result of the passenger arrival distribution law parameters, the average passenger walking speed, the first baggage carousel time and the baggage arrival distribution law parameters. The parameter adjustment results output by each reinforcement learning can be used as the parameter value of the next simulation model. When the error is stable, the optimal simulation model parameter combination can be obtained. For example, the optimal simulation model parameter combination is λ=7,V=0.86,T=9.25,τ=10.5.

[0080] The second training process is: input the historical diversion ratio data into the diversion ratio prediction module of the initial simulation model to obtain the diversion ratio prediction result; according to the error between the diversion ratio prediction result and the historical diversion ratio data, the diversion ratio prediction module is trained to obtain the spatial correlation characteristics and time dynamic correlation characteristics of the passenger flow corresponding to each diversion point.

[0081] Specifically, the operation of obtaining the spatial correlation characteristics of the passenger flow corresponding to each diversion point is as follows.

[0082] The diversion ratio prediction problem can be expressed as:

[0083]

[0084] Among them, the function f(·) represents mapping the historical key diversion node diversion ratio information to the future diversion ratio information; It represents the diversion ratio of diversion node j at time t+i, where n represents the number of nodes, M represents the length of the historical time series, N represents the data to be predicted, and the sampling interval is once per minute, that is, the video counts the diversion ratio of the key diversion node once per minute. It represents the diversion ratio to two different transfer areas at time t+i.

[0085] Each group of data on the diversion ratio of multiple key nodes at the same time t+i with a sampling time interval of one minute is input into the spatiotemporal graph convolutional neural network, and the spatiotemporal graph convolutional neural network is used to capture the spatial correlation characteristics of passenger flow. Figure 4 As shown, the input conversion layer will input data The spatial correlation of is converted into the format required by the spatiotemporal graph convolutional neural network. Enter the L-layer GCN network of the STGCN network for spatial feature extraction, and the propagation formula is:

[0086]

[0087] Among them, X is the input of the L layer in the STGCN network obtained by transforming the input data by the input transformation layer. A is the proximity matrix of multiple key shunt nodes, which consists of the spatial structure of multiple key shunt nodes. If the node is directly connected, the corresponding position is 1, otherwise it is 0. W is the weight matrix, and σ(·) is the nonlinear transformation activation function. I is the identity matrix. W (0) Initial weight coefficient matrix, W (L) is the linear transformation of the weight matrix after L layers, where L represents the number of layers.

[0088] After training, the deviation is corrected through the loss function, and finally the node feature Z at a single moment is obtained L =[X L ,W L ].

[0089] Specifically, the operation of obtaining the time dynamic correlation characteristics of the passenger flow corresponding to each diversion point is as follows.

[0090] The temporal attention encoding model is adopted as the temporal dimension modeling structure. The method consists of two gated recurrent unit (GRU) modules with independent parameters.

[0091] H i,t =GRU(Z i,t,H i,t-1 ) (4)

[0092] Among them, H i,t is the output of sensor node i at time t. i,t is the feature sequence obtained by graph convolution operation. i,t-1 is the output of sensor node i at time t-1.

[0093] GRU uses a gating mechanism to control the input level of the current and previous state. The detailed process is shown in the following formula:

[0094] u i,t =σ(W u [Z i,t ,H i,t-1 ]+b u ) (5)

[0095] r i,t =σ(W r [Z i,t ,H i,t-1 ]+b r ) (6)

[0096]

[0097]

[0098] Among them, u i,t and r i,t Respectively represent the update gate and reset gate size in GRU, σ represents the Sigmoid activation function, W and b are learnable parameters, Represents the state of the hidden layer at time t. It should be noted that the larger the value of the update gate and the smaller the value of the reset gate, the less information about the state at the previous moment is written. In this way, the gated recurrent unit can obtain the long-term time correlation of the key node diversion ratio.

[0099] According to the above operation of obtaining spatial correlation characteristics and time dynamic correlation characteristics, the input historical diversion ratio data is used as the training input data of the diversion ratio prediction module for training, and the spatial correlation characteristics and time dynamic correlation characteristics corresponding to each historical diversion ratio data can be obtained. The spatial correlation characteristics and time dynamic correlation characteristics are input into the fully connected layer of the spatiotemporal graph convolutional neural network for processing, and the diversion ratio prediction results corresponding to each historical diversion ratio data can be obtained. According to the diversion ratio prediction results output by the model and the error of the historical diversion ratio data, the diversion ratio prediction module is adaptively iterated and optimized, and finally the spatial correlation characteristics and time dynamic correlation characteristics of the passenger flow corresponding to each diversion point are obtained.

