CN118135799A - Flow prediction method and equipment for layered expressway toll station outlet - Google Patents
Flow prediction method and equipment for layered expressway toll station outlet Download PDFInfo
- Publication number
- CN118135799A CN118135799A CN202410544141.3A CN202410544141A CN118135799A CN 118135799 A CN118135799 A CN 118135799A CN 202410544141 A CN202410544141 A CN 202410544141A CN 118135799 A CN118135799 A CN 118135799A
- Authority
- CN
- China
- Prior art keywords
- cell
- network
- node
- outlet flow
- traffic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000012216 screening Methods 0.000 claims abstract description 18
- 238000004364 calculation method Methods 0.000 claims abstract description 10
- 238000003062 neural network model Methods 0.000 claims abstract description 10
- 230000003068 static effect Effects 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 41
- 238000013527 convolutional neural network Methods 0.000 claims description 19
- 230000007246 mechanism Effects 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 17
- 230000002457 bidirectional effect Effects 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 8
- 238000013135 deep learning Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 7
- 230000001364 causal effect Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000007499 fusion processing Methods 0.000 claims description 3
- 210000004027 cell Anatomy 0.000 description 132
- 238000010586 diagram Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 210000004460 N cell Anatomy 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention discloses a flow prediction method and equipment for layered expressway toll gate outlets, belongs to the technical field of highway traffic flow, and is used for solving the technical problems that the existing user is difficult to select the optimal expressway toll gate, the user is difficult to select the optimal toll gate outlet in a nearby space area, traffic flow prediction data is not accurate enough, and high-efficiency travel of the user is not facilitated. The method comprises the following steps: carrying out similarity calculation under relevant dynamic and static characteristics among nodes of the expressway network to obtain network node similarity; screening the similarity of the network nodes to determine a toll station cell; combining the multiple neural network models to obtain an outlet flow prediction model; carrying out data prediction processing of the exit flow of a target cell in a toll station cell to obtain the exit flow of the cell; and calculating the flow rate ratio of the target charging station relative to the cell outlet flow rate, and determining the station outlet flow rate of the target charging station.
Description
Technical Field
The application relates to the field of highway traffic flow, in particular to a flow prediction method and equipment for an outlet of a layered highway toll station.
Background
The prediction of the exit flow of the toll station is very important in the management work of the expressway, and the accurate prediction of the exit flow of the toll station can provide accurate, reliable and visual reference information for traffic control and driving guidance, and has great significance for people to avoid congestion periods and traffic management departments to dredge and plan toll channels in advance.
With the continuous development of expressways, road network structures become more complex, and toll stations become more dense. When a user selects a highway route, he or she tends to select one of a plurality of toll stations near a traffic start point to drive into a high speed, and at the same time, to select one of a plurality of toll stations near a traffic end point to drive out of the high speed. The user's choice of entrance and exit tollgates is driven by his own behavior habits on the one hand and by the traffic conditions at that time on the other hand, for example, when the tollgate that the user is accustomed to choosing is congested, the user will choose a relatively idle tollgate nearby at a high speed. At the same time, the path selection behavior of the user inevitably affects the toll gate exit traffic.
In view of the above-mentioned condition of selecting toll stations, the existing highway toll stations are inconvenient to select the entrance and exit, and it is difficult for users to select the optimal toll station entrance and exit according to the traffic state changing in real time. Meanwhile, the traffic flow prediction of different toll stations in the adjacent space area is not accurate enough, so that the user cannot easily predict and acquire traffic flow data in a plurality of toll stations in the adjacent space area, and an optimal in-out high-speed selection mode is generated without using the user according to the predicted traffic data, so that the travel efficiency of the user is reduced.
Disclosure of Invention
The embodiment of the application provides a flow prediction method and equipment for a layered highway toll station outlet, which are used for solving the following technical problems: the existing users have difficulty in selecting the optimal expressway toll gate, so that the users are difficult to select the optimal toll gate outlet in the adjacent space area, and the high-efficiency travel of the users is not facilitated.
The embodiment of the application adopts the following technical scheme:
In one aspect, an embodiment of the present application provides a method for predicting traffic at an exit of a layered highway toll station, including: based on a preset highway network model, similarity calculation under relevant dynamic and static characteristics is carried out among highway network nodes, and network node similarity is obtained; wherein the highway network node is a highway toll station; screening the similarity of the network nodes, and determining a toll station cell according to a screening result; wherein the toll station cell is a network node set under high similarity; model combination is carried out on a plurality of preset neural network models to obtain an outlet flow prediction model based on the expressway toll station; carrying out data prediction processing of the outlet flow of a target cell in the toll station cell through the outlet flow prediction model to obtain the cell outlet flow of the target cell; and calculating the flow rate ratio of the target charging station according to the cell outlet flow rate, and determining the station outlet flow rate of the target charging station.
The embodiment of the application groups expressway toll stations, each group is called a toll station cell, then carries out exit high-speed flow prediction by taking the cell as a unit, and finally further obtains the exit flow prediction result of a single toll station in the cell based on the exit flow prediction result of the cell. The method is beneficial to the user to select the optimal expressway toll gate, so that the user can select the optimal toll gate outlet in the adjacent space area, the accuracy of traffic flow prediction data in the adjacent space area is improved, and the travel efficiency of the user is further improved.
