A traffic network bottleneck identification method based on flow cascade

Abstract

The application discloses a traffic network bottleneck identification method based on traffic cascade in the technical field of traffic information control, which comprises the following steps: acquiring road data and commuting behavior data in a target area, extracting the midpoint of each road in the road data, connecting the start and end points of the commuting behavior data to the midpoint of the nearest road, and recording the corresponding index relationship between the start and end points and the midpoint of the road. The application uses an incremental traffic assignment algorithm to assign the trip matrix of the spatial position information, simulates the congestion effect generated by the network traffic cascade by iteratively updating the network edge weight, identifies the network bottleneck under the network traffic reconstruction, can systematically depict the cascade conduction mechanism of the traffic network traffic under the disturbance condition, identify the network bottleneck which has a key influence on the system stability, and improve the response ability and structural resilience of the city risk.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic information control, and particularly relates to a traffic network bottleneck identification method based on traffic cascading. BACKGROUND

[0002] Under the background of climate change intensification and frequent occurrence of various risk events, how to accurately quantify and improve urban resilience has become a core issue of urban governance and planning intervention. As a typical complex system, the city is often abstracted as a network structure to help researchers understand its internal dynamic mechanism. Network science thus provides a basic research paradigm for resilience measurement and evaluation. In the classical method of evaluating the resilience of urban networks, researchers model the city system by setting rules for nodes and edges, and introduce attack strategies to simulate the damage of network structure. The resilience or robustness of the system is measured by evaluating the changes in performance indicators such as network efficiency.

[0003] Among many urban systems, the transportation network, as the core spatial carrier of material circulation, has always been the focus of network resilience research. However, existing research often fails to accurately depict the dynamic response process of the transportation network when responding to external shocks, and it is also difficult to identify key spatial elements to support the needs of precise governance. Although the transportation field has made significant progress in individual travel behavior simulation and traffic flow dynamic modeling, existing research usually focuses on the micro-evolution process of traffic flow, lacks analysis of the operational status of the city transportation system from the perspective of network structure and resilience mechanism, and lacks identification of key bottleneck sections and cascading propagation paths in the network, making it difficult to provide targeted structural intervention basis for the city system. SUMMARY

[0004] The purpose of the present application is to provide a traffic network bottleneck identification method based on traffic cascading, which provides spatial location information of O-D pairs through traffic travel data, uses an incremental traffic assignment algorithm to assign the commuting O-D travel matrix, iteratively updates the network edge weight through multiple traffic assignments, simulates the congestion effect caused by network traffic cascading, and identifies the network bottleneck under network traffic reconstruction, thereby improving the response capacity and structural resilience of the city risk.

[0005] To achieve the above purpose, the present application provides the following technical scheme:

[0006] A traffic network bottleneck identification method based on traffic cascading, comprising:

[0007] S1: Obtain road data and commuting behavior data in a target area, wherein the road data includes free-flow travel time t0 of each road, capacity C of the road, and spatial location information of the road, and the commuting behavior data includes spatial location information of commuting O-D and commuting traffic corresponding to the O-D pairs;

[0008] S2: extracting the midpoint of each road in the road data, connecting the O-D pair space of the commuting behavior data to the nearest road midpoint, and recording the corresponding index relationship between the O-D pair and the road midpoint;

[0009] S3: establishing a first road network G1(V,E) by using an initial modeling method, wherein V is a first network node, V is a set composed of each intersection of a road, E is a first edge, E is a set composed of each road, each edge in E corresponds to the free-flow travel time t0 and the capacity C of the road, and the free-flow travel time t0 is taken as the weight of E;

[0010] S4: establishing a second road network G2(V,E) by using a dual modeling method, wherein V is a second network node, V is a set composed of roads, E is a second edge, and E is a set of edges composed of each intersection of a road;

[0011] S5: dividing the commuting flow data corresponding to each O-D pair in the commuting behavior data into a first part of flow, a second part of flow, a third part of flow and a fourth part of flow by using an incremental traffic assignment algorithm, and readjusting the network edge weight t of each network edge of the second road network G2(V,E);

[0012] S6: outputting the flow V of each O-D pair, taking the flow V of the O-D pair as a road flow distribution result FM0 of the incremental traffic assignment, removing a certain proportion of road segment sets and a certain proportion of network node sets corresponding to the road segment sets from G2(V,E), representing the removed network as a third road network G3(V,E), and initializing the node attributes and edge weights of the third road network G3(V,E);

[0013] S7: according to the index relationship between the O-D pair in the commuting behavior data and the original road segment space, screening the road segments in the removed network as index data of the O-D pair to obtain index removal data, and eliminating the index removal data in the incremental traffic assignment, executing the incremental traffic assignment algorithm on the third road network G3(V,E) based on the removed commuting behavior data, outputting a road flow distribution result FM1 after a network attack, and identifying a bottleneck based on the road flow distribution result FM1.

