An abnormal traffic event identification system and method based on road network operation data fusion

Abstract

The application relates to the technical field of intelligent traffic management, in particular to an abnormal traffic event identification system and method based on road network operation data fusion. The system and method form a road network space-time state matrix by fusing multi-source road network operation data, identify abnormal traffic events, generate candidate traffic intervention schemes, determine intervention counterattack indexes through counterfactual simulation and coupling evaluation, output abnormal traffic event identification results and dynamic regulation suggestions after closed-loop regulation, the application generates a road network space-time state matrix through multi-source road network operation data fusion, improves the accuracy of abnormal traffic event identification, determines intervention counterattack indexes through counterfactual simulation and coupling evaluation, and regulates candidate traffic intervention schemes in a closed loop, thereby improving the robustness and executability of dynamic regulation suggestions.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management technology, and more specifically, to an abnormal traffic event identification system and method based on road network operation data fusion. Background Technology

[0002] As urban road traffic continues to expand, the operational status of the road network exhibits characteristics such as multi-source perception, rapid changes, and strong correlation propagation. Road detectors, video recognition equipment, floating car trajectories, navigation route requests, signal control records, checkpoint vehicle data, and traffic incident reporting information can reflect changes in road segment speed, traffic flow, node queuing, vehicle turning, route selection, and road capacity from different perspectives. By fusing and processing the above-mentioned multi-source road network operational data, a relatively complete expression of the road network operational status can be formed, providing a data foundation for the identification of abnormal traffic events and traffic organization adjustments.

[0003] Existing traffic management technologies can typically identify abnormal traffic events such as traffic accidents, temporary lane closures, and congestion spread based on information such as sudden speed drops, abnormal traffic flow, increased queue lengths, and reduced road capacity. These technologies can then be combined with route guidance, signal timing adjustments, ramp control, and speed limit control for traffic intervention. These technologies play a positive role in improving the efficiency of abnormal event detection and alleviating localized congestion.

[0004] However, in complex road networks, the impact of abnormal traffic events often extends beyond a single road segment. After receiving guidance, some traffic flows may concentrate on several alternative routes. The entrance nodes of these alternative routes, key intersections, and adjacent road segments may experience new operational risks due to sudden increases in turning traffic, queue overflow, or insufficient remaining capacity. If traffic intervention plans are generated directly based on the results of abnormal event identification without comparing and contrasting the non-interventional and interventional states, it becomes difficult to promptly determine whether intervention measures will trigger adverse effects such as concentrated guided traffic, insufficient capacity of alternative routes, or overflow from adjacent nodes.

[0005] Therefore, this application proposes an abnormal traffic event identification system and method based on road network operation data fusion. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide an abnormal traffic event identification system and method based on road network operation data fusion.

[0007] To achieve the above objectives, the present invention provides the following technical solution: An abnormal traffic event identification system based on road network operation data fusion includes: a data fusion module, used to fuse multi-source road network operation data of the target road network to generate a road network spatiotemporal state matrix of the target road network; The event recognition module is used to identify abnormal traffic events based on the road network spatiotemporal state matrix, generate an abnormal event profile, and generate at least one candidate traffic intervention plan based on the abnormal event profile. The counterfactual simulation module is used to construct, under the same counterfactual simulation baseline, a non-intervention counterfactual simulation state and an intervention counterfactual simulation state within a preset future time window for each candidate traffic intervention plan, and generate a candidate intervention response matrix based on the intervention counterfactual simulation state; The coupled evaluation module is used to perform coupled evaluation on candidate traffic intervention schemes based on the road network spatiotemporal state matrix and candidate intervention response matrix, to obtain the induced flow herding coefficient and alternative path remaining capacity coefficient corresponding to the candidate traffic intervention scheme, and further determine the intervention backlash index corresponding to the candidate traffic intervention scheme. The closed-loop control module is used to determine whether there is a risk of intervention backlash in candidate traffic intervention schemes based on whether the intervention backlash index meets the preset backlash judgment rules. When it is determined that there is a risk of intervention backlash in candidate traffic intervention schemes, it generates a parameterized control candidate traffic intervention scheme, feeds the parameterized control candidate traffic intervention scheme back to the counterfactual simulation module, and outputs abnormal traffic event identification results and dynamic control suggestions based on the recalculation results of the counterfactual simulation module and the coupled evaluation module.

[0008] In one embodiment, the road network spatiotemporal state matrix includes road segment operation status indicators, node queuing spillover indicators, path remaining capacity indicators, and road network topology indicators.

[0009] In one embodiment, the abnormal event profile is generated as follows: the road network spatiotemporal state matrix is ​​spatiotemporally anomaly detected to determine the road segment where the anomaly occurred, the time of the anomaly, the road segment affected by the anomaly, and the direction of anomaly propagation. Based on the road segment operation status index, node queuing spillover index, and path remaining capacity index, the scope of the anomaly impact and the initial risk level are determined to generate the abnormal event profile.

[0010] In one embodiment, the candidate traffic intervention scheme includes at least one of the following: a set of target alternative paths, the traffic flow to be induced, the induced ratio, the induced timing, the induced priority corresponding to each target alternative path, and traffic control parameters; wherein, the induced ratio includes the proportion of induced vehicles, the induced timing includes the induced release time and the induced release batch, and the traffic control parameters include at least one of the following: path restriction parameters, speed limit parameters, and ramp control parameters.

[0011] In one embodiment, the candidate intervention response matrix is ​​generated by extracting the predicted transfer scale, predicted transfer timing, transfer path distribution, and node turning increment of the induced traffic flow object relative to the target alternative path set within a preset future time window from the intervention counterfactual simulation state, and generating the candidate intervention response matrix according to the target alternative path set.

[0012] In one embodiment, the induced flow herding coefficient is obtained as follows: the path and time induced response distribution of the induced traffic flow object on the target alternative path set is constructed based on the candidate intervention response matrix, and the scale concentration sub-coefficient, time sequence synchronization sub-coefficient, path clustering sub-coefficient and turning impact sub-coefficient are determined based on the predicted transfer scale, predicted transfer time sequence, transfer path distribution and node turning increment, respectively. Then, the spatial correlation weight between each target alternative path is determined by combining the road network topology index, and the scale concentration sub-coefficient, time sequence synchronization sub-coefficient, path clustering sub-coefficient and turning impact sub-coefficient are coupled and fused to obtain the induced flow herding coefficient.