In a possible implementation manner, based on a preset highway network model, similarity calculation under relevant dynamic and static characteristics is performed between highway network nodes to obtain network node similarity, which specifically includes: establishing a highway network model based on a network topology structure of a highway network; according toObtaining the spatial neighbor similarity/>, between the expressway network nodes; Wherein the longitude and latitude of the ith node are/>; Longitude and latitude of the j-th node are/>；/>Representing the Euclidean distance between the longitude and latitude of the ith node and the longitude and latitude of the jth node in the highway network model; according to/>Obtaining the network structure similarity/>, between the expressway network nodes; Wherein/>For the neighbor node set of node i,/>A neighbor node set of the node j; /(I)Representing the number of common neighbor nodes of node i and node j in the highway network model; /(I)The degree of node j; /(I)The degree of node i; /(I)The eccentricity of the node i in the highway network model is represented as the maximum distance in the distance from the node i to any other point in the network; /(I)The eccentricity of the node j in the highway network model is represented as the maximum distance from the node j to any other point in the network; according to/>Obtaining the outlet flow time sequence similarity/>, between the expressway network nodes; Wherein/>Is the i node historical annual sunrise traffic vector, and/>，/>{1, 2..365 } Represents egress traffic on day k of the ith node; /(I)Is the historical annual sunrise traffic vector of the jth node, and，/>{1,2,..365 } Represents egress traffic on day k of node j; wherein the spatial neighbor similarity and the network structure similarity all belong to static features; the outlet flow time sequence similarity belongs to dynamic characteristics; according toObtaining the similarity/>, of network nodes among the expressway network nodes; Wherein/>Weighting the spatial neighbor similarity,/>Weight of similarity of the network structure,/>And weighting the time sequence similarity of the outlet flow.
In a possible implementation manner, the screening of the similarity is performed on the similarity of the network nodes, and the toll station cell is determined according to the screening result, which specifically includes: based on the network node similarity, a node similarity matrix is established; according toObtaining a node similarity matrix/>, after screening; Wherein/>Similarity for the network node; /(I)Is a preset similarity threshold value, and/>50% Quantiles for a set of elements contained in the node similarity matrix; carrying out modular degree maximization processing on the node similarity matrix after screening according to a preset Louvain algorithm, and carrying out node merging on the node similarity matrix after modular degree maximization promotion to obtain the network node set; determining the set of network nodes as the toll station cell; the toll station cell comprises a plurality of expressway toll stations which are in spatial neighbors and have similar outlet flow change trends.
In a possible implementation manner, model combination is performed on a plurality of preset neural network models to obtain an outlet flow prediction model based on the expressway toll station, which specifically includes: respectively connecting the output end in the full connection layer in the preset 1D-CNN model with the input end of the preset TCN network model and the input end of the preset BiGRU network model; the output end of the TCN network model and the output end of the BiGRU network model are connected with the input end of a preset attention mechanism model; the outlet flow prediction model is combined based on the 1D-CNN model, the TCN network model, the BiGRU network model, and the attention mechanism model.
In a possible implementation manner, before performing, by using the exit traffic prediction model, data prediction processing of an exit traffic for a target cell in the toll station cell, to obtain a cell exit traffic of the target cell, the method further includes: dividing a cell set of a toll station cell based on the outlet flow direction requirement of a toll station of an expressway to obtain the target cell and a reference cell set; the target cell is a plurality of toll station cells of which the exit quantity is to be predicted; the reference cell set is a remaining toll station cell set in a plurality of toll station cells; acquiring historical inlet flow data corresponding to each reference cell in the reference cell set; and marking each historical time of day traffic data for the historical ingress traffic data for each reference cell.
In a possible implementation manner, the data prediction processing of the exit flow is performed on the target cell in the toll station cell through the exit flow prediction model, so as to obtain the cell exit flow of the target cell, which specifically includes: performing data fusion processing on the historical inlet flow data corresponding to each reference cell through a 1D-CNN module in the outlet flow prediction model; the data under convolution pooling is overlapped and fused on the fused historical inlet flow data, and an overlapped and fused feature vector based on the historical inlet flow data is obtained; performing residual error connection processing on the causal convolution and the expansion convolution related to the superposition fusion feature vector by a TCN network module in the outlet flow prediction model to obtain TCN network output features; performing bidirectional gating cyclic processing on the relevant forward layer and backward layer by using the BiGRU network module in the outlet flow prediction model to obtain a time sequence output characteristic based on BiGRU network; according to the attention mechanism module in the outlet flow prediction model, giving weights to the TCN network output characteristics and the time sequence output characteristics, and according to the flow input-output dependency relationship between the historical inlet flow data and the flow data of the target cell after the weights are given, performing feedback learning on the 1D-CNN module to obtain an outlet flow prediction model after deep learning; and carrying out outlet flow data prediction on the target cell based on the outlet flow prediction model after deep learning to obtain the cell outlet flow of the target cell.
In a possible implementation manner, before calculating the traffic ratio of the target charging station with respect to the cell exit traffic, determining the station exit traffic of the target charging station, the method further includes: acquiring the historical cell outlet flow of the target cell and the historical site outlet flow of the target charging site; performing time-series data duty ratio calculation on the historical cell outlet flow and the historical site outlet flow to obtain a historical duty ratio time-series data set; inputting the historical duty ratio time sequence data set into a preset Bi-LSTM model; wherein the Bi-LSTM model is a neural network model added with a double-head attention mechanism; performing linear mapping processing on the historical duty ratio time sequence data set through the Bi-LSTM model, and performing bidirectional output related to the multiplication of the zoom point by the attention on the historical duty ratio time sequence data set based on a double-head attention model to determine a first output vector and a second output vector; wherein the dual-headed attention model is contained in the Bi-LSTM model; splicing the first output vector and the second output vector to obtain a spliced vector; and carrying out linear mapping on the splicing vector, and outputting and obtaining the predicted data of the outlet flow rate ratio of the target charging station.