[0014] As a further scheme of the application, the second road network G2(V,E) established by using the dual modeling method, wherein V is a second network node, V is a set composed of roads, E is a second edge, and E is a set of edges composed of each intersection of a road, comprises:

[0015] S41: Extract each road i and the set of roads connected to each road i i_link from the first road network G1(V, E), select each connected road j in the road set i_link by iteration, input the road i and the connected road j into the set second road network G2(V, E) in turn as the second network node, and input the free-flow travel time t0 and the capacity C of the road i and the connected road j in the first road network G1(V, E) into the set second road network G2(V, E).

[0016] As a further scheme of the present application: the second road network G2(V, E) is established by using a dual modeling method, wherein V is the second network node, V is a set of roads, E is the second edge, and E is a set of edges composed of each intersection of the road, and further comprising:

[0017] S42: Network edge data between the road i and the connected road j is established, and the network edge data is input into the set second road network G2(V, E), the average value of the free-flow travel time t0 of the road i and the connected road j is taken as the edge weight t, and step S41 is repeated until all roads are input as i into the set second road network G2(V, E), and the duplicate nodes and edges of the set second road network G2(V, E) are removed to obtain the dual network G2'(V', E').

[0018] As a further scheme of the present application: the second road network G2(V, E) is established by using a dual modeling method, wherein V is the second network node, V is a set of roads, E is the second edge, and E is a set of edges composed of each intersection of the road, and further comprising:

[0019] S43: An attribute V is added to each network node of the dual network G2'(V', E'), and the initial value of the carrying capacity of the road section of the dual network G2'(V', E') is set to 0 to obtain the second road network G2(V, E), wherein the network node of the second road network G2(V, E) is a road section, and the edge of the second road network G2(V, E) is a road intersection.

[0020] As a further scheme of the present application: the commuting flow data corresponding to each pair of O-D pairs in the commuting behavior data is divided into a first part of flow, a second part of flow, a third part of flow and a fourth part of flow by using an incremental traffic assignment algorithm, and the network edge weight t of each network edge of the second road network G2(V, E) is adjusted again, wherein the first part of flow is 40% of the commuting flow data, the second part of flow is 30% of the commuting flow data, the third part of flow is 20% of the commuting flow data, and the fourth part of flow is 10% of the commuting flow data, and comprising:

[0021] S51: Indexes are contained in the midpoint of the O-D pair corresponding to the first part of the traffic flow, and the shortest path algorithm between two points is performed in the second road network G2(V, E), and the edge weight is set as t, and the returned data is the node list passed by the two points, wherein the node list passed is the road section.

[0022] As a further scheme of the application: the incremental traffic assignment algorithm is used to divide the commuting traffic data corresponding to each O-D pair in the commuting behavior data into first part traffic, second part traffic, third part traffic and fourth part traffic, and the network edge weight t of each network edge of the second road network G2(V, E) is adjusted again, wherein the assignment proportion of the commuting traffic data of the first part traffic is 40%, the assignment proportion of the commuting traffic data of the second part traffic is 30%, the assignment proportion of the commuting traffic data of the third part traffic is 20%, and the assignment proportion of the commuting traffic data of the fourth part traffic is 10%, and further comprising:

[0023] S52: For each node list passed by the O-D pair, multiply the traffic of the O-D pair by the assignment proportion of the corresponding commuting traffic data to obtain the assigned traffic, and sum the assigned traffic and the original traffic to obtain the new traffic, and distribute the new traffic to each node in the node list.