[0013] In one embodiment, the remaining capacity coefficient of the alternative path is obtained by: determining the basic carrying capacity of the target alternative path set within a preset future time window based on the path remaining carrying capacity index in the road network spatiotemporal state matrix, and determining the induced occupancy increment of the target alternative path set based on the predicted transfer scale, transfer path distribution and node turning increment in the candidate intervention response matrix. Based on the basic carrying capacity margin and the induced occupancy increment, the distribution of remaining storage capacity along the path and over time is constructed. Combined with the bottleneck constraints determined by the road segment operation status indicators, the node spillover impact determined by the node queuing spillover indicators, and the sustainability of the remaining storage capacity, the remaining storage capacity coefficient of the alternative path is obtained.

[0014] In one embodiment, the intervention backlash index is determined by comparing the uninterrupted counterfactual simulation state and the intervened counterfactual simulation state to obtain the risk reduction within the scope of abnormal influence determined by the abnormal event profile, the risk increase of the target alternative path set, and the risk increase of the spillover of adjacent nodes associated with the target alternative path set determined based on the road network topology index. The intervention backlash index is obtained by coupling the induced flow herding coefficient and the alternative path remaining capacity coefficient.

[0015] In one embodiment, the closed-loop control module is used to feed back the parameterized control candidate traffic intervention scheme to the counterfactual simulation module, so that the counterfactual simulation module reconstructs the intervention counterfactual simulation state and the coupled evaluation module recalculates the induced flow herding coefficient, the alternative path remaining capacity coefficient, the intervention backlash index, and the post-intervention risk level; wherein, the post-intervention risk level is the risk level corresponding to the reconstructed intervention counterfactual simulation state, and the initial risk level is the risk level determined by the abnormal event profile when the abnormal traffic event is identified; When the recalculation results indicate that the candidate traffic intervention scheme after parameterized control does not have the risk of intervention backlash, and the risk level after intervention is lower than the initial risk level, the closed-loop control module outputs the abnormal traffic event identification results and dynamic control suggestions.

[0016] In one embodiment, an abnormal traffic event identification method based on road network operation data fusion includes: Step 1: Perform fusion processing on the multi-source road network operation data of the target road network to generate the spatiotemporal state matrix of the target road network; Step 2: Identify abnormal traffic events based on the road network spatiotemporal state matrix and generate abnormal event profiles; Step 3: Generate at least one candidate traffic intervention plan based on the abnormal event profile; Step 4: For each candidate traffic intervention plan, under the same counterfactual simulation baseline, construct the non-intervention counterfactual simulation state and the intervention counterfactual simulation state within a preset future time window; Step 5: Generate a candidate intervention response matrix based on the counterfactual simulation state of the intervention; Step 6: Based on the road network spatiotemporal state matrix and the candidate intervention response matrix, perform a coupled evaluation of the candidate traffic intervention schemes to obtain the induced flow herding coefficient, the alternative path remaining capacity coefficient, and the intervention backlash index. Step 7: Determine whether the candidate traffic intervention plan has the risk of intervention backlash based on whether the intervention backlash index meets the preset backlash judgment rules. If the intervention backlash risk is determined, generate the parameterized and adjusted candidate traffic intervention plan and provide feedback. Output the abnormal traffic event identification results and dynamic control suggestions based on the recalculation results.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention identifies abnormal traffic events based on the spatiotemporal state matrix of the road network through an event recognition module, generates an abnormal event profile, and further generates at least one candidate traffic intervention plan, thus establishing a correspondence between the anomaly identification results and the traffic intervention plan. Simultaneously, through a counterfactual simulation module, it constructs non-intervention counterfactual simulation states and intervention counterfactual simulation states under the same counterfactual simulation benchmark conditions. This allows for comparative analysis and deduction of road network operation changes within a preset future time window before the implementation of candidate traffic intervention plans, improving the objectivity and comparability of candidate traffic intervention plan evaluations.

[0018] This invention uses a coupled evaluation module to evaluate candidate traffic intervention schemes based on the road network spatiotemporal state matrix and candidate intervention response matrix, obtaining the induced flow herding coefficient, the alternative path remaining capacity coefficient, and the intervention backlash index. This allows candidate traffic intervention schemes to not only be evaluated for their mitigation effect on abnormal traffic events, but also to be further judged whether they will cause induced flow concentration, insufficient alternative path capacity, or new congestion propagation risks, thus improving the precision of risk identification for traffic intervention schemes. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of an abnormal traffic event identification system based on road network operation data fusion according to the present invention; Figure 2 This is a schematic diagram illustrating the construction of the spatiotemporal state matrix of the road network in this invention; Figure 3 This is a schematic diagram illustrating how the coupling evaluation module in this invention determines the induced flow herding coefficient, the alternative path remaining capacity coefficient, and the intervention backlash index. Figure 4 This is a time-series diagram illustrating how the closed-loop control module generates parameterized control candidate traffic intervention schemes and outputs abnormal traffic event identification results and dynamic control suggestions in this invention. Figure 5 This is a flowchart illustrating an abnormal traffic event identification method based on road network operation data fusion according to the present invention. Detailed Implementation

[0020] Example 1: Refer to Figure 1 An abnormal traffic event identification system based on road network operation data fusion includes: Reference Figure 2 The data fusion module is used to fuse multi-source road network operation data of the target road network to generate a spatiotemporal state matrix of the target road network. It selects road detector data, navigation trajectory data, signal control data, floating car speed data, checkpoint vehicle passage data, and traffic incident reporting data within the target road network as multi-source road network operation data. This data is uniformly matched according to road segment number, node number, path number, and collection time. Missing values, abnormal jump values, and duplicate sampled values ​​are cleaned, and a continuous road network operation data sequence is formed according to a preset time granularity. In the target road network, the set of road segments is denoted as... Let the set of nodes be denoted as Let the set of paths be denoted as Let the set of time slices be denoted as The preset time granularity is denoted as Any road network object belong Any time slice belong The spatiotemporal state matrix of the road network is denoted as... Matrix elements For road network objects In time slice The state vector. ,in, The average speed of the road segment. Traffic flow on the road segment For road segment occupancy rate, The queue length of the node. The node queuing growth rate This refers to the remaining travel capacity of the path. Predict the saturation time for the path. This is a road network topology indicator. Let... It is a collection of data sources for multi-source road network operation data. For any data source, For data source In road network objects and time slices Indicators collected The original value, For time slices Centered on, with a length of Data source within the sliding time window For road network objects The collected sample set of similar indicators, This is the median of the sample set. This represents the absolute deviation of the median of the sample set. The threshold for determining abnormal jumps. and Indicators The physical lower limit and physical upper limit, The preset minimum value is greater than zero. Abnormal jump statistics. and the index values ​​after cleaning We obtain the following formula: For the same indicator Time slice Internally, it can provide road network objects. Indicators The data source set is denoted as fusion value fusion weight and reliability We obtain the following formula: in, For data source The missing rate within the preset statistical period, For data source Abnormal jump rate within a preset statistical period For data source Consistency coefficient with adjacent road network objects and adjacent time slice data. It is obtained as the ratio of the number of missing samples to the number of samples that should be collected. It is obtained by the ratio of the number of abnormal jump samples to the number of samples that should be collected. By data source The deviation between the cleaning value and the cleaning value of the same index of adjacent road network objects is obtained. The value ranges from zero to one. The larger the value, the more consistent the data source s is with the same type of indicator of the adjacent road network object. and The parameters are determined based on data acquisition frequency, equipment noise level, and traffic management platform configuration file. After the above processing, the road segment operation status indicators are... Composition, node queuing spillover index is composed of Composition, path remaining carrying capacity index is composed of Composition, road network topology indicators are composed of constitute.