In a possible implementation manner, calculating a flow ratio of the target charging station with respect to the cell outlet flow, and determining a station outlet flow of the target charging station specifically includes: and carrying out flow prediction on the outlet flow of the target charging station based on the cell outlet flow of the target cell and the flow ratio prediction data of the target charging station, and determining the station outlet flow.
In a possible embodiment, according toObtaining the site outlet flow/>, of the target charging site; Wherein/>The cell outlet flow of the target cell at the moment T+1; /(I)And the time T+1 is the predicted value of the flow ratio of the target charging station.
On the other hand, the embodiment of the application also provides a flow prediction device for the exit of the layered highway toll station, which comprises the following components: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method of predicting traffic at a layered highway toll gate exit as described in any one of the embodiments above.
Compared with the prior art, the embodiment of the application has the following beneficial technical effects:
the embodiment of the application groups expressway toll stations, each group is called a toll station cell, then carries out exit high-speed flow prediction by taking the cell as a unit, and finally further obtains the exit flow prediction result of a single toll station in the cell based on the exit flow prediction result of the cell. The method is beneficial to the user to select the optimal expressway toll gate, so that the user can select the optimal toll gate outlet in the adjacent space area, the accuracy of traffic flow prediction data in the adjacent space area is improved, and the travel efficiency of the user is further improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to the drawings without inventive effort to those skilled in the art. In the drawings:
FIG. 1 is a flow chart of a flow prediction method for a layered highway toll station exit provided by an embodiment of the present application;
FIG. 2 is a schematic diagram of a CNN_ (TCN+ BiGRU) _Attention model according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a flow rate duty ratio model according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a flow prediction device for a layered highway toll station exit according to an embodiment of the present application.
Detailed Description
In order to make the technical solution of the present application better understood by those skilled in the art, the technical solution of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
The embodiment of the application provides a flow prediction method for a layered expressway toll station outlet, which specifically comprises the following steps of S101-S105:
Note that, the toll station cell: a combination of several high-speed toll stations that are spatially close together and have similar trends in the exit traffic is called a toll station cell. Cell exit traffic: the sum of all toll station exit flows in the toll station cell. Cell entry traffic: the sum of all toll station ingress traffic in the toll station cell.
S101, based on a preset highway network model, similarity calculation under relevant dynamic and static characteristics is carried out among highway network nodes, and network node similarity is obtained. Wherein the highway network node is a highway toll station.
Specifically, a highway network model needs to be built based on a network topology structure of the highway network. A complex network topology structure constructed by utilizing a highway network: ; wherein/> Representing a collection of highway network nodes (toll stations), H being the number of highway network nodes; /(I)For a connection matrix of highway nodes, if there is a directly connected road between network node i and network node j, then/>Otherwise, then/>；/>Is a node similarity matrix of the expressway network. /(I)Representing the similarity between the i-th node and the j-th node.
Further, it can be according toObtaining the spatial neighbor similarity/>, between the nodes of the expressway network. Wherein the longitude and latitude of the ith node are as follows; Longitude and latitude of the j-th node are/>；/>And the Europe distance between the longitude and latitude of the ith node and the longitude and latitude of the jth node in the highway network model is represented.
Further according toObtaining the network structure similarity/>, between the nodes of the expressway network. Wherein/>For the neighbor node set of node i,/>A neighbor node set of the node j; /(I)Representing the number of common neighbor nodes of node i and node j in the highway network model; /(I)The degree of node j; /(I)The degree of node i; /(I)The eccentricity of the node i in the highway network model is represented as the maximum distance in the distance from the node i to any other point in the network; /(I)Is the eccentricity of node j in the highway network model and represents the maximum distance in distance of node j from any other point in the network.
Further according toObtaining the outlet flow time sequence similarity/>, between the nodes of the expressway network. Wherein/>Is the i node historical annual sunrise traffic vector, and/>，/>{1, 2..365 } Represents egress traffic on day k of the ith node; /(I)Is the historical annual sunrise traffic vector of the jth node, and，/>{1, 2..365 } Represents egress traffic on day k of node j.
As a possible implementation mode, the spatial neighbor similarity and the network structure similarity belong to static characteristics, and the outlet flow time sequence similarity belongs to dynamic characteristics.
Further according toObtaining the similarity/>, of network nodes among expressway network nodes. Wherein/>Weighting spatial neighbor similarity,/>Is the weight of the similarity of network structures,/>The output traffic timing similarity is weighted.
S102, screening the similarity of the network nodes, and determining a toll station cell according to the screening result. The toll station cell is a network node set under high similarity.
Specifically, a node similarity matrix is established based on network node similarity.
Further according toObtaining a node similarity matrix after screening. Wherein/>Similarity is network node; /(I)Is a preset similarity threshold value, and/>Is a 50% quantile of the set of elements contained in the node similarity matrix.
Further, according to a preset Louvain algorithm (community discovery algorithm based on modularity), the screened node similarity matrix is subjected to modularity maximization, and node combination is performed on the node similarity matrix with the modularity maximized and promoted, so that a network node set is obtained. Finally, the set of network nodes is determined as a toll station cell. The toll station cell comprises a plurality of expressway toll stations which are in spatial neighbors and have similar outlet flow change trend.
And S103, combining the preset neural network models to obtain an outlet flow prediction model based on the expressway toll station.