[0100] It should be noted that the spatiotemporal graph convolutional neural network and gated recurrent unit in the diversion ratio prediction module, as well as the above-mentioned operations of obtaining spatial correlation characteristics and temporal dynamic correlation characteristics are well known to those skilled in the art and will not be described here.

[0101] In addition, it should be noted that the above two training processes can be carried out simultaneously; the first training process can be carried out first, and then the second training process; the second training process can be carried out first, and then the first training process, which is not limited here.

[0102] The present application also provides a method for predicting the flow of arriving passengers in an airport transfer area. The method is used to predict the flow of arriving passengers in an airport transfer area. Figure 5 The specific process and steps of a method for predicting the arrival passenger flow in an airport transfer area are described.

[0103] It should be noted that the method for predicting the arrival passenger flow in the airport transfer area provided in the embodiment of the present application is not based on Figure 5 The order shown below is a limitation.

[0104] Step S201: obtaining actual diversion ratio data corresponding to each diversion point for arriving passengers heading to the transfer area at a first moment, and a flight plan for a preset time period starting at the first moment.

[0105] Specifically, obtaining the actual diversion ratio data corresponding to each diversion point of the arriving passengers going to the transfer area at the first moment may specifically include: obtaining the monitoring video of each diversion point at the first moment; and obtaining the actual diversion ratio data according to the monitoring video. In this way, the actual diversion ratio data of each diversion point at the first moment can be obtained quickly and accurately, thereby improving the accuracy of the flow prediction result of the arriving passengers in the airport transfer area at the second moment.

[0106] It should be noted that the above flight plan includes: the flight number corresponding to the preset time period, flight arrival time, boarding gate number, number of passengers, number of luggage, baggage carousel number and aircraft stand identification.

[0107] After obtaining the actual diversion ratio data and the flight plan, the method may continue to execute step S202.

[0108] Step S202: inputting the flight plan and the actual diversion ratio data into a preset simulation model to obtain the flow prediction result of arriving passengers in the airport transfer area at the second moment.

[0109] The second moment is the end moment of the preset time period.

[0110] Specifically, the simulation model includes an arrival passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module. The flight plan and actual diversion ratio data are input into a preset simulation model to obtain the flow prediction result of the arrival passengers in the airport transfer area at the second moment. The method may specifically include: processing the flight plan through the arrival passenger generation module to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; processing the actual diversion ratio data through the diversion ratio prediction module to obtain the diversion ratio prediction result corresponding to each diversion point at the second moment; processing the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module to obtain the flow prediction result.

[0111] In the embodiment of the present application, the baggage carousel exit is a location that all arriving passengers will pass through before entering the airport transfer area. By obtaining the arrival passenger flow generation result at the baggage carousel exit, the total number of arriving passengers at the second moment can be obtained. In addition, the arrival passenger flow generation result and the diversion ratio prediction result are processed by the transfer area passenger flow prediction module, and the flow prediction result of the arrival passengers in the airport transfer area at the second moment can be accurately obtained.

[0112] Optionally, the diversion ratio prediction module includes a spatiotemporal graph convolutional neural network and a gated recurrent unit. The diversion ratio prediction module processes the actual diversion ratio data to obtain the diversion ratio prediction results corresponding to each diversion point at the second moment. Specifically, the diversion ratio prediction results can include: processing the actual diversion ratio data through the spatiotemporal graph convolutional neural network to obtain the spatial correlation characteristics of the passenger flow at the second moment; processing the actual diversion ratio data through the gated recurrent unit to obtain the temporal dynamic correlation characteristics of the passenger flow at the second moment; processing the spatial correlation characteristics and the temporal dynamic correlation characteristics through the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction results. In this way, the diversion ratio prediction results at the second moment can be accurately obtained.