[0021] Taking a target road network composed of urban expressways and adjacent arterial roads as an example, the average speed of road segments, traffic flow, road segment occupancy rate, node queue length, node queue growth rate, remaining capacity of paths, expected saturation time of paths, as well as road segment connectivity, node turning relationships, and path reachability, collected within three minutes, are written into the same spatiotemporal coordinate frame to form the spatiotemporal state matrix of the target road network. This spatiotemporal state matrix includes road segment operation status indicators, node queue spillover indicators, remaining capacity of paths indicators, and road network topology indicators. Among them, the road segment operation status indicators reflect the speed, traffic flow, and congestion changes of each road segment; the node queue spillover indicators reflect the degree of queue diffusion at intersections or ramp nodes; the remaining capacity of paths indicators reflect the remaining space that the target alternative path set can accommodate for traffic flow; and the road network topology indicators reflect the connectivity and spatial correlation strength between road segments, nodes, and paths. The directed graph corresponding to the target road network is denoted as: Road network topology indicators Including the road segment to node connection matrix Node-to-Road-Segment Connection Matrix Path to segment attribution matrix Path to Node Attribution Matrix . Indicates road segment The endpoint is the node. ,otherwise ; Indicates road segment The starting point is the node ,otherwise ; Representing a path Included road sections ,otherwise ; Representing a path Includes nodes ,otherwise .path The set of included road segments is denoted as ,path The set of included nodes is denoted as .path Path length ,path Number of nodes ,path With path Shared road length Number of shared nodes Path adjacency parameters We obtain the following formula: in, For road section The length of the road segment For path With path In directed graphs The shortest topological distance in, The threshold for adjacent ranges, Configuration is based on road grade, road network density, and the adjacent impact range determined by traffic management authorities. Road sections In time slice Maximum capacity Free flow velocity ,node Queue storage length ,node Conversion factor for unit vehicle queue length ,node Towards the path Maximum capacity for turning flow ,path In time slice The maximum amount of redirection traffic that the nodes can handle. and nodes For path Steering occupancy conversion factor We obtain the following formula: in, For road section The design capacity of the passageway For time slices Weather correction factor, For road section In time slice Event correction factor, It was obtained as the median ratio of historical measured traffic capacity to design traffic capacity under similar weather conditions. The data is obtained as the median ratio of historical measured capacity to design capacity under similar events; if no sample is available, one is taken. and The values ​​of all of them are in the range of zero to one; For road section In historical sample time slices Road segment occupancy rate This represents the upper limit of free flow occupancy. Determined based on the distribution of road segment occupancy in historical free-flow samples. For nodes The queue storage length corresponding to the entrance lane. This represents the average vehicle length. For nodes Average parking distance For nodes Towards the path saturated diversion flow rate, For nodes Towards the path Effective green light time For nodes In time slice The signal period, For nodes Towards the path Average turning delay for Quantile function.

[0022] The weights are not based on a single fixed empirical value. Let any weight reorganization be denoted as... , This refers to the number of evaluation components within the weighting restructuring. For the first The weights corresponding to each evaluation component. Let... This is a historical traffic intervention sample set. For any historical traffic intervention sample, Historical traffic intervention samples Risk label, This indicates that the historical traffic intervention sample carries the risk of backlash. This indicates that the historical traffic intervention sample does not pose a risk of backlash from the intervention. Historical traffic intervention samples The Middle A normalized evaluation component, The minimum number of samples required to initiate sample calibration. Here are the weight regularization coefficients. This is the set of weight constraints. It can be expressed as follows: Reorganization loss function It can be expressed as follows: when At that time, Internal selection The minimum value is used as the weighted reorganization. ;when At that time, the weighted reorganization adopted an equal initial value. Historical traffic intervention samples accumulated to Then, the loss function is recalibrated according to the above method. The above calibration rule is adapted to risk weights. Path scoring weights Induced flow sheep herd coefficient weighting Intervention backlash index weight Adjusting weights and spatial correlation component weights .in, This indicates the number of historical traffic intervention samples.

[0023] Various thresholds were determined based on historical sample distribution, traffic management false alarm tolerance, underreporting tolerance, and road operation safety requirements. For the tolerance rate of false alarms due to speed anomalies, Tolerance for false alarms due to queuing anomalies The tolerance rate for false alarms due to remaining load capacity anomalies. To intervene in the backlash and false alarm tolerance rate, To replace the low-capacity underreporting tolerance rate of the alternative path, The tolerance rate for false alarms from adjacent nodes. and The tolerance for false alarms and false alarms is determined by the traffic management department and the road classification configuration file. When historical samples are lacking, historical samples of roads of the same road classification or of the same type in adjacent areas are used for migration calibration; if samples are still lacking, the initial quantiles in the traffic management department's configuration file are used, and updated according to the historical sample distribution after sample accumulation.