Specifically, the output end in the full connection layer in the preset 1D-CNN model is respectively connected with the input end of the preset TCN network model and the input end of the preset BiGRU network model.
Further, the output end of the TCN network model and the output end of the BiGRU network model are connected with the input end of the preset attention mechanism model.
Further, an outlet flow prediction model is combined based on the 1D-CNN model, the TCN network model, the BiGRU network model and the attention mechanism model.
The expressway is a closed road network system, and vehicles enter the expressway network from an entrance toll station and exit from an exit toll station. Essentially, the exit traffic of the toll gate cell depends on the entrance traffic of some cells in the reference cell set. On the one hand, the traffic OD of the highway network determines which cells in the reference cell set have inlet flows determining the outlet flows of the toll station cells; on the other hand, due to the presence of the travel time, there is a time lag depending on the ingress traffic and the egress traffic, and there is a distribution diversity of time lags between different ingress traffic and egress traffic of the toll station cell.
In one embodiment, fig. 2 is a schematic diagram of a cnn_ (tcn+ BiGRU) _attention model according to an embodiment of the present application, where the total number of the outlet flow prediction models is divided into 4 modules as shown in fig. 2. A modified 1D-CNN (one-dimensional convolutional) module, a TCN network module (Temporal Convolutional Network, a convolutional neural network for handling time series problems), a BiGRU (Bidirectional Gated Recurrent Unit, a bi-directional gating cyclic unit) network module, and an attention mechanism module, respectively.
1) The improved 1D-CNN module (1D-CNN model) is used for deep mining of rules of input flow of the reference cell and is responsible for fusing historical inlet flow information of all the reference cells. The input of the full connection layer of the improved 1D-CNN module is used as the input of the TCN network module and the BiGRU network module.
2) The TCN network module (TCN network model) expands the receptive field of the convolution kernel by expanding convolution, and realizes long-term and short-term dependence of input flow of different reference cells and output flow of a target cell. BiGRU the network module can better capture the bidirectional dependency relationship in the time series data.
3) The attention mechanism module (the intention mechanism model) takes the combination of the outputs of the TCN module and the BiGRU module (the BiGRU network model) as input, and utilizes weight distribution to more accurately find the causal relationship between the flow of each reference cell at a certain historical moment and the flow of the target cell at the next moment.
S104, carrying out data prediction processing of the outlet flow of the target cell in the toll station cell through an outlet flow prediction model to obtain the cell outlet flow of the target cell.
Specifically, firstly, dividing a cell set of a toll station cell based on the outlet flow direction requirement of a toll station of an expressway to obtain a target cell and a reference cell set. The target cell is a toll station cell of which the exit quantity is to be predicted in a plurality of toll station cells. The reference cell set is a remaining set of toll station cells of the number of toll station cells.
Further, the historical inlet flow data corresponding to each reference cell in the reference cell set is obtained, and each historical moment flow data of the historical inlet flow data of each reference cell is marked.
Further, through a 1D-CNN module in the outlet flow prediction model, carrying out data fusion processing on the historical inlet flow data corresponding to each reference cell; and performing data superposition fusion under convolution pooling on the fused historical inlet flow data to obtain superposition fusion feature vectors based on the historical inlet flow data.
In one embodiment, as shown in FIG. 2, with the improved 1D-CNN network architecture, the convolution layer convolves the incoming traffic timing data of the historical ingress traffic data with a set of one-dimensional convolution kernels, respectively, to slidingly extract local features of the timing data in one dimension. A convolution kernel size of 3 and a number of convolution kernels of 8 may be used. The pooling layer integrates the local features extracted by the convolution layer, so that feature parameters are reduced, and the calculated amount is reduced. And further, an average pooling mode is used, so that more input traffic information is reserved. And meanwhile, overlapping and fusing convolution-pooling results of the time sequence data output by the cells at the pooling layer, and sending the overlapping and pooling results to the full-connection layer together. The full connection layer integrates the feature vectors of all cells into a superposition fusion feature vector. The output layer serves as input to the TCN network module and BiGRU network modules.
Further, residual error connection processing is carried out on causal convolution and expansion convolution related to the superposition fusion feature vector through a TCN network module in the outlet flow prediction model, so that TCN network output features are obtained.
In one implementation, as shown in fig. 2, the TCN network module includes 3 layers of residual modules, each layer of residual module changes an activation function into GeLU functions based on an original TCN residual module, so that smoothness of a model is improved, two causal convolutions and expansion convolutions are combined with identity mapping to form a residual module, and then a depth network structure is formed by overlapping a plurality of residual modules. Wherein the convolution kernel size is 3*1, the expansion coefficients of all layers are (1, 2,4,8, 16) in sequence, and the residual channel input passes through one 1*1 convolution kernel convolution layer, so that the tensor dimension (TCN network output characteristic) of the input and the output is kept consistent. The expansion convolution expands the receptive field by increasing the expansion coefficient, so that the relationship between the outlet flow of the target cell at the current moment and the inlet flow of the reference cell with a large travel time can be effectively captured. Residual connection can effectively overcome the gradient vanishing problem caused by the field of view and the dilation convolution.
Further, through BiGRU network modules in the outlet flow prediction model, the overlapped and fused feature vectors are subjected to bidirectional gating circulation processing of the relevant forward layer and the backward layer, and time sequence output features based on the BiGRU network are obtained.
Further, according to the attention mechanism module in the outlet flow prediction model, weighting is given to TCN network output characteristics and time sequence output characteristics, and feedback learning is carried out on the 1D-CNN module according to the flow input-output dependency relationship between the historical inlet flow data and the flow data of the target cell after weighting is given, so that the outlet flow prediction model after deep learning is obtained.