[0113] It should be noted that after the prediction is completed using the simulation model, when the time reaches the second moment, the simulation model can be optimized (i.e., parameter correction) using the actual value corresponding to the time and the predicted value of the simulation model for the time.

[0114] Specifically, when the time reaches the second moment, the actual passenger flow at the baggage carousel exit and the actual diversion ratio data corresponding to each diversion point are obtained; the simulation model is optimized according to the arrival passenger flow generation result and the actual passenger flow at the second moment; the simulation model is optimized according to the diversion ratio prediction result and the actual diversion ratio data at the second moment.

[0115] Through the above method, the simulation model can be continuously adjusted during use, thereby improving the prediction accuracy of the simulation model.

[0116] Taking the four diversion points given in the above-mentioned step S101 as an example, the flight plan for November 1, 2021 is input into the simulation model, and before each prediction of the airport transfer area arrival passenger flow corresponding to the next time point, the actual diversion ratio data of each diversion point at the current time is input. For example: at 8 o'clock, the actual diversion ratio data of each diversion point of the airport at 8 o'clock is input to obtain the airport transfer area arrival passenger flow at 8:30. When the time reaches 8:30, the actual diversion ratio data of each diversion point of the airport at 8:30 is input into the simulation model to obtain the airport transfer area arrival passenger flow at 9 o'clock, and then the arrival passenger flow forecast results of each transfer area on November 1, 2021 are obtained.

[0117] It should be noted that the duration between the above time points can be set according to actual needs. For example, at 8 o'clock, the actual diversion ratio data of each diversion point of the airport at 8 o'clock is input to obtain the passenger flow of the airport transfer area at 8:10; when the time reaches 8:10, the actual diversion ratio data of each diversion point of the airport at 8:10 is input to the simulation model to obtain the passenger flow of the airport transfer area at 8:20. It is understandable that after selecting different durations, the simulation historical data needs to be divided into corresponding durations to construct the simulation model. And by setting the appropriate duration, the passenger flow of the airport transfer area can be predicted in real time.

[0118] like Figure 6-Figure 9 As shown, Figure 6-Figure 9 The following are schematic diagrams showing the comparison between the predicted arrival passenger flow and the actual arrival passenger flow in the four different transfer areas (parking lot area, subway area, airport bus area and taxi area). Figure 6-Figure 9 The horizontal axis is time, and the vertical axis is passenger flow. The triangles in the figure are the predicted passenger flow results at each time point, and the dots are the actual passenger flow at each time point. Figure 6-Figure 9 It can be seen that the airport taxi area, airport bus area, and parking area have less passenger flow from 3 to 7 in the morning, and the peak period mainly occurs from 20 to 24 at night; the subway is affected by the departure schedule, and the flow increases significantly after 5 in the morning, reaching a peak at around 20:00. And by comparing the predicted results of the arrival passenger flow with the actual arrival passenger flow, it can be seen that the simulation model predicts the arrival passenger flow in the airport transfer area with high accuracy.

[0119] See also Fig.10Based on the same inventive concept, an embodiment of the present application also provides an airport transfer area arrival passenger flow prediction device 200, and the device 200 includes: an acquisition module 201 and a prediction module 202.

[0120] The acquisition module 201 is used to acquire the actual diversion ratio data corresponding to each diversion point of the arriving passengers heading to the transfer area at the first moment, and the flight plan of the preset time period starting from the first moment;

[0121] The prediction module 202 is used to input the flight plan and the actual diversion ratio data into a preset simulation model to obtain the flow prediction result of the arriving passengers in the airport transfer area at the second moment, where the second moment is the end moment of the preset time period.

[0122] Optionally, the simulation model includes an arrival passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module. Accordingly, the prediction module 202 is specifically used to process the flight plan through the arrival passenger generation module to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; process the actual diversion ratio data through the diversion ratio prediction module to obtain the diversion ratio prediction results corresponding to each diversion point at the second moment; process the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module to obtain the flow prediction result.

[0123] Optionally, the diversion ratio prediction module includes a spatiotemporal graph convolutional neural network and a gated recurrent unit. Accordingly, the prediction module 202 is specifically used to process the actual diversion ratio data through the spatiotemporal graph convolutional neural network to obtain the spatial correlation characteristics of the passenger flow at the second moment; process the actual diversion ratio data through the gated recurrent unit to obtain the temporal dynamic correlation characteristics of the passenger flow at the second moment; process the spatial correlation characteristics and the temporal dynamic correlation characteristics through the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result.