[0024] The event identification module is used to identify abnormal traffic events based on the road network spatiotemporal state matrix, generate abnormal event profiles, and generate at least one candidate traffic intervention plan based on the abnormal event profiles; it also performs joint analysis on the road segment operation status indicators, node queuing spillover indicators, path remaining capacity indicators, and road network topology indicators for each time period in the road network spatiotemporal state matrix. Historical similar periods The time range is determined by the date type, time interval, and traffic management calendar. The date type includes weekdays, rest days, and public holidays. The length of the time interval is related to the preset time granularity. Same. Let's assume... For similar periods in history The corresponding set of historical sample time slices, For road section In historical sample time slices The average speed of the road section For path In historical sample time slices The remaining path capacity, For nodes In historical sample time slices The node queue length. Speed ​​baseline value. Remaining route capacity baseline and node queue length benchmark value We obtain the following formula: The section of road In time slice The average speed of the road segment is denoted as , will node In time slice The node queue length is denoted as , will node The queue storage length is denoted as , path In time slice The remaining path capacity is denoted as The percentage decrease in speed Queue growth rate Remaining load reduction ratio We obtain the following formula: set up For road section The rate of decrease in the historical sample set, For nodes The historical sample set of queuing growth rates For path The historical sample set of the remaining carrying capacity decline ratio. This is a set of risk scores for historical abnormal traffic incident samples. Each from the same historical period composition, It consists of initial risk scores from a sample of historical abnormal traffic events. Let... The threshold for speed decrease, The threshold for queue growth The threshold for the remaining load decreases. and We obtain the following formula: When the same road section exist Within a consecutive time slice, satisfy and at least one node connected to the road segment. satisfy or at least one path associated with that road segment. satisfy At that time, it was determined that there was an abnormal traffic incident on that section of the road. The number of consecutive sampling periods. The duration of the anomaly is determined according to the configuration file on the traffic management platform. The scope of the anomaly's impact is denoted as... , Composed of the road segment where the anomaly occurred, and the road segment topologically connected to the road segment where the anomaly occurred and or Composition of road segments, nodes, and paths affected by anomalies that meet the corresponding thresholds. Scope of anomaly impact. The set of road segments within is denoted as The set of nodes is denoted as The path set is denoted as Initial risk score We obtain the following formula: in, Indicates the number of elements in the set. Risk weighting And all risk weights are not less than zero. Let the low-risk threshold be... The medium-risk threshold is The high-risk threshold is . and We obtain the following formula: The time is low risk. The risk level is currently medium. It is a high-risk time. This is at an extremely high risk.

[0025] Based on the spatial location and temporal order of the abnormal data, the road segment where the abnormal occurred, the time of the abnormal occurrence, the road segment affected by the abnormal, and the direction of the abnormal propagation are determined. Combined with the congestion level reflected by the road segment operation status indicators, the queuing spread level reflected by the node queuing spillover indicators, and the carrying capacity reflected by the path remaining carrying capacity indicators, the scope of the abnormal impact is delineated and the initial risk level is determined to form an abnormal event profile. Taking the traffic congestion near the exit ramp of an urban expressway as an example, the abnormal event profile records that the abnormal section is one kilometer before the expressway exit, the abnormal time is 7:45 am during the morning rush hour, the abnormal affected sections include the exit ramp, adjacent main lanes and connecting main roads, and the abnormal propagation direction is from the exit node to the upstream main line and downstream intersection. Based on the abnormal event profile, at least one candidate traffic intervention plan is generated. The candidate traffic intervention plan includes at least one of the following: a set of target alternative paths, the induced traffic flow objects, the guidance ratio, the guidance time sequence, the guidance priority corresponding to each target alternative path, and traffic control parameters. The guidance ratio includes the proportion of induced vehicles, the guidance time sequence includes the guidance release time and the guidance release batch, and the traffic control parameters include at least one of the following: path restriction parameters, speed limit parameters, and ramp control parameters. set up Time slices for identifying abnormal traffic events To preset the length of the future time window, are positive integers and , Let this be the set of time slices within a pre-defined future time window. The set of traffic flow objects within the target road network. For any traffic flow object, Traffic flow objects The starting point Traffic flow objects The end point Traffic flow objects Planned path in the absence of intervention, For path In time slice The estimated travel time, Traffic flow objects The original planned path in time slice The estimated travel time, This is the upper limit for the increase in detour time. The lower limit for filtering the remaining storage capacity of the path. and The route is determined based on the traffic management department's tolerance for detours and the capacity requirements of alternative routes. In time slice node queuing spillover ,path In time slice Expected travel time ,path In time slice Foundation bearing capacity ,path In time slice Path theory carrying capacity limit Path in uninterrupted state In time slice Remaining storage capacity ,path In time slice Uninterventional remaining storage capacity ratio We obtain the following formula: Traffic flow objects Corresponding target alternative path candidate set , No. The set of traffic flow objects for each candidate traffic intervention scheme and the set of alternative paths to the target We obtain the following formula: in, Representing a path Without passing through the abnormal influence range The road sections within the area. After the candidate traffic intervention plan is determined, the set of target alternative routes is denoted as... and based on Calculate the set of neighboring nodes and path The set of neighboring nodes . For the set of alternative paths to the target The shortest topological distance is no greater than The set of nodes, For path The shortest topological distance is no greater than Let be the set of nodes. For path In time slice Any index to be normalized, , Indicates the remaining travel capacity of the path. Indicates the average speed along the path. Indicates the degree of node queuing overflow. Indicates the estimated travel time. Normalized index. Path scoring ,path Induction priority normalization ratio and path Induction priority We obtain the following formula: in, As path scoring weights, And the score weight of each path is not less than zero. Let... Traffic flow objects In time slice Traffic demand, This represents the total traffic demand of the guided traffic flow within a preset future time window. The total remaining storage capacity of the target alternative path set within a preset future time window. The minimum proportion of vehicles allowed by the traffic management platform. This represents the maximum proportion of vehicles that can be guided by the traffic management platform. and path The proportion of guided vehicles We obtain the following formula: in, For the first The maximum proportion of vehicles allowed by each candidate traffic intervention scheme. For the first The minimum proportion of vehicles allowed by each candidate traffic intervention scheme For the first Risk response coefficient of each candidate traffic intervention scheme For the first The total proportion of vehicles induced by each candidate traffic intervention scheme Scope of abnormal impact In time slice The initial risk score. and The number of guidance release batches is determined by the road management platform's guidance release strategy, road grade, and traffic safety management requirements. and release time of each batch According to the maximum number of guidance targets allowed to be published in a single batch by the traffic management platform. and Sure; For example, some vehicles heading towards the congested exit are considered as the target traffic flow. Two parallel main roads and one detour branch road are combined into a set of target alternative paths. The proportion of guided vehicles is set according to the remaining capacity index of each target alternative path. The timing of batch guidance release is set according to the direction of abnormal propagation. At the same time, ramp control parameters and speed limit parameters are configured to form a candidate traffic intervention plan that matches the abnormal traffic event.

[0026] The counterfactual simulation module is used to construct, under the same counterfactual simulation baseline conditions, a non-intervention counterfactual simulation state and an intervention counterfactual simulation state within a preset future time window for each candidate traffic intervention scheme, and to generate a candidate intervention response matrix based on the intervention counterfactual simulation state. The target road network's spatiotemporal state matrix at the time the abnormal traffic event is identified is used as the simulation starting state. The same traffic demand input, signal timing baseline, road traffic rules, weather conditions, traffic flow composition ratio, and preset future time window are used as the counterfactual simulation baseline conditions. Without loading candidate traffic intervention schemes, the module simulates the natural diffusion process of the abnormal traffic event in the target road network, forming the non-intervention counterfactual simulation state. With candidate traffic intervention schemes loaded, the target alternative path set, the induced traffic flow objects, the induced ratio, the induced timing sequence, the induced priority of each target alternative path, and traffic control parameters are written into the simulation constraints, forming the intervention counterfactual simulation state. For example, when abnormal congestion occurs at the exit of an expressway, the preset future time window is set to thirty minutes. The counterfactual simulation state without intervention records the growth of the main line queue after vehicles continue to gather at the congested exit. The counterfactual simulation state with intervention records the speed, flow and node queue changes of some induced traffic flow objects after they enter the two parallel main roads according to the time and batch of the induced release. Generate a candidate intervention response matrix based on the counterfactual simulation state of the intervention; No. The candidate traffic intervention plans are denoted as , ,in, For the set of alternative paths to the target, For the set of guided traffic flow objects, For path The proportion of guided vehicles, For the first The release time of each induced release batch. path The induction priority, This is a set of traffic control parameters. Including path restriction parameters Speed ​​limit parameters and ramp control parameters The uninterrupted counterfactual simulation state is denoted as... , will load the first The counterfactual simulation state of the intervention after each candidate traffic intervention plan is denoted as... The traffic demand input in the counterfactual simulation baseline is determined by navigation path requests, checkpoint vehicle passage data, floating car trajectories, and historical traffic flow during similar time periods. Let... To intervene in the counterfactual simulation state Downstream traffic flow objects In time slice The chosen alternative path For uninterrupted counterfactual simulation state Downstream traffic flow objects In time slice The original path. Node redirection traffic is denoted as... ,in, To intervene in the counterfactual simulation state Next node Towards the path The turning traffic, For uninterrupted counterfactual simulation state Next node Towards the path The turning flow. The candidate intervention response matrix is ​​denoted as... Matrix rows correspond Matrix columns correspond Matrix elements It can be expressed as follows: path In time slice Predicted transfer scale Original path Switch to target alternative path Traffic demand Vehicle transfer ratio Distribution of transfer paths and node turning increment We obtain the following formula: in, The set of original paths involved by the induced traffic flow object in the uninterrupted counterfactual simulation state. For nodes Towards the path The shift increment, For nodes Towards the path The positive shift in increment, For path The node redirects to the incremental summary value. and Using the same initial simulation conditions, traffic demand input, signal timing reference, road traffic rules, weather conditions, traffic flow composition ratio, and preset future time window, the only difference between the two simulations is whether or not candidate traffic intervention schemes are loaded. The signal timing reference is determined by the target road network during the abnormal traffic event identification time slice. The signal control data is determined, the road traffic rules are determined by road grade, lane attributes, turning restrictions and path restriction parameters, the weather conditions are obtained by the traffic management platform or meteorological interface, and the traffic flow composition ratio is obtained by checkpoint vehicle data and floating car trajectory statistics.

[0027] Taking three target alternative paths as an example, if the first target alternative path receives 40% of the induced traffic flow in the first five-minute time slice, the second target alternative path receives 35% of the induced traffic flow, and the third target alternative path receives 25% of the induced traffic flow, and there is a significant increase in left-turn traffic at the entrance node of the first target alternative path, then the candidate intervention response matrix synchronously records the predicted transfer scale, predicted transfer timing, distribution of the turning path, and node turning increment of the path, providing a quantitative basis for the coupled evaluation of candidate traffic intervention schemes.

[0028] Reference Figure 3 The coupled evaluation module is used to perform coupled evaluation on candidate traffic intervention schemes based on the road network spatiotemporal state matrix and candidate intervention response matrix, to obtain the induced flow herding coefficient and alternative path remaining capacity coefficient corresponding to the candidate traffic intervention scheme, and further determine the intervention backlash index corresponding to the candidate traffic intervention scheme. Expand the candidate intervention response matrix according to the target alternative path set and the preset future time window to construct the path and time-based induced response distribution of the induced traffic flow object on the target alternative path set; For the One candidate traffic intervention plan and target alternative route The proportion of the transfer scale within the preset future time window is denoted as . Time slice The proportion of the transfer scale is denoted as The calculation formula is as follows: Size Concentrator Coefficient Timing Synchronization Sub-coefficients Path segregation coefficient Steering impact coefficient We obtain the following formula: set up These are the spatial association component weights corresponding to the shared road segment length, the number of shared nodes, and the path adjacency relationship, respectively. , All are not less than zero. Target alternative path and Spatial correlation weight between We obtain the following formula: Spatial correlation correction coefficient Induced flow flock coefficient and path Contribution to the sheep herd coefficient of induced flow We obtain the following formula: in, These are the weights of the scale concentration sub-coefficient, the temporal synchronization sub-coefficient, the path segregation sub-coefficient, and the turning impact sub-coefficient, respectively. And each weight is not less than zero. and The value range is from zero to one.

[0029] Target Alternative Path In time slice Induced occupancy increment Path and time remaining storage capacity Bottleneck road constraints Node spillover effects Sustainability of remaining storage capacity We obtain the following formula: Path weight and time weight Allocation based on predicted transfer scale in the candidate intervention response matrix; remaining storage capacity coefficient of alternative paths. and path Remaining storage capacity coefficient We obtain the following formula: in, and The values ​​range from zero to one; the larger the value, the larger the set of alternative paths or paths to the target. To maintain a higher remaining storage capacity within a preset future time window.