[0024] As a further scheme of the application: the incremental traffic assignment algorithm is used to divide the commuting traffic data corresponding to each O-D pair in the commuting behavior data into first part traffic, second part traffic, third part traffic and fourth part traffic, and the network edge weight t of each network edge of the second road network G2(V, E) is adjusted again, wherein the assignment proportion of the commuting traffic data of the first part traffic is 40%, the assignment proportion of the commuting traffic data of the second part traffic is 30%, the assignment proportion of the commuting traffic data of the third part traffic is 20%, and the assignment proportion of the commuting traffic data of the fourth part traffic is 10%, and further comprising:

[0025] S53: A time and traffic road resistance function is used to calculate the current actual travel time t a of the road section of each network node of the second road network G2(V, E) a As the new weight t a , wherein the time and traffic road resistance function is:

[0026] Where t0, C and V are the current node attributes, and α and β are preset parameters, α is 0.15, and β is 4.

[0027] As a further scheme of the present application: the incremental traffic assignment algorithm is used to divide the commuting flow data corresponding to each pair of O-D pairs in the commuting behavior data into first part flow, second part flow, third part flow and fourth part flow, and the network edge weight t of each network edge of the second road network G2(V, E) is readjusted, wherein the assignment proportion of the commuting flow data of the first part flow is 40%, the assignment proportion of the commuting flow data of the second part flow is 30%, the assignment proportion of the commuting flow data of the third part flow is 20%, and the assignment proportion of the commuting flow data of the fourth part flow is 10%, and further comprising:

[0028] S54: Based on the new weight t of each node at both ends of the network edge a , the network edge weight t is readjusted, wherein the network edge weight t is the average value of the weight t of the nodes at both ends. a

[0029] As a further scheme of the present application: the index data of the road segments in the network after removal is screened as the index data of the O-D pairs according to the index relationship between the O-D pairs in the commuting behavior data and the original road segments, index removal data is obtained, and the index removal data is excluded in the incremental traffic assignment, the incremental traffic assignment algorithm is executed on the commuting behavior data after exclusion in the third road network G3(V, E), and the road flow distribution result FM1 after network attack is output, and the bottleneck is identified based on the road flow distribution result FM1, comprising:

[0030] S101: The road segment nodes remaining in the network after attack are screened out, and the ratio of flow and capacity before attack V0OC and the ratio of flow and capacity after attack V1 OC .

[0031] As a further scheme of the present application: the index data of the road segments in the network after removal is screened as the index data of the O-D pairs according to the index relationship between the O-D pairs in the commuting behavior data and the original road segments, index removal data is obtained, and the index removal data is excluded in the incremental traffic assignment, the incremental traffic assignment algorithm is executed on the commuting behavior data after exclusion in the third road network G3(V, E), and the road flow distribution result FM1 after network attack is output, and the bottleneck is identified based on the road flow distribution result FM1, further comprising:

[0032] S102: The VOC change of the remaining road segments in the network before and after network attack is calculated.

[0033] As a further scheme of the present application: the VOC change is ΔVOC = V1OC-V0 OC ​wherein, AVOC greater than 0 indicates that the road traffic is higher than before the attack, V1 OC greater than 1 indicates that the road traffic is higher than its own capacity, and the road satisfying both conditions is identified as the network bottleneck, and V1 OC The higher the value, the higher the road traffic.

[0034] Compared with the prior art, the present application has the beneficial effects that:

[0035] 1. In the present application, O-D pair information is provided through traffic travel data, an incremental traffic assignment algorithm is used to assign the spatial position information travel matrix, the congestion effect generated by the cascade of network traffic is simulated through multiple traffic assignment and iterative updating of network edge weight, and the network bottleneck under network traffic reconstruction is identified. Through the method framework of fusing dynamic traffic evolution modeling and network structure identification, based on the overall traffic network, the cascade conduction mechanism of traffic network traffic under disturbance conditions can be systematically described, the network bottleneck having a key influence on system stability can be identified, and the response ability and structural resilience of urban risk can be improved.

[0036] 2. In the present application, the cascade effect of network traffic is simulated based on the incremental traffic assignment algorithm, which can simplify the calculation steps, does not need to rely on the complex mechanism of microscopic individual simulation or modeling, can realize the traffic assignment of large-scale and complex structure urban traffic network, and reduces the calculation cost of traffic assignment. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is a method flow chart of the present application;

[0038] Figure 2 in which Figure 2 (a) is a road traffic distribution result t graph of the incremental traffic assignment of the embodiment of the present application, Figure 2 (b) is a road traffic distribution result graph after outputting network attack of the embodiment of the present application;

[0039] Figure 3 is a road network bottleneck graph in the embodiment of the present application under the road network of the 500-year return period rainfall flood simulation scenario. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0041] Embodiment:

[0042] Referring to Figure 1 In the embodiments of the present application, a traffic network bottleneck identification method based on flow cascade comprises the following method:

[0043] S1: Obtain road data and commuting behavior data in a target area, wherein the road data comprises free-flow travel time t0 of each road, capacity C of the road, and spatial position information of the road, and the commuting behavior data comprises spatial position information of commuting O-D and commuting flow corresponding to the O-D;

[0044] S2: Extract the midpoint of each road in the road data, connect the O-D of the commuting behavior data to the nearest midpoint of the road, and record the corresponding index relationship between the O-D and the midpoint of the road;

[0045] S3: Establish a first road network G1(V,E) by using an initial modeling method, wherein V is the first network node, V is a set composed of each intersection of the road, E is the first edge, E is a set composed of each road, each edge in E corresponds to the free-flow travel time t0 and the capacity C of the road, and the free-flow travel time t0 is taken as the weight of E;

[0046] S4: Establish a second road network G2(V,E) by using a dual modeling method, wherein V is the second network node, V is a set composed of roads, E is the second edge, and E is a set of edges composed of each intersection of the road;

[0047] S5: Divide the commuting flow data corresponding to each O-D pair in the commuting behavior data into a first part of flow, a second part of flow, a third part of flow, and a fourth part of flow by using an incremental traffic assignment algorithm, and readjust the network edge weight t of each network edge of the second road network G2(V,E);

[0048] S6: Output the flow V of each O-D pair, take the flow V of the O-D pair as the road flow distribution result FM0 of the incremental traffic assignment, remove a certain proportion of road segment sets and a certain proportion of network node sets corresponding to the road segment sets from G2(V,E), represent the removed network as a third road network G3(V,E), and initialize the node attribute and edge weight of the third road network G3(V,E);

[0049] S7: According to the index relationship of the O-D pairs in the commuting behavior data and the spatial connection of the original road segments, the road segments in the network after removal are screened as index data of the O-D pairs, index removal data is obtained, and the index removal data is excluded in the incremental traffic assignment. The commuting behavior data after exclusion is executed in the third road network G3(V, E) to perform an incremental traffic assignment algorithm, and a road flow distribution result FM1 after network attack is output. Based on the road flow distribution result FM1, a bottleneck is identified.

[0050] In this embodiment, the collected area is the road network data within the land range of a certain region, the road is simplified as a single line section, referred to as a road segment, and the format is shapefile. The free travel time t0, the capacity C and the spatial position information of the road of each road segment are required. The coordinate system of the data is converted to EPSG = 32651.

[0051] Preferably, step S2 is used to collect commuting behavior data within the land range of a certain region. The format of the commuting behavior data is csv, and the commuting behavior data includes spatial position information of commuting O-D and commuting flow of the O-D.

[0052] Preferably, in step S3, the Geopandas library in Python language is used to read the road network shapefile data obtained in step S1, the centroid function is used to identify the midpoint of each road segment in the road network data, and the sjoin_nearest function in the Geopandas library is used for spatial connection, so that the road midpoint is matched to the spatial position of the commuting O-D pair obtained in step S2, and the index of the road midpoint corresponding to the start point and the end point of each O-D pair is obtained, which is recorded as ODLinkRoad_dict.

[0053] Preferably, step S4 includes:

[0054] S41: Extracting each road i and the road set i_link connected by each road from the first road network G1(V, E), using an iterative method to select each connected road j in the road set i_link, inputting the road i and the connected road j into the set second road network G2(V, E) in turn as the second network node, and inputting the free flow travel time t0 and the capacity C of the road i and the connected road j in the first road network G1(V, E) into the set second road network G2(V, E);

[0055] S42: Establish network edge data between road i and connecting road j, and input the network edge data into the set second road network G2(V, E), take the average value of free-flow travel time t0 of road i and connecting road j as the edge weight t, repeat step S41 until all roads are input as i into the set second road network G2(V, E), remove the duplicate nodes and edges of the set second road network G2(V, E) to obtain the dual network G2'(V', E');

[0056] S43: Add attribute V to each network node of the dual network G2'(V', E'), and set the initial value of the carrying capacity of the road section of the dual network G2'(V', E') to 0 to obtain the second road network G2(V, E), wherein the network nodes of the second road network G2(V, E) are road sections, and the edges of the second road network G2(V, E) are road intersections.