In one embodiment, as shown in fig. 2, the attention mechanism weights the fusion features of the TCN network and BiGRU network outputs differently, i.e., weights the TCN network output features and the timing output features, the magnitude of the weights essentially corresponds to a stronger traffic input-output dependency. And further improves the prediction accuracy of the exit flow of the target cell. The Attention module learns the characteristic weight vector in the input vector, combines the characteristic weight vector with the input vector of the layer and transmits the combined characteristic weight vector to the full-connection layer, and the feedback learning of the 1D-CNN module is completed to obtain the deep-learned outlet flow prediction model.
Further, based on the outlet flow prediction model after deep learning, the outlet flow data of the target cell is predicted, and the cell outlet flow of the target cell is obtained.
In one embodiment, a collection of toll station cellsThe n+1 cells in (a) are divided into two parts. 1 cell whose exit traffic needs to be predicted, called target cell, is denoted/>. The remaining N cells are all called reference cells, the set of all reference cells is called reference cell set, and is denoted/>. The cell exit traffic prediction problem can be described as: knowing the historical ingress traffic data for N reference cells, noted asWherein/>Is historical traffic data at time 1~T of the nth reference cell. Prediction/>Time of day target cell/>Is denoted as/>：; Wherein/>Is the outlet flow prediction model after deep learning.
S105, calculating the flow rate ratio of the target charging station for the cell outlet flow rate, and determining the station outlet flow rate of the target charging station.
Specifically, it is first necessary to acquire the historical cell exit traffic of the target cell and the historical site exit traffic of the target charging site.
Further, the time sequence data duty ratio calculation is carried out on the historical cell outlet flow and the historical site outlet flow, and a historical duty ratio time sequence data set is obtained. And inputting the historical duty ratio time sequence data set into a preset Bi-LSTM model. The Bi-LSTM model is a neural network model added with a double-head attention mechanism.
Further, linear mapping processing is carried out on the historical duty ratio time sequence data set through the Bi-LSTM model, bidirectional output related to zooming point multiplied attention is carried out on the historical duty ratio time sequence data set based on the double-head attention model, and a first output vector and a second output vector are determined. Wherein the dual-head attention model is contained in the Bi-LSTM model.
Further, the first output vector and the second output are spliced to obtain a spliced vector. And performing linear mapping on the spliced vector, and outputting and obtaining the predicted data of the outlet flow ratio of the target toll station.
In one embodiment, fig. 3 is a schematic structural diagram of a traffic ratio model provided in the embodiment of the present application, as shown in fig. 3, each toll station has a time-varying rule in the proportion of the traffic of the cell exit, and the historic data at different moments has different importance in prediction. Thus, the prediction of the traffic ratio is achieved using the Bi-LSTM model incorporating the self-attention mechanism. It is necessary to time-sequence the historical duty cycle data setAnd sending the data into a Bi-LSTM layer, extracting information in the duty ratio data through a forward LSTM layer and a backward LSTM layer, and splicing output results and sending the output results into a double-head self-attention model. In each head, the data is duplicated in triplicate, mapped linearly, and then passed through a scaled Dot-by-attention (scaled Dot-Product Attention) mechanism as output of the dual-head attention mechanism. And then splicing the output vectors (the first output vector and the second output vector) of the two heads to obtain a spliced vector, and then performing linear mapping again to finally obtain a flow ratio prediction result.
As a possible implementation, the target toll stationIs in the target cellDuty cycle timing data in historical outlet traffic of/>Wherein/>Indicated at time tIs at/>And T represents the length of the historical outlet flow rate duty cycle sequence. Prediction/>Time/>Outlet flow is at/>Flow ratio predictive value/>, in outlet flow. The predictive formula is: /(I); Wherein/>Is a traffic ratio prediction model of the target toll station, namely a Bi-LSTM model.
Further, based on the predicted data of the cell outlet flow of the target cell and the flow ratio of the target charging station, the outlet flow of the target charging station is predicted, and the station outlet flow is determined.
As a possible embodiment, according toObtaining the site outlet flow/>, of the target charging site. Wherein/>The cell outlet flow of the target cell at the moment T+1; /(I)And the time T+1 is the predicted value of the flow ratio of the target toll station.
In addition, the embodiment of the application also provides a flow prediction device for the exit of the layered expressway toll station, as shown in fig. 4, the flow prediction device 400 for the exit of the layered expressway toll station specifically comprises:
At least one processor 401; and a memory 402 communicatively coupled to the at least one processor 401; wherein the memory 402 stores instructions executable by the at least one processor 401 to enable the at least one processor 401 to perform:
Based on a preset highway network model, similarity calculation under relevant dynamic and static characteristics is carried out among highway network nodes, and network node similarity is obtained; wherein the highway network node is a highway toll station;
screening the similarity of the network nodes, and determining a toll station cell according to the screening result; the toll station cell is a network node set under high similarity;
Model combination is carried out on a plurality of preset neural network models to obtain an outlet flow prediction model based on the expressway toll station;
carrying out data prediction processing of the outlet flow on a target cell in the toll station cell through an outlet flow prediction model to obtain the cell outlet flow of the target cell;
And calculating the flow rate ratio of the target charging station relative to the cell outlet flow rate, and determining the station outlet flow rate of the target charging station.