[0030] Based on the path remaining capacity index in the road network spatiotemporal state matrix, the basic capacity reserve of the target alternative path set within a preset future time window is determined. Based on the predicted transfer scale, transfer path distribution, and node turning increment in the candidate intervention response matrix, the induced occupancy increment of the target alternative path set in each time slice is determined. The basic capacity reserve and the induced occupancy increment are deducted accordingly to form the path and time remaining capacity distribution. At the same time, the bottleneck road section constraints determined by the road section operation status index, the node spillover impact determined by the node queuing spillover index, and the sustainability of remaining capacity are incorporated to obtain the alternative path remaining capacity coefficient. For example, among the three target alternative routes, although the first target alternative route has a relatively high remaining capacity, the left turn queue at the entrance node is close to the overflow threshold, and its remaining capacity evaluation needs to deduct the impact of node overflow. The second target alternative route has a stable road speed and a relatively slow increase in node queues, so its remaining capacity evaluation is relatively high. The third target alternative route has a short-term bottleneck section, so its remaining capacity sustainability evaluation is correspondingly reduced. Within the scope of the abnormal impact determined by the abnormal event profile, calculate the risk reduction of the intervened counterfactual simulation state relative to the non-interventional counterfactual simulation state, calculate the risk increase caused by the predicted transfer scale, the distribution of the transfer path and the node turning increment within the target alternative path set, and calculate the spillover risk increase of adjacent nodes associated with the target alternative path set based on road network topology indicators. Any region In simulation state Next, Time Slice Risk score Calculated based on the degree of speed reduction, flow saturation, queuing overflow, and remaining capacity insufficiency. Let... For the region The collection of road sections within, For the region The set of nodes within, For the region The set of paths within, Simulation state Lower section In time slice The average speed of the road section Simulation state Lower section In time slice Traffic flow on the road section Simulation state Next node In time slice The node queue length, Simulation state Next path In time slice The remaining path capacity, For path In time slice The baseline remaining capacity, and .area The degree of average speed decrease Average flow saturation Average queue spillover and average remaining load-bearing capacity We obtain the following formula: in, As the risk score weight, Furthermore, the weight of each risk score is not less than zero. The amount of risk reduction within the scope of the abnormal impact. Increase in risk of the target alternative path set Increase in spillover risk from adjacent nodes ,path Increase in spillover risk associated with adjacent nodes We obtain the following formula: Intervention Backlash Index We obtain the following formula: in, The inducible flock coefficient, To replace the remaining storage capacity coefficient of the alternative path, The weights of the intervention backlash index are determined, and each weight is not less than zero. The value range is from zero to one. Let the comprehensive risk score after intervention be... and post-intervention risk level We obtain the following formula: in, In accordance with A function to determine the risk level. The initial risk level is denoted as... ,and ; For example, in the scenario of abnormal congestion at an expressway exit, candidate traffic intervention schemes will guide some vehicles to two parallel arterial roads and one detour branch road. If the queue on the main line within the abnormal impact area decreases significantly, but a large number of guided traffic flows turn into the first target alternative path in the same time slice, forming a significant increase in left turns at the entrance intersection, and the queues at adjacent nodes show a spillover and diffusion trend, then the herding coefficient of the guided flow increases, the remaining capacity coefficient of the alternative path decreases, and the intervention backlash index increases accordingly. If the guided traffic flows are evenly distributed on the target alternative path set, the guidance release time is dispersed, the node turning increment is lower than the node's carrying capacity, and the target alternative path set maintains a stable remaining capacity throughout the entire preset future time window, then the intervention backlash index remains at a low level.

[0031] Reference Figure 4 The closed-loop control module is used to determine whether there is a risk of intervention backlash in the candidate traffic intervention scheme based on whether the intervention backlash index meets the preset backlash judgment rules. When the candidate traffic intervention scheme is determined to have a risk of intervention backlash, a parameterized control candidate traffic intervention scheme is generated. The parameterized control candidate traffic intervention scheme is fed back to the counterfactual simulation module. Based on the recalculation results of the counterfactual simulation module and the coupled evaluation module, the abnormal traffic event identification results and dynamic control suggestions are output.

[0032] The intervention backlash index corresponding to the candidate traffic intervention plan is jointly compared with the preset backlash judgment threshold, the lower limit of the remaining storage capacity of the alternative path, the spillover growth limit of adjacent nodes, and the conditions for the change of risk level after intervention. set up Historical traffic intervention sample set The sample set in which there is a risk of intervention backlash. For any risk sample, For risk samples Intervention backlash index For risk samples Alternative path remaining storage capacity coefficient, For risk samples Increase in spillover risk from adjacent nodes. Intervention backlash index threshold. Alternative path remaining storage capacity lower limit and spillover growth limit of adjacent nodes We obtain the following formula: Candidate traffic intervention schemes are determined when any of the following conditions are met. There is a risk of backlash from intervention: When there is a risk of intervention backlash, follow the risk path. The risk sources are used to generate parameterized and controlled candidate traffic intervention schemes. Path Regulation risk level We obtain the following formula: in, For path The contribution of sheep herd coefficient to induced flow. For path The remaining storage capacity coefficient, For path The increase in spillover risk associated with adjacent nodes. In order to adjust the weight, And each regulatory weight is not less than zero. Let... For the set of controllable parameter types, For any type of control parameter, For regulating parameter types The corresponding adjustment step size, To adjust the lower limit of the step size, To adjust the upper limit of the step size, For regulating parameter types The corresponding historical effective control sample set, For any historical effective control sample, and Historical effective control samples Types of control parameters Parameter values ​​before and after adjustment, and They are respectively the types of control parameters The upper and lower limits, and Historical effective control samples Intervention backlash index before and after regulation Historical effective regulation sample Types of control parameters The normalized variation range, Historical effective regulation sample The extent of the decline in the backlash index due to intervention. Basic step size and current step size We obtain the following formula: Adjusted proportion of guided vehicles Inducing the release of batches , No. Release time of each induced release batch Induced priority Path restriction parameters Speed ​​limit parameters and ramp control parameters We obtain the following formula: in, and These represent the current adjustment step size for the proportion of guided vehicles, the batch of guided traffic releases, the time of guided traffic releases, the priority of guided traffic, route restriction parameters, speed limit parameters, and ramp control parameters, respectively. When historical effective adjustment samples are lacking, Use historical sample migration values ​​from the same road grade or control type; if samples are still lacking. Pick Parametric control employs boundary constraints. Let... , The lower realm For the upper realm, The parameters to be constrained are: The guided vehicle ratio, guidance priority, path restriction parameters, speed limit parameters, and ramp control parameters after boundary processing are denoted as follows: and and fell into and The boundaries of each parameter are determined by the road management platform, road grade, and traffic safety management requirements. Let... To control the number of iterations in a closed loop, This represents the maximum number of simulation iterations allowed by the road management platform. This represents the median decrease in the backlash index after a single iteration in the historical effective control sample. Determine using the following formula: when and Simultaneously, upon establishment, the closed-loop control module outputs abnormal traffic event identification results and dynamic control suggestions; when... At that time, the closed-loop control module outputs the candidate traffic intervention plan with the lowest current intervention backlash index and its risk warning.

[0033] The parameterized candidate traffic intervention schemes are fed back into the counterfactual simulation process to reconstruct the intervention counterfactual simulation state. The induced flow herding coefficient, alternative path remaining capacity coefficient, intervention backlash index, and post-intervention risk level are recalculated. The post-intervention risk level corresponds to the risk level of the reconstructed intervention counterfactual simulation state, while the initial risk level is the risk level determined by the abnormal event profile when the abnormal traffic event is identified. When the recalculation results show that the parameterized candidate traffic intervention scheme has no intervention backlash risk and the post-intervention risk level is lower than the initial risk level, the abnormal traffic event identification result and dynamic control suggestions are output. For example, the abnormal traffic event identification results record the abnormal road segment, the time of occurrence, the road segment affected by the abnormality, the direction of propagation of the abnormality, the scope of the abnormality, and the changes in risk level. The dynamic control suggestions record the adjusted set of target alternative paths, the traffic flow objects being induced, the induced ratio, the induced timing, the induced priority of each target alternative path, and the traffic control parameters, providing an executable basis for diversion release, ramp control, speed limit adjustment, and route restriction in traffic operation management.

[0034] Example 2: Refer to Figure 5 An abnormal traffic event identification method based on road network operation data fusion includes: Step one involves fusing multi-source road network operation data from the target road network to generate a spatiotemporal state matrix. This process integrates road network operation data from different sources within the target road network, unifying the dispersed data under the same target road network and spatiotemporal framework. The generated spatiotemporal state matrix centrally carries road segment operation status indicators, node queuing spillover indicators, path remaining capacity indicators, and road network topology indicators, laying the data foundation for a comprehensive characterization of the target road network's operation status.

[0035] Step two involves identifying abnormal traffic events based on the road network spatiotemporal state matrix and generating an abnormal event profile. Spatiotemporal anomaly detection is then performed within the road network spatiotemporal state matrix to identify abnormal traffic events deviating from normal operating conditions within the target road network. The abnormal event profile records the road segment where the anomaly occurred, the time of occurrence, the road segments affected, the direction of propagation, the scope of impact, and the initial risk level, forming a structured description of the abnormal traffic event.

[0036] Step 3: Generate at least one candidate traffic intervention plan based on the abnormal event profile. Based on the abnormal event profile's impact range, initial risk level, and target road network operation status, generate candidate traffic intervention plans corresponding to the abnormal traffic event. The candidate traffic intervention plans are configured around the target alternative path set, the induced traffic flow objects, the guidance ratio, the guidance timing, the guidance priority of each target alternative path, and traffic control parameters, so that the same abnormal traffic event has different intervention options.

[0037] Step four: For each candidate traffic intervention plan, under the same counterfactual simulation baseline, construct both a non-intervention counterfactual simulation state and an intervention counterfactual simulation state within a preset future time window. Under the same counterfactual simulation baseline, establish both the non-intervention counterfactual simulation state and the intervention counterfactual simulation state separately. The non-intervention counterfactual simulation state preserves the natural evolution of abnormal traffic events within the preset future time window, while the intervention counterfactual simulation state reflects the changes in road network operation after the intervention plan is implemented. Together, they form the basis for the comparative assessment of the impact of the candidate traffic intervention plan.

[0038] Step 5: Generate a candidate intervention response matrix based on the counterfactual simulation state. Using the counterfactual simulation state as a basis, extract the predicted transfer scale, predicted transfer timing, distribution of incoming paths, and node turning increments of the induced traffic flow relative to the target alternative path set within a preset future time window. The candidate intervention response matrix organizes the above information according to the target alternative path set, reflecting the response results of the candidate traffic intervention scheme to changes in traffic flow transfer and node turning in the target road network.

[0039] Step Six: Based on the road network spatiotemporal state matrix and the candidate intervention response matrix, a coupled evaluation of candidate traffic intervention schemes is conducted to obtain the induced flow herding coefficient, the alternative path remaining capacity coefficient, and the intervention backlash index. Combining the road network spatiotemporal state matrix and the candidate intervention response matrix, a coupled evaluation of traffic flow aggregation, alternative path capacity, and risk changes generated by the candidate traffic intervention schemes is also conducted. The induced flow herding coefficient reflects the concentration of induced traffic flow objects on the target alternative path set; the alternative path remaining capacity coefficient reflects the remaining capacity status of the target alternative path set within a preset future time window; and the intervention backlash index reflects the risk increase / decrease relationship formed between the candidate traffic intervention scheme and the abnormal impact range, the target alternative path set, and adjacent nodes.

[0040] Step 7: Based on whether the intervention backlash index meets the preset backlash judgment rules, determine whether the candidate traffic intervention plan has intervention backlash risk. If intervention backlash risk is determined, generate a parameterized adjusted candidate traffic intervention plan and feed it back. Output abnormal traffic event identification results and dynamic control suggestions based on the recalculation results. Based on the relationship between the intervention backlash index and the preset backlash judgment rules, determine whether the candidate traffic intervention plan has intervention backlash risk. If intervention backlash risk exists, generate a parameterized adjusted candidate traffic intervention plan and feed it back to the counterfactual simulation process. The counterfactual simulation module reconstructs the intervention counterfactual simulation state, and the coupled evaluation module recalculates the induced flow herding coefficient, the alternative path remaining capacity coefficient, the intervention backlash index, and the post-intervention risk level. If the recalculation results indicate that the parameterized adjusted candidate traffic intervention plan does not have intervention backlash risk, and the post-intervention risk level is lower than the initial risk level, output abnormal traffic event identification results and dynamic control suggestions.

[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An abnormal traffic event identification system based on road network operation data fusion, characterized in that, include: The data fusion module is used to fuse multi-source road network operation data of the target road network and generate the spatiotemporal state matrix of the target road network. The event recognition module is used to identify abnormal traffic events based on the road network spatiotemporal state matrix, generate an abnormal event profile, and generate at least one candidate traffic intervention plan based on the abnormal event profile. The counterfactual simulation module is used to construct, under the same counterfactual simulation baseline, a non-intervention counterfactual simulation state and an intervention counterfactual simulation state within a preset future time window for each candidate traffic intervention plan, and generate a candidate intervention response matrix based on the intervention counterfactual simulation state; The coupled evaluation module is used to perform coupled evaluation on candidate traffic intervention schemes based on the road network spatiotemporal state matrix and candidate intervention response matrix, to obtain the induced flow herding coefficient and alternative path remaining capacity coefficient corresponding to the candidate traffic intervention scheme, and further determine the intervention backlash index corresponding to the candidate traffic intervention scheme. The closed-loop control module is used to determine whether there is a risk of intervention backlash in candidate traffic intervention schemes based on whether the intervention backlash index meets the preset backlash judgment rules. When it is determined that there is a risk of intervention backlash in candidate traffic intervention schemes, it generates a parameterized control candidate traffic intervention scheme, feeds the parameterized control candidate traffic intervention scheme back to the counterfactual simulation module, and outputs abnormal traffic event identification results and dynamic control suggestions based on the recalculation results of the counterfactual simulation module and the coupled evaluation module.

