A dynamic construction method for highway digital twin

Abstract

The present invention discloses a method for dynamically constructing a digital twin of a highway, which relates to the field of traffic technology, obtains multi-dimensional data of a highway system and performs preprocessing, builds a traffic flow model of the highway based on deep learning technology and in combination with the multi-dimensional data of the highway system, maps the multi-dimensional data of the highway system collected in real time to the traffic flow model, dynamically updates the traffic status in the traffic flow model, uses the traffic flow model to perform short-term traffic flow prediction, simulates the impact of different traffic events, and analyzes the existing risks. The present invention captures and analyzes traffic flow data in real time, uses machine learning and data mining technology, extracts valuable traffic information from massive data, enables the digital twin to accurately predict future traffic flow, provides decision-making basis, helps reduce traffic congestion and accidents, and improves the traffic capacity and safety of highways.

Description

Technical Field

[0001] The present invention relates to the field of transportation technology, and in particular to a method for dynamically constructing a digital twin of a highway. Background Art

[0002] With the continuous development of smart cities and smart transportation, highways, as the core part of urban and regional transportation networks, require more intelligent and digital management and services. Digital twin technology, as one of the key supporting technologies for smart transportation, realizes comprehensive perception, accurate analysis and intelligent decision-making of the transportation system by constructing virtual mapping of physical entities. In the field of highways, the application of digital twin technology can promote the intelligent, convenient and safe operation of highways.

[0003] Since traffic flow in the highway system is highly variable, the road network scheduling process is difficult to respond quickly to sudden traffic accidents and weather changes, resulting in delayed flow regulation and affecting traffic smoothness. Therefore, how to construct a highway digital twin to achieve dynamic prediction of traffic flow and improve the accuracy of dynamic road network scheduling is the problem we need to solve. To this end, a dynamic construction method for highway digital twins is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a method for dynamically constructing a digital twin of a highway to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A method for dynamically constructing a highway digital twin includes the following steps:

[0007] Step 1: Obtain multi-dimensional data of the highway system and perform preprocessing;

[0008] Step 2: Based on deep learning technology, a highway traffic flow model is constructed in combination with multi-dimensional data of the highway system to reflect changes in traffic flow;

[0009] Step 3: Map the multi-dimensional data of the highway system collected in real time to the traffic flow model, and dynamically update the traffic status in the traffic flow model;

[0010] Step 4: Use the traffic flow model to make short-term traffic flow forecasts, simulate the impact of different traffic events, and analyze the risks involved;

[0011] Step 5: According to the predicted traffic flow and risk analysis results, formulate a dynamic road network scheduling strategy and conduct dynamic road network scheduling and optimization;

[0012] Step 6: Integrate the traffic flow model with the highway management system and display the real-time status and prediction results of traffic flow through visualization technology.

[0013] A further improvement of the technical solution of the present invention is that the multi-dimensional data of the highway system specifically includes: traffic flow, vehicle speed, traffic events, meteorological information, road condition information and infrastructure status, wherein traffic events include traffic accidents, road construction and congestion, meteorological information includes weather conditions, temperature, humidity and wind speed, road condition information includes road surface slipperiness, potholes and icing conditions, and infrastructure status includes tunnel equipment operation status.

[0014] A further improvement of the technical solution of the present invention is that in step 1, the process of acquiring the multi-dimensional data of the highway system includes:

[0015] Collect multi-dimensional data of the highway system through data collection equipment deployed along the highway, at toll booths, key sections and nodes. The data collection equipment includes sensors, cameras, GPS equipment, weather stations and infrastructure detection equipment;

[0016] The data collection equipment collects data at a set frequency and transmits it to the data center via wireless communication;

[0017] The collected multi-dimensional data is preprocessed in the data center, the received raw data is cleaned, invalid, duplicate or erroneous data is removed, and the data is formatted and standardized for subsequent data analysis and processing. The data from different sensors and monitoring equipment are integrated to form a complete data set, and then the data is associated and matched, and the vehicle speed data is associated with the vehicle information to form a comprehensive traffic flow data sequence.