[0124] Optionally, the acquisition module 201 is specifically used to acquire monitoring videos of each diversion point at a first moment; and acquire actual diversion ratio data according to the monitoring videos.

[0125] Optionally, the airport transfer area arrival passenger flow prediction device also includes an optimization module 203, which is used to obtain the actual passenger flow at the baggage carousel exit and the actual diversion ratio data corresponding to each diversion point when the time reaches the second moment; optimize the simulation model according to the arrival passenger flow generation result and the actual passenger flow at the second moment; optimize the simulation model according to the diversion ratio prediction result and the actual diversion ratio data at the second moment.

[0126] Optionally, the airport transfer area arrival passenger flow prediction device also includes a training module 204, which is used to construct an initial simulation model based on flight plans, the walking process of arriving passengers, multiple preset diversion points and the spatial layout of the airport, and initialize the parameter values ​​of the parameters that characterize passenger characteristics and luggage characteristics in the initial simulation model, and set an initial value of the diversion ratio; train the initial simulation model according to historical flight plans and historical diversion ratio data corresponding to each diversion point to obtain a simulation model.

[0127] Optionally, the training module 204 is specifically used to input the historical flight plan into the arrival passenger generation module of the initial simulation model to obtain the simulated passenger flow at the baggage carousel exit; based on the feedback incentive function, determine whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range; if the error is not within the preset range, correct the parameter values ​​characterizing the passenger characteristics and the baggage characteristics based on the error; repeat the above steps until the error is within the preset range to obtain the final parameter value; input the historical diversion ratio data into the diversion ratio prediction module of the initial simulation model to obtain the diversion ratio prediction result; train the diversion ratio prediction module based on the diversion ratio prediction result and the error between the historical diversion ratio data to obtain the spatial correlation characteristics and time dynamic correlation characteristics of the passenger flow corresponding to each diversion point.

[0128] See also Fig.11 Based on the same inventive concept, an embodiment of the present application further provides a simulation model construction device 300 , and the device 300 includes: a processing module 301 and a construction module 302 .

[0129] The processing module 301 is used to construct an initial simulation model according to the flight plan, the walking flow of arriving passengers, multiple preset diversion points and the spatial layout of the airport, initialize the parameter values ​​of the parameters representing the passenger characteristics and luggage characteristics in the initial simulation model, and set the initial value of the diversion ratio.

[0130] The construction module 302 is used to train the initial simulation model according to the historical flight plans and the historical diversion ratio data corresponding to each diversion point to obtain the simulation model.

[0131] See also Fig.12Based on the same inventive concept, the present application embodiment provides a schematic structural block diagram of an electronic device 400, which can be used to implement the above-mentioned method for predicting the flow of arriving passengers in an airport transfer area, or to implement the above-mentioned method for constructing a simulation model. In the embodiment of the present application, the electronic device 400 can be, but is not limited to, a personal computer (PC), a smart phone, a tablet computer, a personal digital assistant (PDA), a mobile Internet device (MID), etc. Structurally, the electronic device 400 may include a processor 410 and a memory 420.

[0132] The processor 410 is electrically connected to the memory 420 directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines. Among them, the processor 410 can be an integrated circuit chip with signal processing capabilities. The processor 410 can also be a general-purpose processor, for example, it can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a discrete gate or transistor logic device, a discrete hardware component, and can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the present application. In addition, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0133] The memory 420 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), and an electric erasable programmable read-only memory (EEPROM). The memory 420 is used to store a program, and the processor 410 executes the program after receiving an execution instruction.

[0134] It should be understood that Fig.12 The structure shown is for illustration only. The electronic device 400 provided in the embodiment of the present application may also have Fig.12Fewer or more components, or with Fig.12 In addition, Fig.12 The components shown may be implemented by software, hardware or a combination thereof.

[0135] It should be noted that, since technicians in the relevant field can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0136] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the method provided in the above embodiment is executed.

[0137] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)).