2. The abnormal traffic event identification system based on road network operation data fusion according to claim 1, characterized in that, The road network spatiotemporal state matrix includes road segment operation status indicators, node queuing spillover indicators, path remaining capacity indicators, and road network topology indicators.

3. The abnormal traffic event identification system based on road network operation data fusion according to claim 2, characterized in that, The abnormal event profile is generated as follows: Spatiotemporal anomaly detection is performed on the spatiotemporal state matrix of the road network to determine the road segment where the anomaly occurred, the time of the anomaly, the road segment affected by the anomaly, and the direction of anomaly propagation. Based on the road segment operation status indicators, node queuing spillover indicators, and path remaining capacity indicators, the scope of the anomaly impact and the initial risk level are determined, and an abnormal event profile is generated.

4. The abnormal traffic event identification system based on road network operation data fusion according to claim 3, characterized in that, The candidate traffic intervention scheme includes at least one of the following: a set of target alternative paths, the target traffic flow objects, the guidance ratio, the guidance timing, the guidance priority corresponding to each target alternative path, and traffic control parameters; wherein, the guidance ratio includes the proportion of guided vehicles, the guidance timing includes the guidance release time and the guidance release batch, and the traffic control parameters include at least one of the following: path restriction parameters, speed limit parameters, and ramp control parameters.

5. The abnormal traffic event identification system based on road network operation data fusion according to claim 1, characterized in that, The candidate intervention response matrix is ​​generated by extracting the predicted transfer scale, predicted transfer timing, transfer path distribution, and node turning increment of the induced traffic flow object relative to the target alternative path set within a preset future time window from the intervention counterfactual simulation state, and generating the candidate intervention response matrix according to the target alternative path set.

6. The abnormal traffic event identification system based on road network operation data fusion according to claim 5, characterized in that, The induced flow herding coefficient is obtained as follows: Based on the candidate intervention response matrix, the path and time induced response distribution of the induced traffic flow object on the target alternative path set is constructed. Based on the predicted transfer scale, predicted transfer time sequence, transfer path distribution and node turning increment, the scale concentration sub-coefficient, time sequence synchronization sub-coefficient, path clustering sub-coefficient and turning impact sub-coefficient are determined respectively. Then, the spatial correlation weight between each target alternative path is determined by combining the road network topology index. The scale concentration sub-coefficient, time sequence synchronization sub-coefficient, path clustering sub-coefficient and turning impact sub-coefficient are coupled and fused to obtain the induced flow herding coefficient.

7. The abnormal traffic event identification system based on road network operation data fusion according to claim 6, characterized in that, The method for obtaining the remaining capacity coefficient of alternative paths is as follows: based on the remaining carrying capacity index of the path in the spatiotemporal state matrix of the road network, determine the basic carrying capacity margin of the target alternative path set within a preset future time window, and based on the predicted transfer scale, the distribution of transfer paths and the node turning increment in the candidate intervention response matrix, determine the induced occupancy increment of the target alternative path set. Based on the basic carrying capacity margin and the induced occupancy increment, the distribution of remaining storage capacity along the path and over time is constructed. Combined with the bottleneck constraints determined by the road segment operation status indicators, the node spillover impact determined by the node queuing spillover indicators, and the sustainability of the remaining storage capacity, the remaining storage capacity coefficient of the alternative path is obtained.

8. The abnormal traffic event identification system based on road network operation data fusion according to claim 7, characterized in that, The intervention backlash index is determined by comparing the uninterrupted counterfactual simulation state and the intervened counterfactual simulation state to obtain the risk reduction within the scope of abnormal influence determined by the abnormal event profile, the risk increase of the target alternative path set, and the risk increase of the spillover of adjacent nodes associated with the target alternative path set determined by the road network topology index. The intervention backlash index is obtained by coupling the induced flow herding coefficient and the alternative path remaining capacity coefficient.

9. The abnormal traffic event identification system based on road network operation data fusion according to claim 8, characterized in that, The closed-loop control module is used to feed back the parameterized control candidate traffic intervention schemes to the counterfactual simulation module, so that the counterfactual simulation module can reconstruct the intervention counterfactual simulation state and enable the coupled evaluation module to recalculate the induced flow herding coefficient, the alternative path remaining capacity coefficient, the intervention backlash index, and the post-intervention risk level. Among them, the post-intervention risk level is the risk level corresponding to the reconstructed intervention counterfactual simulation state, and the initial risk level is the risk level determined by the abnormal event profile when the abnormal traffic event is identified. When the recalculation results indicate that the candidate traffic intervention scheme after parameterized control does not have the risk of intervention backlash, and the risk level after intervention is lower than the initial risk level, the closed-loop control module outputs the abnormal traffic event identification results and dynamic control suggestions.

10. A method for identifying abnormal traffic events based on road network operation data fusion, applied to the abnormal traffic event identification system based on road network operation data fusion as described in any one of claims 1-9, characterized in that, include: Step 1: The multi-source road network operation data of the target road network are fused and processed to generate the spatiotemporal state matrix of the target road network; Step 2: Identify abnormal traffic events based on the road network spatiotemporal state matrix and generate abnormal event profiles; Step 3: Generate at least one candidate traffic intervention plan based on the abnormal event profile; Step 4: For each candidate traffic intervention plan, under the same counterfactual simulation baseline, construct the non-intervention counterfactual simulation state and the intervention counterfactual simulation state within a preset future time window; Step 5: Generate a candidate intervention response matrix based on the counterfactual simulation state of the intervention; Step 6: Based on the road network spatiotemporal state matrix and the candidate intervention response matrix, perform a coupled evaluation of the candidate traffic intervention schemes to obtain the induced flow herding coefficient, the alternative path remaining capacity coefficient, and the intervention backlash index. Step 7: Determine whether the candidate traffic intervention plan has the risk of intervention backlash based on whether the intervention backlash index meets the preset backlash judgment rules. If the intervention backlash risk is determined, generate the parameterized and adjusted candidate traffic intervention plan and provide feedback. Output the abnormal traffic event identification results and dynamic control suggestions based on the recalculation results.

Images