[0057] Preferably, in step S4, the Networkx library is called to establish the road network G1(V, E) using the initial modeling method, wherein V is a set of network nodes, each network node corresponds to a road intersection; E is a set of network edges, each network edge corresponds to a road section. The specific method is as follows: the road data is traversed row by row, the road section is added as the network edge, the free-flow travel time t0 of the road section is set as the edge weight of the network edge, the capacity C of the road section is set as the edge attribute of the network edge, and the starting point of the road section is set as the network node.

[0058] Preferably, step S5 includes:

[0059] S51: Index the midpoint contained in the road section corresponding to the O-D pair of the first part of the flow, execute the shortest path algorithm between two points in the second road network G2(V, E), set the edge weight as t, and return the data as the node list passed through by the two points, wherein the node list passed through is a road section;

[0060] S52: For each node list passed through by the O-D pair, multiply the flow of the O-D pair by the distribution proportion of the corresponding commuting flow data to obtain the distributed flow, sum the distributed flow and the original flow to obtain the new flow, and distribute the new flow to each node in the node list;

[0061] S53: Use the time and flow resistance function to calculate the current actual travel time t a of the road section of each network node of the second road network G2(V, E) a as the new weight t a , wherein the time and flow resistance function is:

[0062] wherein t0,C and V are the current node attributes, and a and b are preset parameters, a is 0.15 and b is 4;

[0063] S54: based on the new weight t of each network edge at both ends of the network edge a , the network edge weight t is adjusted for each network edge, wherein the network edge weight t is the average value of the weights t of both ends of the network edge. a

[0064] Preferably, in step S5, the igraph library is called to establish the road network G2(V,E) using the dual modeling method, wherein V is a set of network nodes, each network node corresponding to a road intersection; E is a set of network edges, each network edge corresponding to a road section, and the specific implementation steps are as follows:

[0065] S501: extract each network edge in G1(V,E) and mark it as edge i; obtain other network edges connected to the starting point of edge i, and mark the set composed of these network edges as i_link;

[0066] S502: iteratively select a network edge in i_link and mark it as edge j; input edge i and edge j into G2(V,E) as network nodes, and set their free-flow travel time t0 and capacity C in G1(V,E) as node attributes;

[0067] S503: for each edge i and edge j, establish a connection between the two and input it into G2(V,E), and set the average value of the free-flow travel time t0 of the two as the edge weight t of the network G2(V,E);

[0068] S504: repeat the above steps S501-S503 until all network edges of G1(V,E) have been processed as edge i;

[0069] S505: use the simplify function of the igraph library to delete duplicate network nodes and network edges in G2(V,E) to obtain the final G2(V,E);

[0070] S506: add a traffic attribute V to each network node in G2(V,E), and set the initial value as 0, indicating the current road section carrying capacity.

[0071] Preferably, in step S5, the network G2(V,E) is obtained, in which the network nodes are road sections, C and V, and the edges are road intersections, the attributes of the road sections are t0, and the weight t of the road intersections is the average value of the weights t0 of both ends of the network edge.

[0072] ​An incremental traffic assignment algorithm is executed. The algorithm first divides the traffic of each pair of O-D pairs in the commuting behavior data into four parts of 40%, 30%, 20%, and 10%. For the selected part of the commuting spatial location information data, the ODLinkRoad_dict index is used to index the corresponding road segment midpoint, and the corresponding network node pair of the O-D pair is found in G2(V, E), and the shortest path algorithm between the two nodes is executed using get_shortest_paths, with the edge weight t, and the returned data being the node list (i.e., road segment) passed through the two points, recorded as vpaths;

[0073] For the node list vpaths passed through each pair of O-D pairs, the traffic of the pair of O-D pairs is multiplied by the current assignment proportion (i.e., 40%, 30%, 20%, or 10%) and added to each node in the list, and summed with the original traffic V attribute to obtain the new traffic V, which replaces the original attribute.

[0074] A "time-flow" road impedance function is executed for each network node of G2(V, E), and the current actual travel time t of the road segment as the node is updated a . The function can be represented as where t0, C, and V are the current node attributes, and a and β are preset parameters, taking 0.15 and 4;

[0075] For each network edge, the source and target are used to find the two end nodes, and the new weight t of the two end nodes is recalculated a The network edge weight t is adjusted, and is calculated as the average of the two node weights t a .

[0076] As shown in Figure 2 (a), after the above steps are completed, the traffic V of each node of the final network is output, representing the road traffic distribution result FM0 of the incremental traffic assignment.

[0077] Preferably, step S10 includes:

[0078] S101: Screening out the remaining road segment nodes of the network after the attack, calculating the traffic and capacity ratio of the remaining road segment nodes before and after the attack, wherein the traffic and capacity ratio before the attack is V0OC, and the traffic and capacity ratio after the attack is V1 OC .

[0079] Preferably, the index data of the road segments in the network after removal is screened as O-D pairs according to the index relationship of the O-D pairs in the commuting behavior data and the original road segment space connection, index removal data is obtained, and the index removal data is excluded in the incremental traffic assignment. The commuting behavior data after exclusion is executed in the third road network G3(V, E) to output the road flow distribution result FM1 after network attack. Bottlenecks are identified based on the road flow distribution result FM1, and the method further comprises:

[0080] S102: Calculate the VOC change of the remaining road segments in the network before and after the network attack.

[0081] Preferably, the VOC change is ΔVOC = V1 OC -V0 OC , wherein ΔVOC > 0 indicates that the road flow after attack is higher than that before attack, V1 OC > 1 indicates that the road flow after attack is higher than its own capacity, and the road satisfying both conditions is identified as a network bottleneck, and V1 OC The higher V1 2 is, the higher the road flow is.

[0082] In this embodiment, the scenario of network attack is a 500-year return period rainfall flood scenario, and the specific implementation method is as follows:

[0083] Obtain the Digital Elevation Model (Digital Elevation Model, hereinafter referred to as DEM) data in the research range, with a precision of 30 meters, and take each grid cell as a catchment unit. Convert the coordinate system of the data to EPSG = 32651, and the data type is TIF;

[0084] Obtain land use type data in the research range, convert the coordinate system of the data to EPSG = 32651, and spatially connect with the DEM data to find the land use type of each catchment unit;

[0085] According to the calculation formula of the “Rainstorm Intensity Formula and Design Rain Type Standard” of the research area, set the 500-year return period and 60-minute rainfall duration to calculate the rainfall q;

[0086] The SCS-CN method is used to calculate the rainfall runoff. First, the corresponding Curve number (CN) value is found according to the land use type, and the saturated water storage S is obtained according to the formula 25400 / (CN-254). Further, the runoff Q is calculated according to the formula (q-0.05S) 2 / (q+0.95S);

[0087] According to the drainage data of the research area, the rainwater drainage pump capacity of the area is about 4996.85 cubic meters per second. Assuming that the drainage pipe network is evenly distributed, the total drainage capacity is evenly distributed to each catchment unit. The final waterlogging volume is calculated as the difference between the runoff Q and the drainage capacity;

[0088] The total waterlogging volume is distributed by the equal volume method. The implementation is to select the lowest terrain in all catchment units as the starting point to distribute the flood volume step by step. The flood volume is iteratively selected until all flood volumes are distributed. The submerged volume of each catchment area is obtained. The submerged depth is calculated according to the area of the catchment area;

[0089] The catchment unit with a submerged depth of not less than 30.5 cm is set as the submerged area, and the data is converted into spatial vector data in the shapefile format;

[0090] The Geopandas library in Python is used to read the submerged area data, and the sjoin_nearest function is used to perform spatial connection with the road network. The overlap length of each road section and the submerged area is calculated. If the length ratio is higher than 7% of the total length (a parameter obtained by robustness test), the road section is identified as a submerged road section;

[0091] All submerged road sections are screened, and the remove function is used to remove the corresponding network node set in G2(V, E). The network after removal is represented as G3(V, E), and the node attributes (t0, C and V) and edge weight (t) are initialized to the original values;

[0092] According to the index relationship ODLinkRoad_dict between the O-D pairs in the commuting behavior data and the spatial connection of the original road section, the removed road section is selected as the data of the O-D pair index, and is excluded in the subsequent incremental traffic assignment;

[0093] The commuting behavior data after exclusion is executed again in G3(V, E) to perform the incremental traffic assignment algorithm. The specific steps are the same as S7. The output is the road traffic distribution result after network attack As Figure 2 (b) shows;

[0094] According to the network traffic reconstruction result caused by the attack scenario, the bottleneck is identified. The remaining nodes (i.e. road sections) of the network after attack are selected, and the ratio of the traffic and capacity of the nodes before and after attack is calculated, i.e. and The ratio of the traffic and capacity of the nodes before and after attack is calculated, i.e. is represented as V0 OC and V1 OC ;

[0095] AsFigure 3 The VOC change of the part of the road section before and after the network attack is calculated, as ΔVOC = V1OC - V0 OC . Wherein, ΔVOC greater than 0 indicates that the road traffic is higher after the attack than before the attack, V1OC greater than 1 indicates that the road traffic is higher than its own capacity after the attack, the road satisfying both conditions is identified as a network bottleneck, and V1 OC The higher, the higher the road traffic.

[0096] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art within the technical range disclosed by the present application, according to the technical scheme and the inventive concept of the present application, equivalent replacement or change, should be covered within the protection scope of the present application.

Claims

1. A method for identifying traffic network bottlenecks based on traffic flow cascading, characterized in that, include: S1: Obtain road data and commuting behavior data within the target area. The road data includes the free passage time t0 of each road, the road capacity C, and the spatial location information of the road. The commuting behavior data includes the spatial location information of commuting OD and the commuting flow corresponding to OD. S2: Extract the midpoint of each road in the road data, connect the OD pairs of the commuting behavior data to the midpoint of the nearest road, and record the corresponding index relationship between the OD pairs and the midpoints of the roads; S3: The first road network G1(V,E) is established using the initial modeling method, where V is the first network node, V is the set of all intersections of the road, E is the first edge, and E is the set of each road. Each edge in E is assigned to the free flow time t0 and capacity C of the road, and the free flow time t0 is used as the weight of E. S4: The second road network G2(V,E) is established using the dual modeling method, where V is the second network node, V is the set of roads, E is the second edge, and E is the set of edges formed by each intersection of the roads. S5: The incremental traffic assignment algorithm is used to divide the commuting traffic data corresponding to each pair of OD pairs in the commuting behavior data into the first part of traffic, the second part of traffic, the third part of traffic and the fourth part of traffic, and to readjust the weight t of each network edge of the second road network G2(V,E). S6: Output the flow V of each OD pair, use the flow V of the OD pair as the road flow distribution result FM0 of incremental traffic allocation, remove a certain proportion of the road segment set and the network node set corresponding to the road segment set from G2(V,E), represent the removed network as the third road network G3(V,E), and initialize the node attributes and edge weights of the third road network G3(V,E); S7: Based on the index relationship between the OD pairs in the commuting behavior data and the original road segments, filter the road segments in the removed network as index data for the OD pairs, obtain the index removal data, remove the index removal data in the incremental traffic assignment, execute the incremental traffic assignment algorithm in the third road network G3(V,E) with the removed commuting behavior data, output the road traffic distribution result FM1 after the network attack, and identify bottlenecks based on the road traffic distribution result FM1.

2. The traffic network bottleneck identification method based on flow cascading as described in claim 1, characterized in that: The dual modeling method is used to establish the second road network G2(V,E), where V is a second network node, V is a set of roads, E is a second edge, and E is the set of edges formed by each intersection of the roads, including: S41: Extract each road i and the set of roads i_link connected to each road from the first road network G1(V,E). Use an iterative method to select each connecting road j in the set of roads i_link. Input road i and connecting road j as second network nodes into the set second road network G2(V,E) in sequence. Input the free flow time t0 and capacity C of road i and connecting road j in the first road network G1(V,E) into the set second road network G2(V,E).

3. The traffic network bottleneck identification method based on flow cascading according to claim 2, characterized in that: The method of establishing a second road network G2(V,E) using dual modeling, where V is a second network node, V is a set of roads, E is a second edge, and E is a set of edges formed by each intersection of the roads, also includes: S42: Establish network edge data between road i and connecting road j, and input the network edge data into the set second road network G2(V,E). Use the average free travel time t0 of road i and connecting road j as the edge weight t. Repeat step S41 until all roads are input as i into the set second road network G2(V,E). Remove duplicate nodes and edges from the set second road network G2(V,E) to obtain the dual network G2'(V',E').

4. The traffic network bottleneck identification method based on flow cascading according to claim 3, characterized in that: The method of establishing a second road network G2(V,E) using dual modeling, where V is a second network node, V is a set of roads, E is a second edge, and E is a set of edges formed by each intersection of the roads, also includes: S43: Add attribute V to each network node of the dual network G2'(V',E'), and set the initial value of the carrying capacity of the road segment of the dual network G2'(V',E') to 0 to obtain the second road network G2(V,E), where the network nodes of the second road network G2(V,E) are road segments, and the edges of the second road network G2(V,E) are road intersections.

5. The traffic network bottleneck identification method based on flow cascading according to claim 4, characterized in that: The incremental traffic assignment algorithm divides the commuting traffic data corresponding to each OD pair in the commuting behavior data into four parts: a first part, a second part, a third part, and a fourth part. The weight t of each network edge in the second road network G2(V,E) is readjusted. The first part of the traffic data comprises 40% of the commuting traffic data, the second part comprises 30%, the third part comprises 20%, and the fourth part comprises 10%. S51: The index contains the midpoint of the road segment corresponding to the OD pair of the first part of the traffic. The shortest path algorithm between the two points is executed in the second road network G2(V,E). The edge weight is set to t. The returned data is a list of nodes passed by the two points, where the list of nodes passed by the two points is the road segment.

6. The traffic network bottleneck identification method based on flow cascading according to claim 5, characterized in that: The incremental traffic assignment algorithm divides the commuting traffic data corresponding to each OD pair in the commuting behavior data into four parts: a first part, a second part, a third part, and a fourth part. The weight t of each network edge in the second road network G2(V,E) is readjusted. The allocation ratio of the commuting traffic data for the first part is 40%, for the second part it is 30%, for the third part it is 20%, and for the fourth part it is 10%. The algorithm also includes: S52: For each OD pair, multiply the traffic of the OD pair by the corresponding commuter traffic data allocation ratio to obtain the allocated traffic. Sum the allocated traffic with the original traffic to obtain the new traffic. Distribute the new traffic to each node in the node list.

7. The traffic network bottleneck identification method based on flow cascading according to claim 1, characterized in that: The incremental traffic assignment algorithm divides the commuting traffic data corresponding to each OD pair in the commuting behavior data into four parts: a first part, a second part, a third part, and a fourth part. The weight t of each network edge in the second road network G2(V,E) is readjusted. The allocation ratio of the commuting traffic data for the first part is 40%, for the second part it is 30%, for the third part it is 20%, and for the fourth part it is 10%. The algorithm also includes: S53: Using time and flow resistance functions, calculate the current actual travel time t of each road segment at each network node of the second road network G2(V,E). a The actual travel time t a As the new weight t a The time-flow-path-resistance function is as follows: Where t0, C, and V are the attributes of the current node, and α and β are preset parameters, with α set to 0.15 and β set to 4.

8. The traffic network bottleneck identification method based on flow cascading according to claim 7, characterized in that: The incremental traffic assignment algorithm divides the commuting traffic data corresponding to each OD pair in the commuting behavior data into four parts: a first part, a second part, a third part, and a fourth part. The weight t of each network edge in the second road network G2(V,E) is readjusted. The allocation ratio of the commuting traffic data for the first part is 40%, for the second part it is 30%, for the third part it is 20%, and for the fourth part it is 10%. The algorithm also includes: S54: Based on the new weights t of the two endpoints of each network edge a For each network edge, the network edge weight t is readjusted, where the network edge weight t is the weight t of the two endpoints. a The average value.

9. The traffic network bottleneck identification method based on flow cascading according to claim 8, characterized in that: The process involves filtering road segments in the removed network as index data for OD pairs based on the index relationship between the commuting behavior data and the original road segments, obtaining index-removed data, and then removing this index-removed data in incremental traffic assignment. The removed commuting behavior data is then used to execute the incremental traffic assignment algorithm in the third road network G3(V,E), outputting the road traffic distribution result FM1 after the network attack. Bottlenecks are identified based on the road traffic distribution result FM1, including: S101: Filter out the remaining road segment nodes in the network after the attack, and calculate the ratio of traffic and capacity of the remaining road segment nodes before and after the attack. The ratio of traffic and capacity before the attack is V0OC, and the ratio of traffic and capacity after the attack is V1OC.

10. The traffic network bottleneck identification method based on flow cascading according to claim 9, characterized in that: The process involves filtering road segments in the removed network as index data for OD pairs based on the index relationship between the commuting behavior data and the original road segments, obtaining index-removed data, removing the index-removed data in incremental traffic assignment, executing the incremental traffic assignment algorithm on the removed commuting behavior data in the third road network G3(V,E), and outputting the road traffic distribution result FM1 after the network attack. Bottlenecks are identified based on the road traffic distribution result FM1. The process also includes: S102: Calculate the VOC changes of the remaining road segments in the network before and after the network attack.

Images