[0138] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0139] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0140] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0141] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0142] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the flow of arriving passengers in an airport transfer area, characterized in that: The method comprises: Acquire actual diversion ratio data corresponding to each diversion point for passengers arriving at the airport at the first moment and heading to the transfer area, as well as a flight plan for a preset time period starting from the first moment; Inputting the flight plan and the actual diversion ratio data into a preset simulation model to obtain a flow forecast result of arriving passengers in the airport transfer area at a second moment, where the second moment is the end moment of the preset time period; The simulation model includes an arrival passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module, and the diversion ratio prediction module includes a spatiotemporal graph convolutional neural network and a gated loop unit; the inputting the flight plan and the actual diversion ratio data into a preset simulation model to obtain the flow prediction result of the arrival passengers in the airport transfer area at the second moment includes: processing the flight plan through the arrival passenger generation module to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; processing the actual diversion ratio data through the spatiotemporal graph convolutional neural network to obtain the spatial correlation characteristics of the passenger flow at the second moment; processing the actual diversion ratio data through the gated loop unit to obtain the time dynamic correlation characteristics of the passenger flow at the second moment; processing the spatial correlation characteristics and the time dynamic correlation characteristics through the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result; processing the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module to obtain the flow prediction result; The simulation model is obtained by the following steps: constructing an initial simulation model according to the flight plan, the walking process of arriving passengers, a plurality of preset diversion points and the spatial layout of the airport, initializing the parameter values ​​of the parameters representing the characteristics of passengers and luggage in the initial simulation model, and setting the initial value of the diversion ratio; training the initial simulation model according to the historical flight plan and the historical diversion ratio data corresponding to each diversion point to obtain the simulation model; Training the arriving passenger generation module in the simulation model includes: inputting the historical flight plan into the initial simulation model to simulate the arriving passenger generation module, and obtaining the simulated passenger flow at the baggage carousel exit; judging whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range according to the feedback incentive function, and if the error is not within the preset range, correcting the parameter value characterizing the passenger characteristics and the baggage characteristics according to the error; and repeating the above training process until the error is within the preset range to obtain the final parameter value.

2. The method according to claim 1, characterized in that The parameters characterizing the characteristics of passengers and luggage in the initial simulation model include passenger arrival distribution law parameters, passenger average walking speed, first luggage carousel loading time and luggage arrival distribution law parameters.

3. A method for constructing a simulation model, characterized in that: The method comprises: According to the flight plan, the walking process of arriving passengers, the preset multiple diversion points and the spatial layout of the airport, an initial simulation model is constructed, and the parameter values ​​of the parameters representing the characteristics of passengers and luggage in the initial simulation model are initialized, and the initial value of the diversion ratio is set; the initial simulation model includes an arriving passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module; The initial simulation model is trained according to the historical flight plan and the historical diversion ratio data corresponding to each diversion point to obtain the simulation model, so as to process the flight plan through the arrival passenger generation module in the simulation model to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; the actual diversion ratio data is processed through the diversion ratio prediction module in the simulation model to obtain the diversion ratio prediction results corresponding to the each diversion point at the second moment; the arrival passenger flow generation result and the diversion ratio prediction result are processed through the transfer area passenger flow prediction module in the simulation model to obtain the flow prediction result; The actual diversion ratio data is processed by the diversion ratio prediction module in the simulation model to obtain the diversion ratio prediction results corresponding to the various diversion points at the second moment, including: the actual diversion ratio data is processed by the spatiotemporal graph convolutional neural network in the diversion ratio prediction module to obtain the spatial correlation characteristics of the passenger flow at the second moment; the actual diversion ratio data is processed by the gated recurrent unit in the diversion ratio prediction module to obtain the time dynamic correlation characteristics of the passenger flow at the second moment; the spatial correlation characteristics and the time dynamic correlation characteristics are processed by the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result; Training the arriving passenger generation module includes: inputting a historical flight plan into an initial simulation model to simulate the arriving passenger generation module, and obtaining a simulated passenger flow at a baggage carousel exit; judging, based on a feedback incentive function, whether an error between an actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range; if the error is not within the preset range, correcting a parameter value characterizing passenger characteristics and baggage characteristics based on the error; and repeating the above training process until the error is within the preset range to obtain a final parameter value.

4. An airport transfer area arrival passenger flow prediction device, characterized in that: The device comprises: An acquisition module, used to acquire actual diversion ratio data corresponding to each diversion point for passengers arriving at the airport to go to the transfer area at a first moment, and a flight plan for a preset time period starting from the first moment; A prediction module, used for inputting the flight plan and the actual diversion ratio data into a preset simulation model to obtain a flow prediction result of arriving passengers in the airport transfer area at a second moment, where the second moment is the end moment of the preset time period; The simulation model includes an arrival passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module, wherein the diversion ratio prediction module includes a spatiotemporal graph convolutional neural network and a gated loop unit; the prediction module is specifically used to: process the flight plan through the arrival passenger generation module to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; process the actual diversion ratio data through the spatiotemporal graph convolutional neural network to obtain the spatial correlation characteristics of the passenger flow at the second moment; process the actual diversion ratio data through the gated loop unit to obtain the temporal dynamic correlation characteristics of the passenger flow at the second moment; process the spatial correlation characteristics and the temporal dynamic correlation characteristics through the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result; process the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module to obtain the flow prediction result; The training module is specifically used to construct an initial simulation model according to the flight plan, the walking process of arriving passengers, a plurality of preset diversion points and the spatial layout of the airport, initialize the parameter values ​​of the parameters representing the characteristics of passengers and luggage in the initial simulation model, and set the initial value of the diversion ratio; train the initial simulation model according to the historical flight plan and the historical diversion ratio data corresponding to each diversion point to obtain the simulation model; The training module is specifically used to input the historical flight plan into the initial simulation model to simulate the arrival passenger generation module, and obtain the simulated passenger flow at the baggage carousel exit; according to the feedback incentive function, determine whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range; if the error is not within the preset range, correct the parameter value characterizing the passenger characteristics and the baggage characteristics according to the error; repeat the above training process until the error is within the preset range to obtain the final parameter value.

5. A device for constructing a simulation model, characterized in that: The device comprises: A processing module, configured to construct an initial simulation model according to the flight plan, the walking flow of arriving passengers, a plurality of preset diversion points and the spatial layout of the airport, initialize the parameter values ​​of the parameters representing the characteristics of passengers and luggage in the initial simulation model, and set an initial value of the diversion ratio; the initial simulation model includes an arriving passenger generation module, a diversion ratio prediction module and a transfer area passenger flow prediction module; A construction module is used to train the initial simulation model according to the historical flight plan and the historical diversion ratio data corresponding to each diversion point to obtain a simulation model, so as to process the flight plan through the arrival passenger generation module in the simulation model to obtain the arrival passenger flow generation result at the baggage carousel exit at the second moment; process the actual diversion ratio data through the diversion ratio prediction module in the simulation model to obtain the diversion ratio prediction result corresponding to each diversion point at the second moment; process the arrival passenger flow generation result and the diversion ratio prediction result through the transfer area passenger flow prediction module in the simulation model to obtain the flow prediction result; The construction module is specifically used to: process the actual diversion ratio data through the spatiotemporal graph convolutional neural network in the diversion ratio prediction module to obtain the spatial correlation characteristics of the passenger flow at the second moment; process the actual diversion ratio data through the gated recurrent unit in the diversion ratio prediction module to obtain the temporal dynamic correlation characteristics of the passenger flow at the second moment; process the spatial correlation characteristics and the temporal dynamic correlation characteristics through the fully connected layer in the spatiotemporal graph convolutional neural network to obtain the diversion ratio prediction result; The construction module is specifically used to: input the historical flight plan into the initial simulation model to simulate the arrival passenger generation module to obtain the simulated passenger flow at the baggage carousel exit; determine whether the error between the actual passenger flow at the baggage carousel exit and the simulated passenger flow is within a preset range according to the feedback incentive function; if the error is not within the preset range, correct the parameter value characterizing the passenger characteristics and the baggage characteristics according to the error; repeat the above training process until the error is within the preset range to obtain the final parameter value.

6. An electronic device, characterized in that: include: A processor and a memory, the processor and the memory being connected; The memory is used to store programs; The processor is configured to run the program stored in the memory, and execute the method according to claim 1 or 2, or execute the method according to claim 3.

7. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a computer, the method according to claim 1 or 2 or the method according to claim 3 is executed.

Images