The application provides a method and equipment for predicting the flow of an exit of a layered expressway toll station, which are characterized in that the expressway toll stations are grouped, each group is called a toll station cell, then the exit high-speed flow prediction is carried out by taking the cell as a unit, and finally the exit flow prediction result of a single toll station in the cell is further obtained based on the exit flow prediction result of the cell. The method is beneficial to the user to select the optimal expressway toll gate, so that the user can select the optimal toll gate outlet in the adjacent space area, the accuracy of traffic flow prediction data in the adjacent space area is improved, and the travel efficiency of the user is further improved.
The embodiments of the present application are described in a progressive manner, and the same and similar parts of the embodiments are all referred to each other, and each embodiment is mainly described in the differences from the other embodiments. In particular, for the apparatus embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and reference is made to the description of the method embodiments in part.
The foregoing describes certain embodiments of the present application. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the embodiments of the application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A method for predicting traffic at an exit of a layered highway toll station, the method comprising:
Based on a preset highway network model, similarity calculation under relevant dynamic and static characteristics is carried out among highway network nodes, and network node similarity is obtained; wherein the highway network node is a highway toll station;
screening the similarity of the network nodes, and determining a toll station cell according to a screening result; wherein the toll station cell is a network node set under high similarity;
Model combination is carried out on a plurality of preset neural network models to obtain an outlet flow prediction model based on the expressway toll station;
Carrying out data prediction processing of the outlet flow of a target cell in the toll station cell through the outlet flow prediction model to obtain the cell outlet flow of the target cell;
and calculating the flow rate ratio of the target charging station according to the cell outlet flow rate, and determining the station outlet flow rate of the target charging station.
2. The method for predicting the traffic of the exit of the layered highway toll station according to claim 1, wherein the similarity calculation under the relevant dynamic and static characteristics is performed between the nodes of the highway network based on a preset highway network model to obtain the similarity of the network nodes, specifically comprising:
establishing a highway network model based on a network topology structure of a highway network;
According to Obtaining the spatial neighbor similarity/>, between the expressway network nodes; Wherein the longitude and latitude of the ith node are/>; Longitude and latitude of the j-th node are/>；/>Representing the Euclidean distance between the longitude and latitude of the ith node and the longitude and latitude of the jth node in the highway network model;
According to Obtaining the network structure similarity/>, between the expressway network nodes; Wherein/>For the neighbor node set of node i,/>A neighbor node set of the node j; Representing the number of common neighbor nodes of node i and node j in the highway network model; The degree of node j; /(I) The degree of node i; /(I)The eccentricity of the node i in the highway network model is represented as the maximum distance in the distance from the node i to any other point in the network; /(I)The eccentricity of the node j in the highway network model is represented as the maximum distance from the node j to any other point in the network;
According to Obtaining the outlet flow time sequence similarity/>, between the expressway network nodes; Wherein/>Is the i node historical annual sunrise traffic vector, and/>，/>{1, 2..365 } Represents egress traffic on day k of the ith node; /(I)Is the historical annual sunrise traffic vector of the jth node, and，/>{1,2,..365 } Represents egress traffic on day k of node j;
wherein the spatial neighbor similarity and the network structure similarity all belong to static features; the outlet flow time sequence similarity belongs to dynamic characteristics;
According to Obtaining the similarity/>, of network nodes among the expressway network nodes; Wherein/>Weighting the spatial neighbor similarity,/>Weight of similarity of the network structure,/>And weighting the time sequence similarity of the outlet flow.
3. The method for predicting traffic at the exit of a layered highway toll station according to claim 1, wherein the screening of the similarity is performed on the network node similarity, and the toll station cell is determined according to the screening result, and specifically comprises:
Based on the network node similarity, a node similarity matrix is established;
According to Obtaining a node similarity matrix/>, after screening; Wherein/>Similarity for the network node; /(I)Is a preset similarity threshold value, and/>50% Quantiles for a set of elements contained in the node similarity matrix;
carrying out modular degree maximization processing on the node similarity matrix after screening according to a preset Louvain algorithm, and carrying out node merging on the node similarity matrix after modular degree maximization promotion to obtain the network node set;
determining the set of network nodes as the toll station cell; the toll station cell comprises a plurality of expressway toll stations which are in spatial neighbors and have similar outlet flow change trends.
4. The method for predicting the traffic of the exit of the layered highway toll station according to claim 1, wherein model combination is performed on a plurality of preset neural network models to obtain an exit traffic prediction model based on the highway toll station, and the method specifically comprises the following steps:
respectively connecting the output end in the full connection layer in the preset 1D-CNN model with the input end of the preset TCN network model and the input end of the preset BiGRU network model;
The output end of the TCN network model and the output end of the BiGRU network model are connected with the input end of a preset attention mechanism model;
The outlet flow prediction model is combined based on the 1D-CNN model, the TCN network model, the BiGRU network model, and the attention mechanism model.
5. The method for predicting traffic at an exit of a layered highway toll station according to claim 1, wherein before performing data prediction processing of exit traffic for a target cell in the toll station cell by the exit traffic prediction model to obtain cell exit traffic of the target cell, the method further comprises:
Dividing a cell set of a toll station cell based on the outlet flow direction requirement of a toll station of an expressway to obtain the target cell and a reference cell set; the target cell is a plurality of toll station cells of which the exit quantity is to be predicted; the reference cell set is a remaining toll station cell set in a plurality of toll station cells;
Acquiring historical inlet flow data corresponding to each reference cell in the reference cell set; and marking each historical time of day traffic data for the historical ingress traffic data for each reference cell.
