A dynamic prediction and control method for traffic accident congestion queue in extra-long tunnels

Abstract

The present invention discloses a method for dynamically predicting and controlling traffic congestion queues in extra-long tunnels, and relates to the technical field of traffic congestion prediction. The present invention divides the tunnel into n sections, and performs filtering processing on the real-time flow and density data of each section where the accident occurs; analyzes the length of the vehicle queue in the tunnel at different stages of the accident and determines the type of accident to determine the relationship between the background traffic volume and the remaining traffic capacity of the tunnel; predicts the maximum length of the vehicle queue; introduces a sliding average filtering method to construct a real-time estimation model for the length of the vehicle queue and calculates it; determines the vehicle queue situation in the tunnel; and takes different control measures for different stages of queue length development. The present invention can reflect the changes in the length of the vehicle queue in the tunnel in real time by establishing a real-time prediction model for the length of the vehicle queue, providing a theoretical basis for the management department to take the next step of control measures, and providing real-time feedback to the management department.

Description

Technical Field

[0001] The invention relates to the technical field of traffic congestion prediction, and in particular to a method for dynamically predicting and controlling traffic accident congestion queues in extra-long tunnels. Background Art

[0002] At present, the research methods on vehicle queuing at home and abroad are mainly concentrated in the following aspects: research methods based on shock wave theory, probability methods, research methods based on input-output models and machine learning methods, etc. Among them, the traffic wave model with traffic flow theory as the core has always occupied a relatively important position. At the same time, with the development of estimation computer technology and intelligent algorithms, artificial intelligence algorithms have gradually become popular.

[0003] Most of the above studies are based on highways and urban trunk roads, and there are fewer studies on the accident queue length in extra-long tunnels. There are even fewer studies on the real-time dynamic prediction of the vehicle queue length during the entire process of congestion generation, spread, and dissipation from the occurrence to the resolution of the accident. Therefore, the present invention proposes a dynamic prediction and control method for traffic accident congestion queues in extra-long tunnels. Summary of the invention

[0004] The purpose of the present invention is to provide a dynamic prediction and control method for congestion queues in extra-long tunnels caused by traffic accidents, to explore the evolution characteristics and laws of queue length in the whole process of congestion generation, spread and dissipation under extra-long tunnel traffic accidents, to establish a vehicle maximum queue length model and a vehicle queue length real-time estimation model for the whole process of congestion generation, spread and dissipation under extra-long tunnel traffic accidents, to grasp the vehicle queue length in real time, to provide key technical methods and data support for tunnel traffic control, and to help improve tunnel traffic operation efficiency and reduce operation safety risks.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a method for dynamically predicting and controlling traffic congestion queues in extra-long tunnels. First, the real-time flow and density data of each section of the tunnel after the accident are collected. Then, based on the traffic flow shock wave theory, the following steps are combined to realize the evolution analysis and modeling of the queue length of vehicles in the extra-long tunnel traffic accident. The steps include:

[0007] S1: Divide the tunnel into n sections, filter the real-time traffic and density data of each section where the accident occurred, obtain smoothed data, and then qualitatively analyze the changing characteristics of the process from the generation to the spread and then to the dissipation of the vehicle queue under the traffic accident, laying the theoretical foundation for S3;

[0008] S2: Analyze the length of the vehicle queue in the tunnel at different stages of the accident and determine the type of accident to determine the relationship between the background traffic volume and the remaining capacity of the tunnel. If the background traffic volume is greater than the remaining capacity, proceed to S3;

[0009] S3: Based on the shock wave theory, a tunnel traffic accident vehicle queue length model is constructed to predict the maximum queue length of vehicles;

[0010] S4: Combine the traffic flow data collected in the tunnel and introduce the sliding average filter method to build a real-time estimation model for the length of the vehicle queue and calculate it;

[0011] S5: using S4 to estimate in real time the relationship between the real-time queue lengths of the vehicles and the time relationship to determine the queue situation of the vehicles in the tunnel;

[0012] S6: Take different control measures for different queue length development stages.

[0013] S1-S2 includes the generation of vehicle queues in the tunnel, the spread of vehicle queues in the tunnel and the dissipation of vehicle queues in the tunnel. The generation of vehicle queues in the tunnel is based on the basic traffic flow diagram, and the traffic flow state is divided by the relationship between flow and density. After the occurrence of the accident, the tunnel management department will implement flow control at the control point to avoid a large number of vehicles entering and causing aggravated congestion, and avoid secondary accidents.

[0014] S3 includes geometric analysis and model building. The geometric analysis combines the traffic flow shock wave theory, the evolution process of the vehicle queue length under the traffic accident analyzed in the previous section, and the change characteristics of the flow, density, and speed during the whole process to make a time-space trajectory line diagram of the vehicle during the whole process of the accident. The model building is to construct a maximum queue length model with flow, density, speed, and the distance between the accident location and the hole as variables after the geometric analysis of the whole process of queue change.

[0015] The steps S4-S5 include a process for implementing real-time estimation of the length of a vehicle queue, a real-time estimation model for the length of a vehicle queue, a method for determining a time interval, and an improved model introducing a sliding average filter. In the process for implementing real-time estimation of the length of a vehicle queue, when an accident occurs in a tunnel and the upstream traffic volume is greater than the remaining capacity of the accident section, a traffic bottleneck is formed at the accident point due to the decrease in capacity. A large number of vehicles cannot pass through in a timely and smooth manner, which will cause congestion. Otherwise, congestion will not occur. In the real-time estimation model for the length of a vehicle queue, through the analysis of the evolution process of the length of a vehicle queue, it is found that the main factors affecting the change in the length of the queue are the magnitude and direction of the shock wave velocity between the congested area and the non-congested area. After the accident occurs, when the shock wave propagates in the direction of the traffic flow, the length of the queue will decrease, and when it propagates in the opposite direction of the traffic flow, the length of the queue will increase. In the method for determining the time interval, due to the randomness of vehicle arrival, the flow and density data extracted in different time intervals are significantly different, and the real-time queue lengths estimated at different time intervals T are different. The improved model introducing a sliding average filter specifically includes the principle of sliding average filtering, the establishment of an improved model, and accuracy comparison.

[0016] The S6 includes dynamic control strategy analysis under extra-long tunnel traffic accidents, dynamic control model constraints under extra-long tunnel traffic accidents and dynamic control model construction under extra-long tunnel traffic accidents. The dynamic control strategy analysis under extra-long tunnel traffic accidents specifically includes queue growth period, queue dissipation period and queue end. The dynamic control model constraints under extra-long tunnel traffic accidents specifically include traffic efficiency function and operation risk function. The dynamic control model construction under extra-long tunnel traffic accidents specifically includes target constraints and model construction.

[0017] The present invention has the following beneficial effects:

[0018] The dynamic prediction and control method for congestion queues in extra-long tunnels caused by traffic accidents in the present invention can reveal the process mechanism of the generation, spread and dissipation of vehicle queues in extra-long tunnels caused by traffic accidents. The vehicle flows on different sections in the tunnel are qualitatively divided into states, and the shock wave theory of traffic flow is used to explain the influence of shock waves generated between vehicle flows under different traffic conditions on the length of vehicle queues.

[0019] The method for dynamically predicting and controlling traffic congestion queues in extra-long tunnels of the present invention establishes a real-time prediction model for the length of vehicle queues, and the accuracy of predicting the maximum queue length can reach 95.62%, and the accuracy of the queue length during the entire process can reach 84.34%. The model can reflect the changes in the length of vehicle queues in the tunnel in real time, provide a theoretical basis for the management department to take the next step of control measures, and provide real-time feedback to the management department.

[0020] The method for dynamic prediction and control of traffic congestion queues in extra-long tunnels of the present invention can realize the regulation of the running state of traffic flow after traffic accidents. Under the premise of ensuring the running safety in the tunnel, the dynamic control model can be controlled 120s in advance, and the improved traffic efficiency can reach 7526veh·km·h -1 Dynamic traffic control in tunnels can reduce vehicle congestion and queues while ensuring that the tunnel's capacity is fully utilized. This method provides scientific and accurate decision-making support for tunnel management and control, allowing the tunnel and its affiliated sections to quickly resume efficient and safe operation, allowing the tunnel to achieve maximum economic and social benefits.

[0021] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0023] Figure 1 This is a technical flow chart of the method for dynamic prediction and control of traffic accident congestion queues in extra-long tunnels according to the present invention;

[0024] Figure 2 It is a flow-density diagram of the vehicle flow in the tunnel of the present invention under normal operating conditions;

[0025] Figure 3 This is a schematic diagram of the normal operation of traffic flow in the tunnel of the present invention;

[0026] Figure 4 It is a flow-density schematic diagram when a traffic accident occurs in a tunnel of the present invention;

[0027] Figure 5 This is a schematic diagram of the traffic flow in the tunnel when an accident occurs in the present invention;

[0028] Figure 6 It is a flow-density schematic diagram of the traffic flow in the tunnel after the traffic accident occurs in the present invention;

[0029] Figure 7 This is a schematic diagram of the traffic flow in the tunnel after the control of the present invention;

[0030] Figure 8 It is a flow-density schematic diagram of the traffic flow in the tunnel after the traffic accident occurs in the present invention;

[0031] Fig. 9 A schematic diagram of traffic flow when a vehicle controlled by the present invention meets a queued vehicle;

[0032] Fig.10 This is a flow-density schematic diagram of the traffic flow in the tunnel after the accident is cleared in the present invention;

[0033] Fig.11 This is a schematic diagram of the traffic flow after the accident is cleared according to the present invention;

[0034] Fig.12 This is a flow-density schematic diagram of the traffic flow in the tunnel after the accident is cleared in the present invention;

[0035] Fig.13 It is a schematic diagram of the traffic flow when the queue is dissipated according to the present invention;

[0036] Fig.14 It is a time-space running trajectory line diagram of the vehicle of the present invention;

[0037] Fig.15 It is the flow-density relationship diagram of the present invention;

[0038] Fig.16 This is a flowchart for estimating the length of a vehicle queue in an accident in an extra-long tunnel according to the present invention;

[0039] Fig.17 A schematic diagram of estimating the length of a vehicle queue using the shock wave model of the present invention;

[0040] Fig.18 A schematic diagram showing comparison of queue lengths at different time intervals T according to the present invention;

[0041] Fig.19 This is a schematic diagram comparing the queue lengths when the time interval T is 30s;

[0042] Fig. 20 It is a schematic diagram of the queue length growth of the present invention;

[0043] Fig.21 It is a schematic diagram of queue length dissipation of the present invention;

[0044] Fig. 22 It is the dynamic control flow chart of the present invention;

[0045] Fig.23 This is a schematic diagram of the traffic efficiency of the present invention;

[0046] Fig.24 It is a schematic diagram of the speed encounter process of the present invention;

[0047] Fig.25 This is an operation flow chart of the method for dynamic prediction and control of traffic accident congestion queues in extra-long tunnels according to the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1 and Fig.25 As shown: The present invention is a dynamic prediction and control method for congestion queues in extra-long tunnels due to traffic accidents. First, the real-time flow and density data of each section of the tunnel after the accident are collected. Then, based on the traffic flow shock wave theory, the following steps are combined to realize the evolution analysis and modeling of the queue length of vehicles in the extra-long tunnel due to traffic accidents. The steps include:

[0050] S1: Divide the tunnel into n sections, filter the real-time traffic and density data of each section where the accident occurred, obtain smoothed data, and then qualitatively analyze the changing characteristics of the process from the generation to the spread and then to the dissipation of the vehicle queue under the traffic accident;

[0051] S2: Analyze the length of the vehicle queue in the tunnel at different stages of the accident and determine the type of accident to determine the relationship between the background traffic volume and the remaining capacity of the tunnel. If the background traffic volume is greater than the remaining capacity, proceed to S3;

[0052] The length of the queue for traffic accidents in extra-long tunnels can be divided into three stages: queue generation, queue spread, and queue dissipation.

[0053] 1. Vehicle queues in tunnels

[0054] First, based on the basic traffic flow diagram, the traffic flow state is divided using the relationship between flow and density, such as Figure 2 As shown, q m1 is the normal traffic capacity in the tunnel, k m1 is the optimal density when the tunnel capacity is maximum, k j1 is the blocking density; defined in k m1 The traffic on the left is smooth, while the traffic on the right is congested. Figure 3 As shown, at this time the traffic in the tunnel is running smoothly.

[0055] When an accident occurs, the accident vehicle occupies the road. The normal flow-density relationship in the tunnel and the flow-density relationship at the accident section are as follows: Figure 4 As shown in the figure, the traffic capacity of the accident section suddenly decreases. At this time, the traffic capacity of the accident section in the tunnel changes from q m1 Reduced to q m2 , k m2 is the optimal density when the accident section in the tunnel has the maximum traffic capacity, kj1 is the blocking density of the accident section; at this time, the traffic flow in the tunnel is as follows Figure 5 As shown, state 1 is a smooth state, state 2' is a congested state, state 2 is a congested state, and state 2" is a smooth state.

[0056] When an accident occurs, the traffic flow at the accident section changes from the original state 1 to state 2 because the traffic flow at this time is greater than the remaining traffic capacity of the accident section. The accident section forms a traffic bottleneck, and vehicles that cannot pass in time begin to queue up. The traffic flow of the adjacent section upstream of the accident section changes from the original state 1 to state 2'. When traffic flows in different traffic states meet, a shock wave w will be generated. 2'1 ,like Figure 6 As shown by arrow line 1-2', a shock wave w will be generated between the accident section state 2 and the upstream state 2'. 22' , a shock wave w will be generated between the accident section state 2 and the downstream state 2” 22” ,like Figure 6 As shown by arrow line 2-2".

[0057] At this time, because Q2=Q 2' ,K2<K 2' , the shock wave formula can be used to obtain the shock wave w 22' Wave speed for:

[0058]

[0059] Shockwave 22' The wave speed is 0, and it will not propagate upstream or downstream. Similarly, the shock wave w 22” The wave speed is also zero. Because Q1>Q 2' ,K1<K 2' , shock wave 2'1 Wave speed for:

[0060]

[0061] Shockwave 2'1 A wave speed less than 0 is a negative value, indicating that the shock wave propagates in the opposite direction of the traffic flow, that is, propagates upstream, congestion occurs, and vehicle queues begin to form.

[0062] 2. Long queues of vehicles in tunnels

[0063] When an accident occurs, the tunnel management department will implement flow control at the flow control point (assuming that the flow control point is located at the tunnel entrance) to prevent a large number of vehicles from entering the tunnel, causing increased congestion, or even a secondary accident. After the flow control point is implemented, the traffic volume entering the tunnel begins to decrease. At this time, the traffic flow in the tunnel is as follows: Figure 7As shown, the flow-density of each state is as follows Figure 6 shown.

[0064] After the implementation of flow control management, the traffic flow status from the accident section to the control point can be divided into five categories (the traffic flow status outside the control point is not considered), which are smooth state 2", smooth state 2 with flow equal to the remaining traffic capacity of the accident section, congested queue state 2', smooth state 1 maintaining normal operation before the accident, and smooth state 3 with reduced flow after flow control. When the traffic in state 3 meets the traffic in state 1, a shock wave w is generated. 13 , because Q3<Q1, K3<K1, the shock wave w 13 Wave speed for:

[0065]

[0066] Shockwave 13 The wave velocity is greater than 0, indicating that the shock wave propagates in the same direction as the traffic flow, that is, propagates downstream, and will not cause traffic congestion. 2'1 Continuously propagating upstream, the traffic flow in state 1 gradually changes to state 2', and the length of the vehicle queue increases.

[0067] 3. The queue of vehicles in the tunnel has been dissipated

[0068] When the wave w 2'1 After all vehicles in state 1 have passed, the traffic flow in the tunnel is as follows: Fig. 9 As shown, the flow-density of each state is as follows Figure 8 shown.

[0069] When the wave w 2'1 After all vehicles in state 1 have passed, the traffic in state 1 has all changed to state 2'. The traffic flow from the accident section to the flow control point can be divided into four categories, namely, smooth state 2", smooth state 2 with flow equal to the remaining traffic capacity of the accident section, congested queue state 2', and smooth state 3 with reduced flow after flow control. At this time, the traffic in state 3 with a smaller flow meets the traffic in state 2' with a congested queue. The two traffic flows in different states meet, generating a shock wave w 32' ,like Figure 7 As shown by line 3-2'. When the traffic flow in state 3 meets the traffic flow in state 2', a shock wave w is generated. 32' , because Q3<Q 2' ,K3<K 2' , shock wave 32' Wave speed for:

[0070]

[0071] Shockwave The wave speed is greater than 0, indicating that the shock wave w 32' Properties and shock waves 13 Similarly, there will be no congestion at the end of the team, and the shock wave w 32' For the forward wave, the queue length no longer increases, and the tail position of the queue begins to move forward, that is, the length of the vehicle queue begins to decrease.

[0072] When the accident vehicles occupying the road are cleared, the traffic capacity of the accident section returns to normal. At this time, the traffic flow in the tunnel is as follows: Fig.11 As shown, the flow-density of each state is as follows Fig.10 shown.

[0073] When the accident vehicles occupying the road are cleared, the traffic capacity of the accident section is reduced by q m2 Increase to q m1 , the traffic capacity returns to normal, and the traffic flow of the accident section begins to increase. Assuming that the traffic flow state of the accident section reaches saturation, the traffic flow state from the accident section to the flow control point can be divided into three categories, state 4 with saturated flow rate, state 2' of congested queue, and state 3 with small flow. The vehicles queued upstream begin to speed up gradually, so the accident section and downstream change from the original state 2 to state 4, and the traffic flow in state 2' changes to state 4, generating a shock wave w 42' ,like Fig.10 As shown in lines 2'-4.

[0074] At this time, Q4>Q 2' ,K4<K 2' , shock wave 42' Wave speed for:

[0075]

[0076] Shockwave 42' The wave speed is less than 0, indicating that the shock wave w 42' It is a receding wave, which propagates in the opposite direction of the traffic flow until it meets the shock wave. meet;

[0077] The traffic flow in state 3 after control meets the saturated traffic flow in state 4. At this time, the traffic flow in the tunnel is as follows: Fig.13 As shown, the flow-density of each state is as follows Fig.12 shown.

[0078] At this time, Q4>Q3, K4>K3, shock wave w 43 Wave speed for:

[0079]

[0080] Shockwave43 The wave speed is greater than 0, indicating that the shock wave w 43 It is a forward wave that propagates in the direction of traffic flow. The queue ends and moves forward until the vehicle at the end of the queue passes the accident section. At this time, the congestion queue caused by the accident is considered to have dissipated.

[0081] S3: Based on the shock wave theory, a tunnel traffic accident vehicle queue length model is constructed to predict the maximum queue length of vehicles;

[0082] 1. Geometric analysis

[0083] Combining the traffic flow shock wave theory, the evolution of the vehicle queue length under the traffic accident analyzed in the previous section, and the changing characteristics of flow, density, and speed during the whole process, a time-space trajectory line diagram of the vehicles during the whole accident process is drawn, such as Fig.14 As shown, the horizontal axis represents time, and the vertical axis represents the distance between the accident site and the tunnel entrance.

[0084] Fig.14 Middle: t A Indicates the time when the accident occurred; t B Indicates the start time of traffic control; t C represents the time when the controlled traffic flow meets the queued traffic flow, that is, the time when the queue length reaches the maximum; t D Indicates the accident clearance time; t E Represents shock wave w 32' With shock wave 42' Time of meeting; t F It indicates the time it takes for the vehicle at the end of the queue with the maximum queue length to pass through the accident section. The different sparseness of the lines on the figure represent different traffic flow states. Area 1 represents the traffic flow state of normal operation of the tunnel before the accident. Area 2' represents the traffic flow state of vehicle queuing congestion caused by the flow rate being greater than the remaining capacity of the accident section after the accident. 2" represents the traffic flow state downstream of the accident section. Area 3 represents the traffic flow state of reduced flow after the flow control point takes flow control measures after the accident. Area 4 represents the traffic flow state after the accident is cleared and the capacity of the accident section is restored.

[0085] On the vehicle time-space trajectory diagram, line AC represents the shock wave w generated when state 1 and state 2' meet. 2'1 The propagation path is the path where vehicles start to queue. C After the time control, the traffic flow meets the queued traffic flow. At this time, the traffic flow reached is less than the traffic flow leaving the accident section. The queue no longer increases and the queue length reaches the maximum. Line CE can represent the shock wave w generated when state 3 and state 2' meet. 32' Line DE represents the shock wave w generated when state 2' and state 4 meet. 2'1Line EF represents the shock wave w generated when state 3 and state 4 meet. 43 The flow-density relationship of each state is as follows: Fig.15 shown.

[0086] The estimation formula of each shock wave velocity (taking absolute value) is as follows:

[0087]

[0088]

[0089]

[0090]

[0091] It can be uniformly expressed as:

[0092]

[0093] Where q i is the flow rate under traffic state i, veh / h, i=1,2,…,n; K i is the density of traffic flow in state i, veh / km, i=1,2,…,n; w ij is the shock wave generated when traffic state i meets traffic state j, i = 1, 2, ..., n, j = 1, 2, ..., n; The shock wave w ij Wave speed, km / h, i=1,2,…,n, j=1,2,…,n.

[0094] The time t at which the traffic flow after control meets the queued traffic flow C The estimation formula is:

[0095]

[0096] Through the geometric analysis of the vehicle time-space running trajectory line graph, we can get:

[0097]

[0098] Model the maximum queue length:

[0099]

[0100] Due to t A The time when the accident occurred, Simplified to:

[0101]

[0102] In the above formula, lmax is the maximum queue length, km; L is the distance from the accident location, which is defined here as the distance between the accident location and the tunnel entrance, km; V3 is the vehicle speed after control, km / h; t A is the time of the accident; t B is the start time of traffic control; t C is the time when the controlled traffic flow meets the queued traffic flow, that is, the time when the queue length reaches the maximum; t D is the end time of the accident; t E is the time when the two evanescent waves meet; For the gathering wave w 2'1 The speed is km / h. The time unit corresponds to the time in the flow, density and speed units.

[0103] Substitute the maximum queue length into the formula The time t when the maximum queue length is generated can be obtained C In addition, the time t when the two evanescent waves meet can be obtained by geometric analysis of the vehicle's time-space trajectory diagram. E , the maximum queue length of the rear vehicle passing the accident section, the time t for the rear vehicle to pass the accident section F And the duration of the accident impact, the formula is as follows:

[0104]

[0105]

[0106] t max =t A -t F

[0107] In the above formula, l max is the maximum queue length; t C is the time when the controlled traffic flow meets the queued traffic flow, that is, the time when the queue length reaches the maximum; t D is the end time of the accident; t E : the time when two evanescent waves meet; t F t is the time it takes for the vehicle at the end of the maximum queue to pass the accident section; max The duration of the accident impact; The shock wave w 32' speed; For shock wave w 42' speed; For shock wave w 43 speed.

[0108] Through the above estimation, the spatial scope l affected by the accident can be obtained max , and the duration of the accident impact t max .

[0109] 2. Model building

[0110] After the geometric analysis of the whole process of queue change, a maximum queue length model with flow, density, speed, and the distance between the accident location and the opening as variables can be constructed, as shown in the following formula:

[0111] l max =(q u -q jam )[L+v u (t B -t A )] / (v u k u -v u k jam +q u -q jam )

[0112] And order:

[0113] v u k u =q u , Δv=t B -t A

[0114] Therefore, the formula l max =(q u -q jam )[L+v u (t B -t A )] / (v u k u -v u k jam +q u -q jam ) is simplified to:

[0115] l max =(q u -q jam )(L+v u Δt) / (2q u -v u k jam -q jam )

[0116] In the above formula, q u is the upstream traffic volume in the tunnel, veh / h; v u is the upstream speed in the tunnel, km / h; k u is the upstream traffic density in the tunnel, veh / km; q jam is the remaining traffic capacity at the accident point, veh / km; kjam is the blocking density, veh / km; L is the distance of the accident location, which is defined here as the distance between the accident location and the tunnel entrance, km; Δt is the start time of control, which is the difference between the start time of control and the time of accident occurrence.

[0117] By formula l max =(q u -q jam )(L+v u Δt) / (2q u -v u k jam -q jam ) It can be seen that the maximum queue length is affected by factors such as upstream traffic volume (the upstream speed and flow rate obey the Greenshields flow-speed model relationship, so the influence of upstream speed is no longer explored separately), remaining capacity, blockage density, the distance between the accident location and the tunnel entrance, and the start time of control.

[0118] S4: Combine the traffic flow data collected in the tunnel and introduce the sliding average filter method to build a real-time estimation model for the length of the vehicle queue and calculate it;

[0119] S5: using S4 to estimate in real time the relationship between the real-time queue lengths of the vehicles and the time relationship to determine the queue situation of the vehicles in the tunnel;

[0120] 1. Implementation process of real-time estimation of vehicle queue length

[0121] When an accident occurs in a tunnel, when the upstream traffic volume is greater than the remaining capacity of the accident section, the accident point forms a traffic bottleneck due to the decrease in capacity. A large number of vehicles cannot pass through in a timely and smooth manner, which will cause congestion, and vice versa. In addition, the vehicle queuing phenomenon caused by an accident in a tunnel is different from the queuing phenomenon caused by signal intersections. The vehicles queuing at signal intersections are completely stationary, and the traffic parameters such as flow and speed at this time are all zero. Therefore, the vehicle queuing situation can be judged based on the changes in traffic flow parameters, while the vehicles queuing in the case of an accident in a tunnel are not completely stationary, and the vehicle queuing situation cannot be accurately judged by the changes in traffic flow parameters. Therefore, the present invention establishes a real-time estimation model for vehicle queue length based on the shock wave theory, and the specific implementation process is as follows. Fig.16 shown.

[0122] After the accident occurs, the first step is to determine whether the traffic volume upstream of the accident section is higher than the remaining capacity of the accident section, so as to determine whether there will be a vehicle queue in the tunnel. When the traffic volume upstream of the accident section is higher than the remaining capacity of the accident section, the second step begins, and the length of the vehicle queue is estimated in real time. The queue length value changes with time. When the queue length value is less than 0, it is considered that the queue has dissipated, and the estimation of the vehicle queue length ends at this time.

[0123] 2. Real-time estimation model of vehicle queue length

[0124] Through the analysis of the evolution of the vehicle queue length, it is found that the main factors affecting the change of queue length are the velocity and direction of the shock wave between the congested area and the non-congested area, such as Fig.17 As shown in the figure, after the accident, when the shock wave propagates in the direction of the traffic flow, the queue length will decrease, and when it propagates in the opposite direction of the traffic flow, the queue length will increase.

[0125] Real-time estimation of queue length requires real-time input of traffic flow parameters. Video detectors in tunnels can be used to obtain real-time flow, density and other data within a certain time interval, so as to estimate the real-time wave velocity of shock waves generated by the encounter of vehicles in different traffic states. To facilitate data acquisition, the detectors from the accident point (or near the upstream) to the flow control point are numbered 1, 2, ..., S, the average distance between adjacent detectors is s, and the time after the accident is divided into 1, 2, ..., m time periods with a time interval of T. The flow of the detector upstream near the tail of the congested queue is selected as the background traffic demand at this time, and the average flow of each detector in the congested area is taken as the queue flow at this time. The density parameters are the same.

[0126] After the accident, when the background traffic demand is greater than the remaining capacity of the accident section, queues begin to form. In the mth time interval T, the queue flow and density in the congestion area are:

[0127]

[0128]

[0129] In the above formula represents the queue flow in the mth time interval of the congestion area, veh / h; The queue density at the mth time interval in the congestion area, veh / km; Q(S,T m ) represents the flow rate of the detector with position number S in the mth time interval, veh / h; K(S,T m ) represents the density of the detector with position number S in the mth time interval, veh / km; S is the detector number, S = 1, 2, ..., n; T mis the mth time interval, m=1,2,…,n. Indicates the detector number at the end of the queue in the congested area, which is also the number of detectors in the congested area.

[0130] The background traffic volume and density in the non-congested area are:

[0131]

[0132]

[0133] In the above formula Indicates that the traffic flow in the non-congested area state j is in T m The flow rate in the time interval, veh / h, j = 1, 2, ..., n; Indicates that the traffic flow in the non-congested area state j is in T m Density in the time interval, veh / km, j = 1, 2, …, n; Indicates the detector number in the non-congested area.

[0134] At this time, the shock wave w ij The wave speed and the changing length of the vehicle queue within this time interval are:

[0135]

[0136]

[0137] In the above formula Represents shock wave w ij In T m Wave speed in the time interval, km / h, i = 1, 2, ..., n, j = 1, 2, ..., n; l(T m ) represents the change in queue length in the mth time interval T, km.

[0138] In summary, the real-time queue length estimation model is:

[0139]

[0140] Where l(t) represents the real-time queue length, km; n represents the number of time intervals; q u represents the flow rate upstream of the tunnel, veh / h; q jam Indicates the remaining capacity of the tunnel at the accident point, veh / h.

[0141] The accuracy (AC) of the estimated vehicle queue length is obtained by estimating the relative error (RE). Relative error is a simple statistical measurement method to estimate the error between the estimated value and the actual value. The relative error estimation formula of the maximum queue length is: The corresponding accuracy estimation formula is:

[0142] in, Indicates the relative error of the maximum queue length of vehicles; It represents the estimated value of the maximum queue length of vehicles; represents the actual measured value of the maximum queue length of vehicles; Indicates the accuracy of the maximum queue length of vehicles.

[0143] The relative error estimation formula of real-time queue length is: The accuracy estimation formula of the vehicle queue length in the whole process is:

[0144] Among them, RE queue (T m ) represents the relative error of the length of the vehicle queue at the mth time interval; represents the estimated value of the vehicle queue length at the mth time interval; represents the actual measured value of the length of the vehicle queue at the mth time interval; AC queue Indicates the accuracy of the vehicle queue length during the entire process.

[0145] When a queue is generated, the shock wave velocity is negative, and the estimated length of the vehicle queue is also negative; when the shock wave velocity is positive, the length of the vehicle queue begins to shorten until it shortens to zero. Therefore, the queue length estimated by the model is negative. For ease of understanding, the absolute value of the vehicle queue length is taken. At the same time, in order to ensure the accuracy of the model, an initial calibration value of the queue length needs to be given, that is, the first queue length value estimated by the model needs to be calibrated according to the measured value.

[0146] 3. Method for determining time interval

[0147] Due to the randomness of vehicle arrival, the traffic and density data extracted in different time intervals are significantly different, and the real-time queue lengths estimated at different time intervals T are different. In order to improve the accuracy of the model and explore the value of the optimal time interval T, the present invention is based on an accident data from a typical extra-long tunnel. Through the real-time queue length estimation model, the queue length is estimated at time intervals T of 5s, 15s, 30s, and 60s. The initial calibration value of the queue length is different for different time intervals and estimation times. The estimation results are as follows: Fig.18 As shown, the accuracy of different time intervals T is shown in the following table.

[0148] Comparison of accuracy at different time intervals T

[0149]

[0150] Depend on Fig.18 It can be seen that when the time interval T is 30s, the estimated queue length is in good agreement with the measured queue length. As shown in Table 5.1, when the time interval T is 30s, the accuracy of the maximum queue length is 95.62%, and the accuracy of the entire queue length is 84.34%. In general, when the time interval T is 30s, the accuracy of the maximum queue length is the highest, and the accuracy of the entire queue length is also the highest; 60s and 15s are the second most accurate.

[0151] 4. Introduce sliding average filter to improve the model

[0152] ①. The principle of sliding average filtering

[0153] The principle of sliding average filtering is: dynamic test data is divided into deterministic components and random components. The deterministic component data is a high-accuracy measurement result or effective signal, and the random component data is a randomly fluctuating test error or noise. After discrete sampling, the dynamic test data can be expressed as:

[0154] y j =f j +e j j=1,2,…,N

[0155] y j is the dynamic test data; f j is the deterministic component of dynamic test data; e j It is the random component of dynamic test data.

[0156] ② Improve model building

[0157] In order to obtain more accurate dynamic test data, it is necessary to suppress the error caused by the random component as much as possible while ensuring the deterministic component of the dynamic test data. j Random errors are often reduced by smoothing and filtering. j The specific method is to use a non-stationary data y j , is considered to be close to stationary in an appropriate small interval, and some local average is made to reduce e j In this way, by continuously performing local averaging on the N data points along the entire length, a smoother measurement result f can be obtained. j , and filter out the random errors of frequent fluctuations. Therefore, in order to reduce the random errors of flow and density data, the improved vehicle queue length real-time estimation model with sliding average filtering is shown as follows:

[0158]

[0159] Where e(r) is the random component in the dynamic test data; r is the sliding average window length.

[0160] ③. Accuracy comparison

[0161] The window lengths of the sliding average are 30s, 60s, and 90s respectively. The queue length estimation results are obtained by introducing an improved model of sliding average filtering into the three sets of flow and density data after smoothing. This is to explore whether the random error can be reduced after smoothing and the optimal smoothing window length.

[0162] The queue lengths obtained after unsmoothing, smoothing for 30s, smoothing for 60s, and smoothing for 90s are compared with the measured queue lengths. Fig.19 It can be seen that when the time interval T is 30s, the queue length obtained after smoothing is higher than the measured queue length, and the queue length obtained without smoothing is closer to the measured queue length.

[0163] Comparison of queue length accuracy

[0164]

[0165] Combined with the above table, it can be obtained that after smoothing the 5s data, the accuracy is significantly improved when the time interval T is 5s, and the results of smoothing 30s, 60s and 90s are close, with the maximum queue length accuracy of 94.86% and the accuracy of the whole process queue length of 81.89%; when the time interval T is 15s, the rule is consistent with that when it is 5s, the maximum queue length accuracy is 93.22% and the accuracy of the whole process queue length is 82.23%. The accuracy decreases when the time interval T is 30s and 60s. This shows that when the time interval T is 5s and 15s, smoothing the flow and density data can improve the accuracy of model estimation, and the difference between the sliding average window length of 30s, 60s and 90s is not much.

[0166] In general, the overall accuracy is the highest when the time interval T is 30s and the data is not smoothed, and the accuracy of the maximum queue length and the whole process queue length are 92.22% and 84.34% respectively. Data smoothing does not improve the highest accuracy of the model, but it has a significant improvement when the time interval is 5s and 15s. Different sliding window lengths have no significant effect on the results.

[0167] S6: Take different control measures for different queue length development stages.

[0168] 1. Analysis of dynamic control strategies in long tunnel traffic accidents

[0169] ①、Queue growth period

[0170] like Fig. 20As shown in Figure 2, the time period from when the queue is generated to when the queue length reaches its maximum value is defined as the queue growth period. The judgment condition of the queue growth period can be obtained through the real-time estimation model of vehicle queue length:

[0171]

[0172] t A Indicates the time when the accident occurred; T indicates the time interval of real-time estimation.

[0173] After the accident, a traffic bottleneck was formed at the accident point, causing congestion and queues of vehicles upstream in the tunnel. During the period when the queue length increased, in order to ensure the safety of the tunnel, the lanes outside the tunnel should be closed and the vehicle input should be controlled to avoid further congestion after vehicles entered the tunnel.

[0174] ②、Queue dissipation period

[0175] like Fig.21 As shown in Figure 2, the time period from when the queue length reaches the maximum value to when the queue length drops to zero is defined as the queue dissipation period. The judgment condition of the queue dissipation period can be obtained through the real-time estimation model of vehicle queue length:

[0176]

[0177] During the queue length dissipation period, under the premise of tunnel lane closure, the blank section from the tunnel entrance to the end of the queue continues to grow. The maximum value of the blank section is related to the location of the accident. The farther the accident location is from the tunnel entrance, the larger the blank section length value, and the more serious the waste of time and space resources. Therefore, under the premise of ensuring tunnel safety, it is possible to consider opening the lanes outside the tunnel in advance during the queue dissipation period to allow some vehicles to enter the tunnel and reduce the waste of time and space resources.

[0178] This control concept needs to meet the conditions of optimal traffic efficiency and lowest operating risk.

[0179] ③、End of queue

[0180] When the queue ends, normal operation will resume in the tunnel and traffic control will end.

[0181] The dynamic control process considering the change of queue length is as follows Fig. 22 shown.

[0182] 2. Constraints of dynamic control model under traffic accidents in extra-long tunnels

[0183] ① Traffic efficiency function

[0184] In order to meet the conditions for optimal traffic efficiency, the present invention constructs a traffic efficiency function and seeks its optimal solution. Brilon proposed a traffic efficiency evaluation index that draws on the concepts of work and power in physics, and its expression is:

[0185] E=qvt

[0186] Where E is the traffic efficiency index, veh·km·h -1 ; q is the flow rate per unit time t; v is the average vehicle speed, km / h.

[0187] The above formula shows that traffic efficiency is related to flow and speed per unit time, and the relationship between speed and flow can be described by Greenshields' flow-speed model (as shown in the following formula).

[0188]

[0189] Combining the above flow-speed model with the traffic efficiency evaluation index, we can obtain the traffic efficiency objective function with speed and time as variables, as shown in the following formula:

[0190]

[0191] By the blocking density k j 、Smooth flow speed v f By calibrating the parameters and considering time t as unit time, that is, as a constant, the maximum value of the traffic efficiency function can be obtained. Fig.23 This is a traffic efficiency diagram obtained by calibrating the accident traffic flow parameters in a typical extra-long tunnel. As shown in the figure, as the speed increases, the traffic efficiency gradually increases, and after reaching the maximum value, it begins to decrease as the speed increases. This shows that the higher the speed, the higher the traffic efficiency. When the speed is equal to the optimal speed v e When , the traffic efficiency is maximum.

[0192] ②Running risk function

[0193] When there is a large speed difference between the vehicle entering the tunnel later and the vehicle inside the tunnel, the operational risk will increase. TTC (time to collision) is widely used in the field of traffic flow safety assessment. It is defined as the remaining time before a collision occurs when the speed of the rear vehicle is greater than that of the front vehicle if the two vehicles maintain this speed difference. The TTC estimation formula is as follows:

[0194]

[0195] Where Δl represents the distance between vehicles; v n represents the speed of the nth vehicle; v n-1 Represents the speed of the n-1th vehicle.

[0196] Therefore, the operation risk in the tunnel can be characterized by the speed difference between the vehicles that meet. Here, the traffic flow is defined as a macroscopic stable flow, and the differences between individual vehicles are ignored. The operation risk objective function with speed as the variable is constructed as shown in the following formula:

[0197] minf1(x)=Δv=v i -v j Δv≥0

[0198] Where Δv represents the speed difference, v i represents the average speed of traffic on the section affected by the accident in the tunnel, v j Indicates the average speed of the traffic outside the tunnel that will enter the tunnel. The greater the speed difference, the greater the operational risk.

[0199] 3. Construction of dynamic control model for traffic accidents in extra-long tunnels

[0200] ① Target constraints

[0201] The target constraints include traffic efficiency constraints and operation risk constraints. To ensure the optimal traffic efficiency of the tunnel, the formula Solve for the maximum value and get the optimal speed v for traffic efficiency e , that is, the control speed is best close to the optimal speed v for traffic efficiency e .

[0202] In terms of operational risk, the speed of entering the tunnel should meet the following conditions: To ensure driving safety in the tunnel, the speed of traffic entering the tunnel from outside the tunnel should not be higher than the maximum speed limit and not lower than the minimum speed limit, satisfying the following formula:

[0203] v min ≤v j ≤v max

[0204] To ensure smooth connection of traffic flow in the tunnel, the speed of traffic entering the tunnel from outside the tunnel should not be higher than the speed of traffic in the tunnel:

[0205] v j ≤v i

[0206] To ensure optimal traffic efficiency, the speed of vehicles entering the tunnel from outside the tunnel approaches the optimal speed for traffic efficiency:

[0207] v j →v e

[0208] Among them, the tail position speed v of the queued vehicle flow in the tunnel during the queue dissipation period is tailIt changes with time. In the process of traffic flow from congestion to parking and queuing to dissipation, the speed of the tail of the queue gradually increases from zero to the smooth speed. The law of speed change has a certain linear relationship with time. In the process of speeding up, in addition to the acceleration factor of the vehicle itself, the traffic environment in the tunnel is also affected. The viscosity coefficient h in physics can be used to represent the viscosity of the traffic flow in the extra-long tunnel during acceleration, and to describe the influence of the environment in the extra-long tunnel on the acceleration of the traffic flow from congestion to smooth flow. Therefore, the tail position speed v of the traffic flow considering the viscosity of the traffic flow in the extra-long tunnel is constructed. tail Calculation formula:

[0209]

[0210] Where a represents the vehicle acceleration. Generally, a car can accelerate from zero to 100 km / h within 10 seconds, with an average acceleration of 2.78 m / s^2. Here, the acceleration is defined as a constant. h represents the viscosity coefficient during the acceleration of traffic flow in a long tunnel, which can be obtained through parameter calibration.

[0211] In addition, whether to release in advance during the queue dissipation period also needs to meet the distance requirement between the accident location and the tunnel entrance. The operation risk is lowest when the speed difference between the traffic outside the tunnel entering the tunnel and the traffic inside the tunnel is less than or equal to zero, that is, the speed of the rear position of the traffic inside the tunnel is v when they meet. tail It has been increased to the smooth running speed v in the tunnel, satisfying the following constraints, such as Fig.24 shown.

[0212] Therefore, under the premise of ensuring the safety of tunnel traffic operation, the distance L between the accident location and the tunnel entrance needs to meet the following conditions:

[0213]

[0214] ②Model construction

[0215] Based on the above analysis, the following dynamic control model is established:

[0216]

[0217] Where C t (q,v) represents the recommended control values ​​of flow rate q and speed v at time t; q m1 Indicates the normal traffic capacity of the basic section of the tunnel; q jam Indicates the remaining traffic capacity of the accident section.

[0218] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0219] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for dynamic prediction and control of traffic congestion queues in extra-long tunnels, characterized in that: Firstly, the real-time traffic and density data of each section of the tunnel after the accident are collected. Then, based on the traffic flow shock wave theory, the following steps are combined to realize the evolution analysis and modeling of the queue length of vehicles in the extra-long tunnel due to traffic accidents. The steps include: S1: Divide the tunnel into n sections, filter the real-time traffic and density data of each section where the accident occurred, obtain smoothed data, and then qualitatively analyze the changing characteristics of the process from the generation to the spread and then to the dissipation of the vehicle queue under the traffic accident; S2: Analyze the length of the vehicle queue in the tunnel at different stages of the accident and determine the type of accident to determine the relationship between the background traffic volume and the remaining capacity of the tunnel. If the background traffic volume is greater than the remaining capacity, proceed to S3; S3: Based on the shock wave theory, a tunnel traffic accident vehicle queue length model is constructed to predict the maximum queue length of vehicles; S4: Combine the traffic flow data collected in the tunnel and introduce the sliding average filter method to build a real-time estimation model for the length of the vehicle queue and calculate it; S5: using S4 to estimate in real time the relationship between the real-time queue lengths of the vehicles and the time relationship to determine the queue situation of the vehicles in the tunnel; S6: Take different control measures for different queue length development stages; S3 includes geometric analysis and model building. The geometric analysis combines the traffic flow shock wave theory, the evolution process of the vehicle queue length under the traffic accident analyzed in the previous section, and the change characteristics of the flow, density, and speed during the whole process to make a time-space trajectory line diagram of the vehicle during the whole process of the accident. The model building is to construct a maximum queue length model with flow, density, speed, and the distance between the accident location and the hole as variables after the geometric analysis of the whole process of queue change. The steps S4-S5 include a vehicle queue length real-time estimation implementation process, a vehicle queue length real-time estimation model, a time interval determination method, and an improved model introducing a sliding average filter. In the vehicle queue length real-time estimation implementation process, when an accident occurs in the tunnel, when the upstream traffic volume is greater than the remaining traffic capacity of the accident section, the accident point forms a traffic bottleneck due to the decrease in traffic capacity, and a large number of vehicles cannot pass through in a timely and smooth manner, which will cause congestion, otherwise it will not; In the real-time estimation model of vehicle queue length, through the analysis of the evolution process of vehicle queue length, it is found that the main factors affecting the change of queue length are the size and direction of the shock wave velocity between the congested area and the non-congested area. After the accident, when the shock wave propagates in the direction of the traffic flow, the queue length will decrease, and when it propagates in the opposite direction of the traffic flow, the queue length will increase. In the time interval determination method, since the arrival of vehicles is random, the flow and density data extracted in different time intervals are significantly different, and the real-time queue lengths estimated at different time intervals T are different. The improved model introducing sliding average filtering specifically includes the principle of sliding average filtering, the establishment of improved model and accuracy comparison.

2. The method for dynamic prediction and control of traffic accident congestion queue in a super-long tunnel according to claim 1 is characterized in that: S1-S2 includes the generation of vehicle queues in the tunnel, the spread of vehicle queues in the tunnel and the dissipation of vehicle queues in the tunnel. The generation of vehicle queues in the tunnel is based on the basic traffic flow diagram, and the traffic flow state is divided by the relationship between flow and density. After the accident occurs, the tunnel management department will implement flow control at the flow control point to avoid a large number of vehicles from entering and causing increased congestion, and to avoid secondary accidents.

3. The method for dynamic prediction and control of traffic accident congestion queue in a super-long tunnel according to claim 1 is characterized in that: The S6 includes dynamic control strategy analysis under extra-long tunnel traffic accidents, dynamic control model constraints under extra-long tunnel traffic accidents and dynamic control model construction under extra-long tunnel traffic accidents. The dynamic control strategy analysis under extra-long tunnel traffic accidents specifically includes queue growth period, queue dissipation period and queue end. The dynamic control model constraints under extra-long tunnel traffic accidents specifically include traffic efficiency function and operation risk function. The dynamic control model construction under extra-long tunnel traffic accidents specifically includes target constraints and model construction.

Images