[0018] A further improvement of the technical solution of the present invention is that in step 2, the process of constructing the traffic flow model of the expressway includes:

[0019] Extract the pre-processed multi-dimensional data of the highway system, ensure data quality, remove noise and outliers, and divide the pre-processed multi-dimensional data into training set, validation set and test set in chronological order;

[0020] The traffic flow model is constructed by integrating graph convolutional network (GCN) and gated recurrent unit (GRU), namely T-GCN model, which captures the topological structure of the road network and the dynamic changes of traffic data at the same time. The integrated graph convolutional network part is used to capture the topological structure of the road network, that is, the connection relationship between roads, and the gated recurrent unit part is used to capture the dynamic changes of traffic data, that is, the time dependency in time series data. The deep learning model is trained using the training set data, and the model parameters are optimized through the back propagation algorithm to minimize the prediction error. The model in the training process is evaluated using the validation set data, the model structure and hyperparameters are adjusted, and the training and adjustment process is repeated until the model achieves satisfactory performance on the validation set.

[0021] After the T-GCN model training is completed, the trained T-GCN model is evaluated using the test set data, the prediction error and accuracy indicators are calculated, and the prediction performance of the model is analyzed, including the accuracy and stability of the prediction. The trained T-GCN model is then deployed to the highway traffic flow prediction system.

[0022] A further improvement of the technical solution of the present invention is that in step 3, the process of dynamically updating the traffic status in the traffic flow model includes:

[0023] The multi-dimensional data of the highway system collected in real time is converted into a form suitable for the input of the T-GCN model, and then the real-time collected road network connection relationship is mapped to the adjacency matrix of the graph convolutional network part, and the node features (road section flow, vehicle speed) are updated to reflect the current traffic status;

[0024] The historical highway system multi-dimensional data and the latest real-time data are combined into a time series as the input of the gated recursive unit part. The time window is set to the data of the most recent hour. Then the spatial features (extracted from the GCN part) and the temporal features (extracted from the GRU part) are fused to form a complete input feature vector.

[0025] The adapted input data is sent to the trained T-GCN model for inference, and the predicted results including traffic flow and vehicle speed in the short future period are output. Then, according to the results output by the model, the traffic status in the T-GCN model is dynamically updated. The updated content includes section flow, vehicle speed distribution and status marking. The section flow is to adjust the expected flow value of each section, update the average vehicle speed and speed distribution of each section, and preset the thresholds of traffic flow and vehicle speed to mark congestion, smoothness or slowness.

[0026] A further improvement of the technical solution of the present invention is that the output process of the traffic flow prediction result is:

[0027] Extract real-time traffic flow data for each section along the highway , ensure the consistency of data timestamps and handle missing values ​​or outliers;

[0028] Use historical traffic data to train the T-GCN model, adjust the weights to reflect the impact of different road sections on the overall traffic flow, and optimize the loss function during the training process to derive the traffic adjustment weights and bias ;

[0029] Real-time traffic flow for each road segment Perform weighted processing and sum the weighted traffic flow values ​​of all road sections, calculate the square sum, include the data of all road sections in the calculation range, and then take the square root of the square sum result to obtain the final traffic flow prediction value ;

[0030] The output process of the vehicle speed prediction result is as follows:

[0031] Use GPS devices, radar speed guns or video analysis technology to collect real-time vehicle speed data on each road section , while extracting real-time traffic flow data , make sure the timestamps of both are consistent;

[0032] Determine the vehicle speed adjustment weight through the T-GCN model , Speed ​​adjustment bias , Flow adjustment coefficient and flow bias , where the weight Reflects the influence of vehicle speed on the overall average speed, bias Correction model error, flow adjustment factor Indicates the reverse effect of traffic flow on vehicle speed, flow bias Balance the impact of traffic flow on vehicle speed;

[0033] Real-time speed for each road segment Perform weighted processing and sum the weighted vehicle speed values ​​of all road sections;

[0034] Weighted processing of real-time traffic flow on each road segment , and sum the weighted traffic flow values ​​of all road sections and take the square root;

[0035] The final speed prediction value is obtained by dividing the sum of the weighted speed values ​​by the square root of the weighted traffic flow. .

[0036] A further improvement of the technical solution of the present invention is that in step 4, the risk analysis process includes:

[0037] Load the trained traffic flow model (T-GCN model) and set the time range for short-term prediction, specifically the next hour;

[0038] Extract the real-time collected traffic data (including road traffic, vehicle speed, weather conditions, etc.) as model input, and output the predicted results of traffic flow and vehicle speed in the short future time through traffic flow model inference;

[0039] Define the definitions and characteristics of traffic accidents, road construction and severe weather events, and adjust the relevant parameters in the input data according to the event type. For traffic accidents, reduce the capacity of the accident section and increase the queue length. For road construction, close some lanes and reduce road capacity. For severe weather, reduce the speed limit and increase the distance between vehicles. Then, feed the adjusted input data into the traffic flow model, re-infer, and output the traffic flow prediction results under the influence of the event.

[0040] Compare the prediction results before and after the event, determine the focus indicators and analyze the specific impact of the event on traffic flow. The focus indicators include: the decline in traffic flow, the degree of speed reduction, and the scope of congestion diffusion;

[0041] Based on the simulation results, identify the road sections or areas that are susceptible to the incident, calculate the sensitivity index, analyze the sensitivity of the analyzed road sections to traffic incidents, and calculate the comprehensive risk index of the analyzed area in combination with the sensitivity index to measure the overall risk level of all road sections in the area;

[0042] According to the severity of the impact of the incident, risks are divided into three levels: low, medium and high, and the classification criteria for each risk level are determined.

[0043] A further improvement of the technical solution of the present invention is that the classification criteria of the risk levels are specifically:

[0044] The low-risk classification criteria are that the traffic flow decreases by less than 10%, the vehicle speed decreases by less than 10%, and the congestion spread is small;

[0045] The standard for dividing into medium risk is that the traffic volume decreases by 10%-30%, the vehicle speed decreases by 10%-30%, and the congestion spread is medium;

[0046] The criteria for high risk are a decrease in traffic flow greater than 30%, a decrease in vehicle speed greater than 30%, and a large spread of congestion.

[0047] A further improvement of the technical solution of the present invention is that in step 5, the process of formulating the road network dynamic scheduling strategy includes:

[0048] Use the trained traffic flow model to perform short-term traffic flow forecasts to obtain traffic flow and risk analysis results, and identify high-risk sections or areas;

[0049] Clarify the objectives of road network dispatching, determine the priority dispatching sections according to the risk level and importance of the sections, and give priority to high-risk sections to ensure that their impact is minimized. The dispatching objectives include relieving congestion, improving traffic efficiency and ensuring safety. Relieving congestion means reducing the pressure on high-traffic sections, improving traffic efficiency means optimizing vehicle speed distribution and reducing vehicle waiting time, and ensuring safety means avoiding potential safety hazards caused by traffic incidents;

[0050] Based on the predicted traffic flow and risk analysis results, the traffic information release system guides vehicles to choose reasonable routes, and releases traffic warning information in advance in high-risk areas or sections to remind drivers to pay attention to changes in road conditions. When traffic accidents or road construction occur, emergency plans are activated, driving routes are replanned, and congested sections are avoided.

[0051] A further improvement of the technical solution of the present invention is that in step 6, the process of visual display includes:

[0052] Select visualization tools to display the real-time status and prediction results of traffic flow, display the road network structure of the highway on the GIS map, and mark the status of each section with different colors, and then dynamically update the flow and speed information on the map to reflect the real-time traffic status. Among them, green means smooth, yellow means slow, and red means congestion;

[0053] Use charts to show the changing trends of traffic flow and vehicle speed in the short term, provide forecast results by time period, and display warning information in high-risk sections or areas to remind users to pay attention to changes in road conditions.

[0054] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0055] 1. The present invention provides a method for dynamically constructing a digital twin of a highway. By capturing and analyzing traffic flow data in real time and using machine learning and data mining techniques, valuable traffic information is extracted from massive data, so that the digital twin can accurately predict future traffic flow and provide a basis for decision-making, which helps to reduce traffic congestion and accidents and improve the capacity and safety of highways.

[0056] 2. The present invention provides a method for dynamically constructing a digital twin of a highway. By integrating multi-dimensional data collected in real time and traffic flow models, the accuracy of traffic flow prediction can be significantly improved. It not only takes into account the current traffic status, but also combines historical data and road network topology, thereby realizing accurate prediction of traffic flow and vehicle speed change trends in the short term in the future. It not only improves the efficiency of traffic management, but also reduces large-scale congestion and safety hazards caused by emergencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0058] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0059] Figure 2 A schematic diagram of a workflow for dynamically updating traffic status in a traffic flow model of the present invention;

[0060] Figure 3 Schematic diagram of the workflow of risk analysis of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0062] Embodiment 1, as Figure 1 , Figure 2 As shown, the present invention provides a method for dynamically constructing a digital twin of a highway, comprising the following steps:

[0063] Step 1: Obtain and preprocess the multi-dimensional data of the highway system. The multi-dimensional data of the highway system specifically include: traffic flow, vehicle speed, traffic events, meteorological information, road condition information and infrastructure status, among which traffic events include traffic accidents, road construction and congestion, meteorological information includes weather conditions, temperature, humidity and wind speed, road condition information includes road slipperiness, potholes and icing, and infrastructure status includes tunnel equipment operation status. The multi-dimensional data of the highway system is collected through data collection equipment deployed along the highway, toll stations, key sections and nodes. The data collection equipment includes sensors, cameras, GPS equipment, meteorological stations and infrastructure detection equipment. Geomagnetic sensors, microwave detectors and video cameras are used to collect data, and GPS positioning technology is combined to obtain the real-time position and speed of the vehicle. Traffic events are identified based on video analysis technology. Failures or abnormal behaviors, combining manual reporting and intelligent algorithms to detect construction areas or congestion points, deploying meteorological stations, real-time monitoring of temperature, humidity, wind speed, rainfall, etc., using satellite remote sensing technology to obtain large-scale meteorological data, using LiDAR to scan the road surface, detect potholes, cracks, etc., and obtain high-resolution images through drone inspections. Data acquisition equipment collects data at a set frequency and transmits it to the data center through wireless communication. The collected multi-dimensional data is pre-processed in the data center, the received raw data is cleaned, invalid, duplicated or erroneous data is removed, and the data is formatted and standardized for subsequent data analysis and processing. The data from different sensors and monitoring equipment are integrated to form a complete data set, and then the data is associated and matched, and the speed data is associated with the vehicle information to form a comprehensive traffic flow data sequence;

[0064] Step 2: Based on deep learning technology, combined with the multi-dimensional data of the highway system, a traffic flow model of the highway is constructed to reflect the changes in traffic flow. The pre-processed multi-dimensional data of the highway system is extracted to ensure data quality, remove noise and outliers, and divide the pre-processed multi-dimensional data into training set, validation set and test set in chronological order. The training set is used for the initial training of the model, the validation set is used to adjust the model parameters and structure during the training process to avoid overfitting, and the test set is used to finally evaluate the performance of the model. The graph convolutional network (GCN) and the gated recurrent unit (GRU) are integrated to construct a traffic flow model, namely the T-GCN model, which captures the topological structure of the road network and the dynamic changes of traffic data at the same time. Among them, the integrated graph convolutional network part is used to capture the topological structure of the road network, that is, the connection relationship between roads, and the gated recurrent unit (GRU) is used to capture the topological structure of the road network. The return unit part is used to capture the dynamic changes of traffic data, that is, the time dependency in time series data. The deep learning model is trained using the training set data, and the model parameters are optimized through the back propagation algorithm to minimize the prediction error. The model in the training process is evaluated using the validation set data, and the model structure and hyperparameters are adjusted. The training and adjustment process is repeated until the model achieves satisfactory performance on the validation set. After the T-GCN model training is completed, the trained T-GCN model is evaluated using the test set data, and the prediction error and accuracy indicators are calculated. The prediction performance of the model is analyzed, including the accuracy and stability of the prediction. The trained T-GCN model is then deployed to the highway traffic flow prediction system to ensure that the model can receive and process multi-dimensional data from the highway system in real time to achieve real-time traffic flow prediction.

[0065] Step 3: Map the multi-dimensional data of the highway system collected in real time to the traffic flow model, dynamically update the traffic status in the traffic flow model, convert the multi-dimensional data of the highway system collected in real time into a form suitable for the input of the T-GCN model, and then map the road network connection relationship collected in real time to the adjacency matrix of the graph convolutional network part, and update the node features (road section flow, vehicle speed) to reflect the current traffic status. For the construction of the adjacency matrix, the connection relationship between sections is defined according to the topological structure of the highway. For example, if section A is connected to section B, the corresponding position in the adjacency matrix is ​​marked as 1, otherwise it is marked as 0. The historical highway system multi-dimensional data and the latest real-time data are combined. The time series are combined as the input of the gated recursive unit part, and the time window is set to the data of the most recent hour. Then the spatial features (extracted from the GCN part) and the temporal features (extracted from the GRU part) are fused to form a complete input feature vector. The adapted input data is sent to the trained T-GCN model for inference, and the predicted results of traffic flow and vehicle speed in the short future period are output. Then, according to the results of the model output, the traffic status in the T-GCN model is dynamically updated. The updated content includes the section flow, vehicle speed distribution and status mark. The section flow is to adjust the expected flow value of each section, update the average vehicle speed and speed distribution of each section, and preset the traffic flow threshold. Speed ​​threshold , to mark the congested, smooth or slow status;

[0066] In addition, the output process of traffic flow prediction results is:

[0067] Extract real-time traffic flow data for each section along the highway , ensure the consistency of the data timestamps, handle missing values ​​or outliers, use historical traffic data to train the T-GCN model, adjust the weights to reflect the impact of different road sections on the overall traffic flow, and optimize the loss function during the training process to obtain the traffic adjustment weights and bias , real-time traffic flow for each road segment Perform weighted processing and sum the weighted traffic flow values ​​of all road sections, calculate the square sum, include the data of all road sections in the calculation range, and then take the square root of the square sum result to obtain the final traffic flow prediction value ;

[0068] The calculation formula for traffic flow prediction value is as follows:

[0069] ;

[0070] In the formula, For the short future Traffic flow prediction value within is the number of road sections, For the The traffic flow adjustment weight of each road section reflects its impact on the overall traffic flow. For the Road section in time Real-time traffic flow data, For the The bias term of the road segment is used to correct the model error. As the traffic flow increases, It is on an upward trend;

[0071] The output process of vehicle speed prediction results is as follows:

[0072] Use GPS devices, radar speed guns or video analysis technology to collect real-time vehicle speed data on each road section , while extracting real-time traffic flow data , ensure that the timestamps of the two are consistent, and determine the speed adjustment weight through the T-GCN model , Speed ​​adjustment bias , Flow adjustment coefficient and flow bias , where the weight Reflects the influence of vehicle speed on the overall average speed, bias Correction model error, flow adjustment factor Indicates the reverse effect of traffic flow on vehicle speed, flow bias Balance the impact of traffic flow on vehicle speed, and calculate the real-time vehicle speed for each road section Perform weighted processing and sum the weighted speed values ​​of all road sections to weight the real-time traffic flow of each road section The weighted traffic flow values ​​of all road sections are summed up, the square root is taken, and the sum of the weighted speed values ​​is divided by the square root of the weighted traffic flow to obtain the final speed prediction value. ;

[0073] The calculation formula of the vehicle speed prediction value is as follows:

[0074] ;

[0075] In the formula, For the short future The average speed prediction value within For the The speed adjustment weight of the road section reflects its impact on the speed. For the Road section in time Real-time vehicle speed data, For the The speed adjustment bias of the road section is used to correct the model error. For the The flow adjustment coefficient of the road section reflects the reverse effect of flow on vehicle speed. For the Road section in time Real-time traffic data, For the The traffic flow bias of each road section is used to balance the impact of traffic flow on vehicle speed. When the traffic flow is low, the vehicle speed is high, and when the traffic flow is high, the vehicle speed is reduced. Combined with the preset traffic flow threshold Speed ​​threshold , congestion: and , smooth: and , Slow: between congestion and smooth flow;

[0076] Step 4: Use the traffic flow model to make short-term traffic flow forecasts, simulate the impact of different traffic events, and analyze the risks involved;

[0077] Step 5: According to the predicted traffic flow and risk analysis results, formulate a dynamic road network scheduling strategy and conduct dynamic road network scheduling and optimization;

[0078] Step 6: Integrate the traffic flow model with the highway management system and display the real-time status and prediction results of traffic flow through visualization technology.

[0079] Embodiment 2, as Figure 3 As shown, based on Example 1, the present invention provides a technical solution: Preferably, in step 4, the risk analysis process includes:

[0080] Load the trained traffic flow model (T-GCN model) and set the time range for short-term prediction, specifically the next hour. Extract the real-time collected traffic data (including road flow, vehicle speed, weather conditions, etc.) as model input, and use the traffic flow model to infer the predicted results of traffic flow and vehicle speed in the short term in the future. Clarify the definitions and characteristics of traffic accidents, road construction and severe weather traffic events, and adjust the relevant parameters in the input data according to the type of event. For traffic accidents, reduce the capacity of the accident section and increase the queue length. For road construction, close some lanes and reduce road capacity. For severe weather, reduce the speed limit and increase the distance between vehicles. Then send the adjusted input data to the traffic flow model and re-infer to output the traffic flow prediction results under the influence of the event. Compare the prediction results before and after the event, determine the focus indicators and analyze the specific impact of the event on traffic flow. Impact, focus on indicators including: traffic flow decline, speed reduction and congestion diffusion range. According to the simulation results, identify the road sections or areas susceptible to the incident, calculate the sensitivity index, analyze the sensitivity of the analyzed road sections to traffic incidents, and calculate the risk comprehensive index of the analyzed area in combination with the sensitivity index to measure the overall risk level of all road sections in the area. According to the severity of the impact of the incident, the risk is divided into three levels: low, medium and high, and the classification standards for each risk level are determined. The classification standard for low risk is that the traffic flow decline is less than 10%, the speed reduction is less than 10%, and the congestion diffusion range is small. The classification standard for medium risk is that the traffic flow decline is between 10%-30%, the speed reduction is between 10%-30%, and the congestion diffusion range is medium. The classification standard for high risk is that the traffic flow decline is greater than 30%, the speed reduction is greater than 30%, and the congestion diffusion range is large.

[0081] The calculation formula of the sensitivity index is as follows:

[0082] ;

[0083] In the formula, For the The sensitivity index of a road section is used to measure the sensitivity of the road section to traffic events. is the number of prediction steps in the time window, For the The road segment at time step The decrease in flow rate, For the The road segment at time step The speed reduction of For the The road segment at time step The congestion diffusion weight reflects its impact on the downstream road section. For the The basic importance weight of a road section indicates the criticality of the road section in the entire road network. is the adjustment coefficient of the basic importance weight, controlling The impact on the sensitivity index, as the flow rate decreases and speed reduction increase, It is on an upward trend;

[0084] The calculation formula of the comprehensive risk index is as follows:

[0085] ;

[0086] In the formula, For Region The comprehensive risk index is used to measure the overall risk level of all road sections in the area. For the The sensitivity index of the road section, For the The weight of each road section reflects its relative importance in the region. For the The length of the road segment, For the The traffic volume of the road section, For Region The number of road sections within is the adjustment coefficient of the number of road sections, controlling The impact on the risk index is as follows: increase, It is on an upward trend;

[0087] In step 5, the process of formulating the road network dynamic scheduling strategy includes:

[0088] Using the trained traffic flow model to predict short-term traffic flow, obtain the traffic flow and risk analysis results, identify high-risk sections or areas, clarify the goals of road network scheduling, and determine the priority sections according to the risk level and the importance of the sections. High-risk sections are given priority to ensure that their impact is minimized. Among them, the scheduling goals include relieving congestion, improving traffic efficiency and ensuring safety. Relieving congestion means reducing the pressure on high-flow sections, improving traffic efficiency means optimizing the speed distribution and reducing vehicle waiting time, and ensuring safety means avoiding safety hazards caused by traffic incidents. According to the predicted traffic flow and risk analysis results, the traffic information release system guides vehicles to choose routes reasonably, and releases traffic warning information in advance in high-risk areas or sections to remind drivers to pay attention to changes in road conditions. When traffic accidents or road construction occur, the emergency plan is activated to re-plan the driving route to avoid congested sections.

[0089] In step 6, the visualization process includes:

[0090] Select visualization tools to display the real-time status and prediction results of traffic flow, display the road network structure of the highway on the GIS map, and mark the status of each section with different colors, and then dynamically update the traffic flow and speed information on the map to reflect the real-time traffic status. Green indicates smooth traffic, yellow indicates slow traffic, and red indicates congestion. Use charts to display the traffic flow and speed change trends in the short term in the future, provide prediction results for different time periods, and display warning information in high-risk sections or areas to remind users to pay attention to changes in road conditions.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for dynamically constructing a highway digital twin, characterized in that: The following steps are involved: Step 1: Obtain multi-dimensional data of the highway system and perform preprocessing; Step 2: Based on deep learning technology, the traffic flow model of the highway is constructed by combining the multi-dimensional data of the highway system. The traffic flow model is constructed by integrating the graph convolutional network and the gated recursive unit, which is the T-GCN model. Step 3: Map the multi-dimensional data of the highway system collected in real time to the traffic flow model, dynamically update the traffic status in the traffic flow model, convert the multi-dimensional data of the highway system collected in real time into a form suitable for the input of the T-GCN model, and then map the real-time collected road network connection relationship to the adjacency matrix of the graph convolutional network part, and update the node features to reflect the current traffic status; The multi-dimensional data of the historical highway system and the latest real-time data are combined into a time series as the input of the gating recursive unit part. The time window is set to the data of the most recent hour, and then the spatial features and time features are fused to form a complete input feature vector. The adapted input data is sent to the trained T-GCN model for inference, and the predicted results of traffic flow and vehicle speed in the short future period are output. Then, the traffic status in the T-GCN model is dynamically updated according to the results output by the model. The updated content includes the section flow, vehicle speed distribution and status mark. The section flow is to adjust the expected flow value of each section, update the average vehicle speed and speed distribution of each section, and preset the thresholds of traffic flow and vehicle speed to mark the congested, unblocked or slow status; Step 4: Use the traffic flow model to make short-term traffic flow forecasts, simulate the impact of different traffic events, and analyze the risks involved; Step 5: According to the predicted traffic flow and risk analysis results, formulate a dynamic road network scheduling strategy and conduct dynamic road network scheduling and optimization; Step 6: Integrate the traffic flow model with the highway management system and display the real-time status and prediction results of traffic flow through visualization technology.

2. A method for dynamically constructing a highway digital twin according to claim 1, characterized in that: The multi-dimensional data of the highway system specifically include: traffic flow, vehicle speed, traffic events, meteorological information, road condition information and infrastructure status, among which traffic events include traffic accidents, road construction and congestion, meteorological information includes weather conditions, temperature, humidity and wind speed, road condition information includes road surface slipperiness, potholes and icing conditions, and infrastructure status includes tunnel equipment operation status.

3. The method for dynamically constructing a highway digital twin according to claim 2 is characterized in that: In the step 1, the process of acquiring the multi-dimensional data of the highway system includes: Collect multi-dimensional data of the highway system through data collection equipment deployed along the highway, at toll booths, key sections and nodes. The data collection equipment includes sensors, cameras, GPS equipment, weather stations and infrastructure detection equipment; The data collection equipment collects data at a set frequency and transmits it to the data center via wireless communication; The collected multi-dimensional data is preprocessed in the data center, the received raw data is cleaned, formatted and standardized, and the data from different sensors and monitoring equipment are integrated to form a complete data set. The data is then associated and matched, and the vehicle speed data is associated with the vehicle information to form a comprehensive traffic flow data sequence.

4. A method for dynamically constructing a highway digital twin according to claim 3, characterized in that: In the step 2, the process of constructing the traffic flow model of the expressway includes: Extract the preprocessed multi-dimensional data of the highway system, and divide the preprocessed multi-dimensional data into a training set, a validation set, and a test set in chronological order; At the same time, it captures the dynamic changes of the topological structure of the road network and traffic data, uses the training set data to train the deep learning model, optimizes the model parameters through the back propagation algorithm, and uses the validation set data to evaluate the model during the training process and adjust the model structure and hyperparameters; After the T-GCN model training is completed, the trained T-GCN model is evaluated using the test set data, the prediction error and accuracy indicators are calculated, the prediction performance of the model is analyzed, and then the trained T-GCN model is deployed to the highway traffic flow prediction system.

5. The method for dynamically constructing a highway digital twin according to claim 1, characterized in that: The output process of the traffic flow prediction result is as follows: Extract real-time traffic flow data for each section along the highway , and handle missing values ​​or outliers; Use historical traffic data to train the T-GCN model, adjust the weights to reflect the impact of different road sections on the overall traffic flow, and optimize the loss function during the training process to derive the traffic adjustment weights and bias ; Real-time traffic flow for each road segment Perform weighted processing and sum the weighted traffic flow values ​​of all road sections, calculate the square sum, include the data of all road sections in the calculation range, and then take the square root of the square sum result to obtain the final traffic flow prediction value ; The output process of the vehicle speed prediction result is as follows: Use GPS devices, radar speed guns or video analysis technology to collect real-time vehicle speed data on each road section , while extracting real-time traffic flow data , make sure the timestamps of both are consistent; Determine the vehicle speed adjustment weight through the T-GCN model , Speed ​​adjustment bias , Flow adjustment coefficient and flow bias ; Real-time vehicle speed for each road segment Perform weighted processing and sum the weighted vehicle speed values ​​of all road sections; Weighted processing of real-time traffic flow on each road segment , and sum the weighted traffic flow values ​​of all road sections and take the square root; The final speed prediction value is obtained by dividing the sum of the weighted speed values ​​by the square root of the weighted traffic flow. .

6. A method for dynamically constructing a highway digital twin according to claim 5, characterized in that: In step 4, the risk analysis process includes: Load the trained traffic flow model and set the short-term prediction time range to the next hour; Extract the real-time collected traffic data as model input, and output the predicted results of traffic flow and vehicle speed in the short future time by traffic flow model reasoning; Define the definitions and characteristics of traffic accidents, road construction and severe weather events, and adjust the relevant parameters in the input data according to the event type. For traffic accidents, reduce the capacity of the accident section and increase the queue length. For road construction, close some lanes and reduce road capacity. For severe weather, reduce the speed limit and increase the distance between vehicles. Then, feed the adjusted input data into the traffic flow model, re-infer, and output the traffic flow prediction results under the influence of the event. Compare the prediction results before and after the event, determine the focus indicators and analyze the specific impact of the event on traffic flow. The focus indicators include: the decline in traffic flow, the degree of speed reduction, and the scope of congestion diffusion; Based on the simulation results, identify the road sections or areas that are susceptible to the incident, calculate the sensitivity index, analyze the sensitivity of the analyzed road sections to traffic incidents, and calculate the comprehensive risk index of the analyzed area in combination with the sensitivity index to measure the overall risk level of all road sections in the area; According to the severity of the impact of the incident, the risks are divided into three levels: low, medium and high, and the classification standards for each risk level are determined; The calculation formula of the sensitivity index is as follows: ； In the formula, For the The sensitivity index of the road section, is the number of prediction steps in the time window, For the The road segment at time step The decrease in flow rate, For the The road segment at time step The speed reduction of For the The road segment at time step The congestion diffusion weight, For the The basic importance weight of the road segment, is the adjustment coefficient of the basic importance weight; The calculation formula of the comprehensive risk index is as follows: ； In the formula, For Region The risk composite index, For the The sensitivity index of the road section, For the The weight of the road segment, For the The length of the road segment, For the The traffic volume of the road section, For Region The number of road sections within is the adjustment coefficient of the number of road sections.

7. A method for dynamically constructing a highway digital twin according to claim 6, characterized in that: The specific classification standards for each risk level are as follows: The low-risk classification criteria are that the traffic flow decreases by less than 10%, the vehicle speed decreases by less than 10%, and the congestion spread is small; The standard for dividing into medium risk is that the traffic volume decreases by 10%-30%, the vehicle speed decreases by 10%-30%, and the congestion spread is medium; The criteria for high risk are a decrease in traffic flow greater than 30%, a decrease in vehicle speed greater than 30%, and a large spread of congestion.

8. A method for dynamically constructing a highway digital twin according to claim 7, characterized in that: In step 5, the process of formulating the road network dynamic scheduling strategy includes: Use the trained traffic flow model to perform short-term traffic flow forecasts to obtain traffic flow and risk analysis results, and identify high-risk sections or areas; Clarify the objectives of road network dispatching, determine the priority dispatching sections according to the risk level and importance of the sections, and give priority to high-risk sections. The dispatching objectives include alleviating congestion, improving traffic efficiency and ensuring safety; Based on the predicted traffic flow and risk analysis results, the traffic information release system guides vehicles to choose reasonable routes, and releases traffic warning information in advance in high-risk areas or sections to remind drivers to pay attention to changes in road conditions. When traffic accidents or road construction occur, emergency plans are activated, driving routes are replanned, and congested sections are avoided.

9. A method for dynamically constructing a highway digital twin according to claim 8, characterized in that: In step 6, the process of visual display includes: Select visualization tools to display the real-time status and prediction results of traffic flow, display the road network structure of the highway on the GIS map, and mark the status of each section with different colors, and then dynamically update the flow and speed information on the map to reflect the real-time traffic status. Among them, green means smooth, yellow means slow, and red means congestion; Use charts to show the changing trends of traffic flow and vehicle speed in the short term, provide forecast results by time period, and display warning information in high-risk sections or areas to remind users to pay attention to changes in road conditions.

Images