6. The method for predicting traffic of layered expressway toll station exit of claim 5, wherein said predicting traffic data of a target cell in said toll station cell by said exit traffic prediction model, to obtain said target cell exit traffic, comprises:
performing data fusion processing on the historical inlet flow data corresponding to each reference cell through a 1D-CNN module in the outlet flow prediction model; the data under convolution pooling is overlapped and fused on the fused historical inlet flow data, and an overlapped and fused feature vector based on the historical inlet flow data is obtained;
Performing residual error connection processing on the causal convolution and the expansion convolution related to the superposition fusion feature vector by a TCN network module in the outlet flow prediction model to obtain TCN network output features;
Performing bidirectional gating cyclic processing on the relevant forward layer and backward layer by using the BiGRU network module in the outlet flow prediction model to obtain a time sequence output characteristic based on BiGRU network;
According to the attention mechanism module in the outlet flow prediction model, giving weights to the TCN network output characteristics and the time sequence output characteristics, and according to the flow input-output dependency relationship between the historical inlet flow data and the flow data of the target cell after the weights are given, performing feedback learning on the 1D-CNN module to obtain an outlet flow prediction model after deep learning;
And carrying out outlet flow data prediction on the target cell based on the outlet flow prediction model after deep learning to obtain the cell outlet flow of the target cell.
7. A method for predicting traffic at a layered highway toll station exit according to claim 1 wherein prior to calculating traffic duty cycle for a target toll station for said cell exit traffic and determining station exit traffic for said target toll station, said method further comprises:
acquiring the historical cell outlet flow of the target cell and the historical site outlet flow of the target charging site;
Performing time-series data duty ratio calculation on the historical cell outlet flow and the historical site outlet flow to obtain a historical duty ratio time-series data set;
Inputting the historical duty ratio time sequence data set into a preset Bi-LSTM model; wherein the Bi-LSTM model is a neural network model added with a double-head attention mechanism;
Performing linear mapping processing on the historical duty ratio time sequence data set through the Bi-LSTM model, and performing bidirectional output related to the multiplication of the zoom point by the attention on the historical duty ratio time sequence data set based on a double-head attention model to determine a first output vector and a second output vector; wherein the dual-headed attention model is contained in the Bi-LSTM model;
Splicing the first output vector and the second output vector to obtain a spliced vector; and carrying out linear mapping on the splicing vector, and outputting and obtaining the predicted data of the outlet flow rate ratio of the target charging station.
8. The method for predicting traffic of layered highway toll station exit according to claim 7, wherein calculating traffic duty ratio of a target toll station for the cell exit traffic to determine station exit traffic of the target toll station comprises:
And carrying out flow prediction on the outlet flow of the target charging station based on the cell outlet flow of the target cell and the flow ratio prediction data of the target charging station, and determining the station outlet flow.
9. The method for predicting traffic at the exit of a layered highway toll station according to claim 8, wherein the method comprises the steps ofObtaining the site outlet flow/>, of the target charging site; Wherein/>The cell outlet flow of the target cell at the moment T+1; /(I)And the time T+1 is the predicted value of the flow ratio of the target charging station.
10. A flow prediction apparatus for a layered highway toll gate outlet, the apparatus comprising:
at least one processor; and
A memory communicatively coupled to the at least one processor; wherein,
The memory stores instructions executable by the at least one processor to enable the at least one processor to perform a layered highway toll gate exit flow prediction method according to any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410544141.3A CN118135799A (en) | 2024-05-06 | 2024-05-06 | Flow prediction method and equipment for layered expressway toll station outlet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410544141.3A CN118135799A (en) | 2024-05-06 | 2024-05-06 | Flow prediction method and equipment for layered expressway toll station outlet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118135799A true CN118135799A (en) | 2024-06-04 |
Family
ID=91236234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410544141.3A Pending CN118135799A (en) | 2024-05-06 | 2024-05-06 | Flow prediction method and equipment for layered expressway toll station outlet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118135799A (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513370A (en) * | 2015-12-29 | 2016-04-20 | 浙江大学 | Traffic zone dividing method based on sparse vehicle license identification data |
| CN111260927A (en) * | 2020-05-07 | 2020-06-09 | 北京航空航天大学 | Road network flow prediction method |
| US20220190596A1 (en) * | 2020-12-11 | 2022-06-16 | Korea Institute Of Science And Technology | Method for predicting power generation and remaining useful life per system and system for performing the same |
| CN115442253A (en) * | 2022-09-15 | 2022-12-06 | 西安电子科技大学 | Network flow prediction method using attention mechanism |
| CN115578852A (en) * | 2022-07-14 | 2023-01-06 | 西北师范大学 | Traffic prediction method based on DSTGCN |
| CN116052421A (en) * | 2023-01-30 | 2023-05-02 | 东北大学 | Traffic flow prediction method based on static and dynamic multi-graph fusion |
| CN116486624A (en) * | 2023-04-24 | 2023-07-25 | 山东大学 | Traffic flow prediction method and system based on space-time diagram convolutional neural network |
| CN116885691A (en) * | 2023-05-30 | 2023-10-13 | 淮阴工学院 | Wind power climbing event indirect prediction method |
| CN117744705A (en) * | 2023-11-17 | 2024-03-22 | 南京熊猫电子股份有限公司 | Short-time station outbound passenger flow prediction method based on self-attention mechanism and TCN-BiGRU model |
| CN117829201A (en) * | 2024-01-02 | 2024-04-05 | 京上智数科技(北京)有限公司 | Network abnormal flow detection method based on mixed space-time neural network |
-
2024
- 2024-05-06 CN CN202410544141.3A patent/CN118135799A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513370A (en) * | 2015-12-29 | 2016-04-20 | 浙江大学 | Traffic zone dividing method based on sparse vehicle license identification data |
| CN111260927A (en) * | 2020-05-07 | 2020-06-09 | 北京航空航天大学 | Road network flow prediction method |
| US20220190596A1 (en) * | 2020-12-11 | 2022-06-16 | Korea Institute Of Science And Technology | Method for predicting power generation and remaining useful life per system and system for performing the same |
| CN115578852A (en) * | 2022-07-14 | 2023-01-06 | 西北师范大学 | Traffic prediction method based on DSTGCN |
| CN115442253A (en) * | 2022-09-15 | 2022-12-06 | 西安电子科技大学 | Network flow prediction method using attention mechanism |
| CN116052421A (en) * | 2023-01-30 | 2023-05-02 | 东北大学 | Traffic flow prediction method based on static and dynamic multi-graph fusion |
| CN116486624A (en) * | 2023-04-24 | 2023-07-25 | 山东大学 | Traffic flow prediction method and system based on space-time diagram convolutional neural network |
| CN116885691A (en) * | 2023-05-30 | 2023-10-13 | 淮阴工学院 | Wind power climbing event indirect prediction method |
| CN117744705A (en) * | 2023-11-17 | 2024-03-22 | 南京熊猫电子股份有限公司 | Short-time station outbound passenger flow prediction method based on self-attention mechanism and TCN-BiGRU model |
| CN117829201A (en) * | 2024-01-02 | 2024-04-05 | 京上智数科技(北京)有限公司 | Network abnormal flow detection method based on mixed space-time neural network |
Non-Patent Citations (6)
| Title |
|---|
| CHANG, ZH等: "STA-GCN: Spatial-Temporal Self-Attention Graph Convolutional Networks for Traffic-Flow Prediction", APPLIED SCIENCES-BASEL, 2 June 2023 (2023-06-02), pages 1 - 14 * |
| JING BI等: "Web Traffic Anomaly Detection Using a Hybrid Spatio-temporal Neural Network", 2023 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC), 29 January 2024 (2024-01-29), pages 5009 - 5014 * |
| LEI LIU等: "Multilevel Federated Learning-Based Intelligent Traffic Flow Forecasting for Transportation Network Management", IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, 29 May 2023 (2023-05-29), pages 1446 - 1458, XP011944010, DOI: 10.1109/TNSM.2023.3280515 * |
| 寇敏等: "融合天气特征的WOA-GRU高速公路短时交通流预测", 公路, 30 April 2024 (2024-04-30), pages 193 - 199 * |
| 林靖皓;秦亮曦;苏永秀;秦川;: "基于自注意力机制的双向门控循环单元和卷积神经网络的芒果产量预测", 计算机应用, no. 1, 10 July 2020 (2020-07-10), pages 56 - 60 * |
| 阴艳超等: "融合多通道CNN-BiGRU与时间模式注意力机制的多工序工艺质量预测方法", 计算机集成制造系统, 2 August 2023 (2023-08-02), pages 1 - 19 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Contextualized spatial–temporal network for taxi origin-destination demand prediction | |
| Xu et al. | Spatial-temporal transformer networks for traffic flow forecasting | |
| Mallick et al. | Transfer learning with graph neural networks for short-term highway traffic forecasting | |
| Zhang et al. | Long-term mobile traffic forecasting using deep spatio-temporal neural networks | |
| CN111383452A (en) | Method for estimating and predicting short-term traffic running state of urban road network | |
| CN110570651A (en) | Road network traffic situation prediction method and system based on deep learning | |
| CN110543543A (en) | user movement behavior prediction method and device based on multi-granularity neural network | |
| CN114299728B (en) | Vehicle flow prediction method combining attention mechanism and dynamic space-time convolution model | |
| CN112949931B (en) | Method and device for predicting charging station data by mixing data driving and models | |
| Huang et al. | A GAN framework-based dynamic multi-graph convolutional network for origin–destination-based ride-hailing demand prediction | |
| Ni et al. | STGMN: A gated multi-graph convolutional network framework for traffic flow prediction | |
| CN114925836B (en) | Urban traffic flow reasoning method based on dynamic multi-view graph neural network | |
| CN105933857A (en) | Mobile terminal position prediction method and apparatus | |
| Qiu et al. | Taxi origin-destination demand prediction with contextualized spatial-temporal network | |
| Su et al. | Operating status prediction model at EV charging stations with fusing spatiotemporal graph convolutional network | |
| CN115206092A (en) | Traffic prediction method of BiLSTM and LightGBM model based on attention mechanism | |
| Zhang et al. | Trajgail: Trajectory generative adversarial imitation learning for long-term decision analysis | |
| Storani et al. | A traffic responsive control framework for signalized junctions based on hybrid traffic flow representation | |
| CN116823572B (en) | Population flow data acquisition method and device and computer readable storage medium | |
| CN118135799A (en) | Flow prediction method and equipment for layered expressway toll station outlet | |
| Sun et al. | Ada-STNet: A Dynamic AdaBoost Spatio-Temporal Network for Traffic Flow Prediction | |
| He et al. | Dyna-PTM: OD-enhanced GCN for metro passenger flow prediction | |
| Lee et al. | Relocation Framework to Improve the Usage Efficiency of Bicycle-sharing Systems | |
| Xing et al. | GRU‐CNN Neural Network Method for Regional Traffic Congestion Prediction Serving Traffic Diversion Demand | |
| Emami et al. | Rc-tl: Reinforcement convolutional transfer learning for large-scale trajectory prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |