Expressway new energy automobile travel characteristic analysis method and system

Abstract

The invention discloses an expressway new energy vehicle travel characteristic analysis method and system, and the method comprises the steps: collecting and preprocessing expressway data, obtaining images, traffic flow, vehicle speed, roads and facility point location information, carrying out the partitioning according to facility point locations, extracting the characteristics of acceleration waves and deceleration waves and vehicle type images, defining a dynamic region, and employing the vehicle type images of the dynamic region to carry out the analysis of the travel characteristics of new energy vehicles. The method comprises the following steps: acquiring short-time motion characteristics, identifying a first vehicle model and a second vehicle model, calculating a charging demand index, fitting motion data of the short-time motion characteristics of the two vehicle models in combination with the charging demand index and the traffic flow, calculating a conflict index, acquiring traffic flow change information based on the data by utilizing a cellular transmission model, and analyzing by virtue of a graph neural network. And inputting the motion data, the conflict index, the traffic flow change data and the high-speed topological structure, and outputting the travel characteristics of the first vehicle type. According to the method, travel feature analysis of the new energy vehicle in a high-speed scene is realized through quantification of the charging demand index, and auxiliary decision support can be provided for high-speed construction of charging facilities.

Description

Technical Field

[0001] The present invention relates to the field of analysis of automobile travel characteristics, and particularly to a method and system for analyzing the travel characteristics of new energy vehicles on expressways. Background Art

[0002] As the proportion of new energy vehicles in expressway travel gradually increases, traditional expressway traffic characteristic analysis methods are difficult to meet the accurate analysis requirements for the travel characteristics of new energy vehicles, such as charging needs and special driving behaviors. In existing expressway traffic analysis, a fixed partition method is mostly used to divide the sections where vehicles enter and exit facilities, and it is unable to respond in real time to the fluctuation characteristics such as acceleration waves and deceleration waves in the traffic flow, resulting in a mismatch between the partition range and the actual traffic impact area, and reducing the accuracy of traffic characteristic analysis.

[0003] More and more models with different power combinations appear in the relatively closed environment of expressways, which has a greater impact on the movement characteristics of vehicles in heterogeneous traffic flows and the traffic flow conditions of different sections. These are all problems that need to be solved urgently at present. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for analyzing the travel characteristics of new energy vehicles on expressways.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions: In the first aspect of the present invention, a method for analyzing the travel characteristics of new energy vehicles on expressways is provided, including: Collecting the expressway data to be processed for preprocessing to obtain image data, traffic volume, vehicle speed, road data, and facility points; Dividing the sections where vehicles enter and exit facilities based on the facility points, obtaining the acceleration wave characteristics, deceleration wave characteristics, and vehicle type images of the partitions according to the image data, traffic volume, and vehicle speed of the partitions, and correcting the partition range using the acceleration wave characteristics and deceleration wave characteristics to obtain dynamic areas; Obtaining short-term movement characteristics using the vehicle type images of the dynamic areas, obtaining the first vehicle type in the intervals before and after the charging facilities according to the vehicle type images, taking the remaining vehicle types as the second vehicle type, and calculating the charging demand index according to the first vehicle type that enters and does not enter the charging facilities; Fitting the short-term movement characteristics of the first vehicle type and the second vehicle type according to the charging demand index and traffic volume of the dynamic areas to obtain the movement data of the first vehicle type and the movement data of the second vehicle type, calculating the conflict index of the first vehicle type and the conflict index of the second vehicle type using the movement data of the first vehicle type, the movement data of the second vehicle type, and the partition road data, and obtaining the traffic flow change data through the cell transmission model based on the conflict index of the first vehicle type, the conflict index of the second vehicle type, the movement data of the first vehicle type, the movement data of the second vehicle type, and the partition road data; The travel characteristics of the first vehicle on the highway to be processed are obtained through graph neural network analysis based on the motion data of the first vehicle type, the conflict index of the first vehicle type, the traffic flow change data, and the highway topology structure.

[0006] As a further method, the method for collecting the data of the highway to be processed and performing preprocessing to obtain image data, traffic flow, vehicle speed, and facility points includes: Collect the image data, traffic flow, and vehicle speed data of the highway to be processed, obtain the original coordinates of charging facilities, fueling facilities, and service area facilities and the highway topology structure, perform filtering and image enhancement on the image data in sequence, use the interpolation method to repair abnormal data for the traffic flow and vehicle speed data, and match and correct the original coordinates of the facility points with the high-precision map to obtain the preprocessed image data, traffic flow, vehicle speed, road data, and facility points.

[0007] As a further method, the method for dividing the sections where vehicles enter and exit the facilities based on the facility points, and obtaining the acceleration wave characteristics, deceleration wave characteristics, and vehicle type images of the divided areas according to the image data, traffic flow, and vehicle speed of the divided areas, and using the acceleration wave characteristics and deceleration wave characteristics to correct the division range to obtain the dynamic area includes: Based on the transition area design standard of the highway facilities points, divide the direction along the highway extension into the pre-sequence influence area and the subsequent influence area, align the time stamps of the image data, traffic flow, and vehicle speed according to the division, slice the image data, extract the vehicle type images and vehicle position changes in the slices, calculate the vehicle spacing using the vehicle position changes and vehicle speed, and use the vehicle spacing, traffic flow, and vehicle speed to analyze the traffic flow speed sequence using the car-following model to obtain the acceleration wave characteristics and deceleration wave characteristics. If three wave superpositions occur in the acceleration wave and deceleration wave, expand the pre-sequence influence area and the subsequent influence area according to the multiple.

[0008] As a further method, the method for obtaining the short-term motion characteristics using the vehicle type images of the dynamic area includes: Based on the vehicle type images and vehicle speed of the dynamic area, extract the vehicle positions and vehicle speeds of different vehicles at different times in the image data slices as short-term motion data, use the short-term motion data to construct a spatio-temporal propagation matrix, and calculate the fluctuation energy density of multi-vehicle motion at different positions and times. The formula for the spatio-temporal propagation matrix is: where is the spatio-temporal propagation matrix, n is the total number of vehicles, representing at position and time the fluctuation energy density at, is the vehicle's speed at time , is the speed of the i-th vehicle at time The speed, is that the time interval is determined by the time stamp scale, is the position coordinate of the i-th vehicle, is the Dirac delta function, which takes an infinite value only at and is 0 at other times, is the wave propagation speed and the average wave speed of highway traffic flow, is that the wave propagates from to the target position The required time delay; According to the short-term motion data, the car-following model is used to analyze the vehicle motion characteristics. The wave energy density is input into the car-following model as an additional stimulus term to correct the acceleration in the vehicle motion characteristics. The jerk is calculated according to the time change rate of the acceleration, the speed is obtained by integrating the acceleration, the vehicle positions at different times are connected in chronological order to form a motion trajectory, and the lateral offset and heading angle are obtained based on the lane markings and vehicle positions; the relative speed and distance of adjacent vehicles are calculated using the vehicle positions and vehicle speeds of the vehicle and adjacent vehicles to obtain short-term motion characteristics, and the short-term motion characteristics include jerk, acceleration, speed, motion trajectory, heading angle, lateral offset, relative speed of adjacent vehicles, and relative distance of adjacent vehicles.

[0009] As a further method, the method of obtaining the first vehicle type in the front and rear intervals of the charging facility according to the vehicle type image, taking the remaining vehicle types as the second vehicle type, and calculating the charging demand index according to the first vehicle type that enters and does not enter the charging facility includes: Based on the vehicle type image in the dynamic area of the charging facility, the license plate color is extracted by the license plate color recognition algorithm as vehicle type information. The green background with white characters or gradually changing green background license plates are classified as the first vehicle type using the vehicle type information, and other vehicles are classified as the second vehicle type; the first vehicle type is divided into two categories: those that enter and do not enter the charging facility. The lane-changing selection data is extracted from the motion trajectory of the first vehicle type that enters and does not enter the charging facility, and the queuing selection data of the first vehicle type is obtained based on the lane-changing selection data according to the two categories of those that enter and do not enter the charging facility. The charging demand index is calculated according to the first vehicle type that enters and does not enter the charging facility. The formula for calculating the charging demand index is: where is the number of charging vehicles in the first vehicle type, is the number of the first vehicle type in the dynamic area of the charging facility, is the vehicle 's charging identifier, is the charging vehicle in the first vehicle type 's wave energy density, is the wave energy density of the first vehicle type in the dynamic area of the charging facility, is the charging vehicle in the first vehicle type Heading angle change amount Is the heading angle change amount of the first vehicle type in the dynamic area of the charging facility Is the average value of the heading angle change amount of the first vehicle type in the dynamic area of the charging facility Among the first vehicle type, for the charging vehicle Lane change selection Is the lane change selection of the first vehicle type in the dynamic area of the charging facility Is the lane change average value of the first vehicle type, expressed as a lane change ratio Among the first vehicle type, for the charging vehicle Speed change amount Is the speed change amount of the first vehicle type in the dynamic area of the charging facility Is the average value of the speed change amount of the first vehicle type in the dynamic area of the charging facility Among the first vehicle type, for the charging vehicle Queue selection Is the queue selection of the first vehicle type in the dynamic area of the charging facility Is the average value of the queue selection of the first vehicle type in the dynamic area of the charging facility, expressed as a queue ratio

[0010] As a further method, the method of obtaining the motion data of the first vehicle type and the second vehicle type by fitting the short-term motion characteristics of the first vehicle type and the second vehicle type according to the charging demand index and traffic flow in the dynamic area, and calculating the conflict index of the first vehicle type and the conflict index of the second vehicle type using the motion data of the first vehicle type, the motion data of the second vehicle type and the partition road data, includes: Align the time stamps of the charging demand index, traffic flow in the dynamic area, and the short-term motion characteristics of the first vehicle type and the second vehicle type. Use the input charging demand index and traffic flow as input features, and use the short-term motion characteristics of the first vehicle type and the second vehicle type as output labels. Perform non-linear fitting through a random forest according to the input features and output labels to obtain the motion data of the first vehicle type and the motion data of the second vehicle type. The motion data of the first vehicle type and the motion data of the second vehicle type include the lane change lead amount and the deceleration curve slope; Obtain the partition road data of the dynamic area according to the road data. The partition road data includes the lane width and the curve radius. Calculate the conflict index of the first vehicle type and the conflict index of the second vehicle type through the conflict index formula using the motion data of the first vehicle type, the motion data of the second vehicle type, the corresponding short-term motion characteristics, and the partition road data. The conflict index formula is: Where Is the Relative distance between adjacent vehicles Is the time conflict point, indicating the time for the two vehicles to maintain their current motion states until collision Is the jerk The absolute value of is the course angle gradient, is the vehicle speed change amount, CDI is the charging demand index, is the lane change advance amount, is the deceleration curve slope.

[0011] As a further method, the method for obtaining traffic flow change data through the cellular transmission model based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type movement data, the second vehicle type movement data, and the zoned road data includes: Uniformly divide based on the dynamic zone length to obtain cellular units, use the sections of the acceleration wave starting point and the deceleration wave ending point as the cellular boundaries, use the lane width, the curve radius, and the facility point type as cellular attributes, obtain the traffic flow density based on the divisor of the number of vehicles and the cellular section area, obtain the vehicle position and vehicle speed according to the short-term movement data, obtain the vehicle type ratio using the proportion of the first vehicle type, use the first vehicle type conflict index and the second vehicle type conflict index as the cellular risk coefficients, use the charging demand index as the cellular behavior driving parameter, determine the inflow based on the lane width, conflict index, and lane change advance amount at the exit of the upstream cell, determine the outflow according to the lane width and deceleration curve slope of the downstream cell, trigger the charging induction behavior for the first vehicle type using the charging demand index, and use the vehicle acceleration, jerk, and course angle gradient in the short-term movement characteristics as the initial car-following parameters to obtain the cellular transmission model; Based on the cellular transmission model, use the traffic flow fluctuation time scale as the time step, update the states of each cell in the order of the upstream cell, the current cell, and the downstream cell, calculate the inflow, outflow, internal vehicle type ratio change, and speed decay in sequence, and obtain the traffic flow change data according to the output of the cellular transmission model of the traffic flow density, the proportion of the first vehicle type, the average speed, and the queue length of each cell at different times.

[0012] As a further method, the method for analyzing the travel characteristics of the first vehicle type on the highway to be processed through a graph neural network based on the first vehicle type movement data, the first vehicle type conflict index, the traffic flow change data, and the highway topology structure includes: Extract the highway topology structure according to the road data and facility points, obtain the graph structure based on the highway topology structure, map the dynamic zone to the nodes in the graph structure, endow the nodes with the first movement data, the first conflict index, the traffic flow change data, and the position information of the highway topology structure, use the connection relationship of the road data as the edges of the nodes to obtain the graph structure data, and divide it into a training set, a validation set, and a test set Input the training set into the graph neural network. The model aggregates node features through at least two layers of graph convolutional layers, uses the attention mechanism layer to aggregate the value vectors of adjacent nodes through weighted coefficients, and uses the time series graph convolutional layer to process the temporal evolution of node states. During training, calculate the differences between the prediction results and the image data, traffic flow, and vehicle speed through the cross-entropy loss function. Input the validation set into the model, and verify the overfitting state according to the changes in the loss function value and accuracy. If the accuracy improvement of the validation set is less than 0.7% for 5 consecutive rounds, end the training. Use the backpropagation algorithm to feedback the differences to each layer of the model to adjust the parameters until the maximum number of iterations is reached to complete the training. Input the test set into the trained graph neural network, and evaluate the model performance using accuracy and recall. Output the travel characteristics of the first vehicle type based on the trained model.

[0013] The second aspect of the present invention provides a highway new energy vehicle travel characteristic analysis system, including: Data acquisition module: used to collect the to-be-processed highway data for preprocessing to obtain image data, traffic flow, vehicle speed, road data, and facility points. Data processing module: partition the sections where vehicles enter and exit the facilities based on the facility points, obtain the acceleration wave characteristics, deceleration wave characteristics, and vehicle type images of the partitions according to the image data, traffic flow, and vehicle speed of the partitions, and use the acceleration wave characteristics and deceleration wave characteristics to correct the partition range to obtain the dynamic area. Traffic flow change simulation module: used to obtain the short-term motion characteristics using the vehicle type images in the dynamic area, obtain the first vehicle type in the intervals before and after the charging facilities according to the vehicle type images, regard the remaining vehicle types as the second vehicle type, calculate the charging demand index according to the first vehicle types that enter and do not enter the charging facilities, fit the short-term motion characteristics of the first vehicle type and the second vehicle type according to the charging demand index and traffic flow in the dynamic area to obtain the first vehicle type motion data and the second vehicle type motion data, calculate the first vehicle type conflict index and the second vehicle type conflict index using the first vehicle type motion data, the second vehicle type motion data, and the partition road data, and obtain the traffic flow change data through the cell transmission model based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type motion data, the second vehicle type motion data, and the partition road data. Travel characteristic output module: analyze and obtain the travel characteristics of the first vehicle type on the to-be-processed highway through the graph neural network based on the first vehicle type motion data, the first vehicle type conflict index, the traffic flow change data, and the highway topology structure.

[0014] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention obtains a dynamic area by correcting the partition range of acceleration wave and deceleration wave characteristics, can respond to traffic flow fluctuations in real time, is more in line with actual traffic changes compared with fixed partitions, accurately captures the impact of vehicles entering and leaving facilities on traffic flow, lays a more reliable foundation for subsequent traffic characteristic analysis, and improves the accuracy of overall traffic analysis.

[0015] (2) The present invention combines a car-following model, a cellular transmission model and a graph neural network to analyze the traffic conditions on highways multi-dimensionally from microscopic vehicle motion acceleration, jerk, etc., macroscopic traffic flow simulation of traffic flow density, vehicle type ratio, etc., to comprehensive feature extraction, improving the accuracy and comprehensiveness of the analysis of traffic flow changes and the travel characteristics of new energy vehicles.

[0016] (3) The present invention uses license plate color recognition to distinguish new energy vehicles from other vehicle types, calculates the charging demand index and analyzes their travel characteristics, takes into account the charging demand characteristics of new energy vehicles under heterogeneous traffic flow, provides data support for the layout of highway charging facilities, the formulation of traffic guidance strategies, etc., promotes the convenience of new energy vehicles traveling on highways, and optimizes traffic management and resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of the steps of a method for analyzing the travel characteristics of new energy vehicles on highways in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Referring to Figure 1 as shown, the present invention provides a method for analyzing the travel characteristics of new energy vehicles on highways, including: Collecting the data of the highway to be processed for preprocessing to obtain image data, traffic flow, vehicle speed, road data and facility points; In the actual evaluation, a section of highway to be processed is selected, and image data with a resolution of 1920×1080 and a frame rate of 25fps is collected through road monitoring cameras. Median filtering is used to remove noise, and then histogram equalization is used to enhance the image contrast. The traffic flow and vehicle speed data are collected by microwave sensors along the road section and recorded every 5 minutes. Cubic spline interpolation is used to repair missing or abnormal data. The GPS coordinates (latitude 30.2°, longitude 120.5°) of gas stations, service areas, ramps, and charging facilities are obtained, and road data such as lane width (3.75 meters) and bend radius are obtained by combining with high-precision maps.

[0020] Based on the facility points, the sections where vehicles enter and exit the facilities are divided into zones. According to the image data, traffic flow, and vehicle speed of the divided zones, the acceleration wave characteristics, deceleration wave characteristics, and vehicle type images of the zones are obtained. The dynamic zone is obtained by using the acceleration wave characteristics and deceleration wave characteristics to correct the zone range. In the actual evaluation, according to the design standard of the transition zone at the charging facility points, the 500 meters upstream along the extension direction of the highway is divided into the pre-sequence influence zone and the subsequent influence zone of 300 meters downstream. The image data, average traffic flow of 200 vehicles / hour, and average vehicle speed of 80km / h in the divided zones are time-stamped and aligned at 5-minute intervals. The image data is sliced at 10-second intervals per frame. The vehicle type images are extracted through the YOLOv5 model. The vehicle spacing is calculated by using the change in the vehicle displacement position between adjacent frames of the video edge and the vehicle speed, and the average vehicle spacing is obtained as 30 meters. The GM car-following model is used to analyze the traffic flow speed sequence. At a certain moment, three wave superpositions of acceleration waves and deceleration waves occur, and the pre-sequence influence zone and the subsequent influence zone are each extended by 100 meters to form the dynamic zone.

[0021] The short-term motion characteristics are obtained by using the vehicle type images in the dynamic zone. The first vehicle type in the interval before and after the charging facility is obtained according to the vehicle type image, and the remaining vehicle types are used as the second vehicle type. The charging demand index is calculated according to the first vehicle type that enters and does not enter the charging facility. It should be explained that the fluctuation energy density is input into the car-following model as an additional stimulus item to correct the acceleration in the vehicle motion characteristics. The core logic is that the fluctuation energy density reflects the degree of dynamic disturbance of the traffic flow around the vehicle, and this disturbance will affect the acceleration adjustment of the vehicle. For example, when the traffic flow around the vehicle fluctuates greatly, the vehicle needs to adjust the acceleration more frequently to maintain safe car-following. By incorporating the fluctuation energy density into the car-following model, it becomes a factor for correcting the acceleration, so as to more accurately simulate the acceleration change of the vehicle in a complex traffic environment.

[0022] In the actual evaluation, based on the vehicle type images and vehicle speeds in a dynamic area, the positions of different vehicles at different times in the image data slices are extracted through the YOLOv5 model. The speed is calculated by dividing the position difference between two adjacent frames by the time interval. Among them, the positions of 50 vehicles are integrated into one position, the wave propagation speed is 60 km / h, and the time interval is 2 seconds. The wave energy density of the multi-vehicle movement is calculated to be 28.37. This value and the short-term movement data are input into the GM model to correct the vehicle acceleration. For example, a vehicle accelerates from 105 km / h to 124 km / h, and the acceleration is 0.72 m / s². The jerk is calculated to be 0.1 m / s³ according to the acceleration change rate. The speed is obtained by integrating the acceleration. The vehicle positions are connected in chronological order to form a movement trajectory. Based on the lane markings and vehicle positions, the lateral offset with an average of 0.2 meters and the heading angle with an angle of 5° with the lane line are obtained. The relative speed of adjacent vehicles is calculated. For example, the speed difference between two adjacent vehicles is 5 km / h and the distance is 25 meters, obtaining short-term movement characteristics, including jerk, acceleration, speed, movement trajectory, heading angle, lateral offset, relative speed of adjacent vehicles, and distance.

[0023] In the actual evaluation, based on the vehicle type images and vehicle speeds in a dynamic area, the license plate colors are identified through the YOLOv5 model. New energy vehicles with green license plates on white backgrounds are the first vehicle type, and other vehicles are the second vehicle type. The first vehicle type is divided into two categories: those entering and not entering the charging facilities. Through identification, =1 indicates entering the charging facility, =0 indicates not entering the charging facility. The lane-changing selection data of the first vehicle type entering and not entering the charging facilities is extracted through the movement trajectory. Based on the lane-changing selection data, the queuing selection data of the first vehicle type is obtained for the two categories of entering and not entering the charging facilities. The lane-changing selection data and the queuing selection data are normalized and the values are taken as 0 and 1. 1 0 Lane change, 1 Queue 0 Queue There are a total of 50 first vehicle types (new energy vehicles) in the dynamic area, among which 20 are actual charging vehicles entering the charging facilities. For the 20 vehicles entering the charging facilities, 15 are judged to be queuing through the trajectory, and the 30 vehicles not entering the charging facilities are defaulted not to queue. Among them, the wave energy density of the actual charging vehicles = 32.5, and the wave energy density of all first vehicle types = 28.37. Substituting into the charging demand index formula, the calculated charging demand index CDI is 0.53.

[0024] According to the charging demand index and traffic flow in the dynamic area, the short-term motion characteristics of the first vehicle type and the second vehicle type are fitted to obtain the first vehicle type motion data and the second vehicle type motion data. The first vehicle type conflict index and the second vehicle type conflict index are calculated using the first vehicle type motion data, the second vehicle type motion data, and the zoned road data. Based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type motion data, the second vehicle type motion data, and the zoned road data, traffic flow change data is obtained through the cellular transmission model; In the actual evaluation, the charging demand index, traffic flow data, and the short-term motion characteristics of the first vehicle type and the second vehicle type are collected. These data are time-stamped and aligned at a time interval of every 5 minutes, and a total of 288 time-stamped data are collected. The charging demand index and traffic flow data are used as input features to form a matrix, and the short-term motion characteristics of the first and second vehicle types are used as output labels. The input features are divided into a training set and a test set at a ratio of 7:3. Using the scikit-learn library, a random forest regression model is initialized, with the number of estimators set to 100, the maximum depth set to 10, and the random state set to 42. The data of the first and second vehicle types are trained separately. During the model training, each decision tree in the forest learns the input features to predict the output labels. After training, the mean squared error (MSE) is calculated using the test set to evaluate the performance. For example, the MSE of the first vehicle type is 0.03, and that of the second vehicle type is 0.04, indicating that the model performs well. Finally, the trained model is used to fit the data. For example, for new data with a CDI of 0.6 and a traffic flow of 1500 vehicles / hour, the fitted values of the lane change advance amount of the first vehicle type are 0.6 seconds, and the slope of the deceleration curve is -2.8 m / s²; for the second vehicle type, the lane change advance amount is 0.5 seconds, and the slope of the deceleration curve is -2.5 m / s². These fitted lane change advance amounts and slopes of the deceleration curve are the first and second motion data, providing key inputs for subsequent analysis. For example, the lane change advance amount is 0.5 seconds, and the slope of the deceleration curve is -2 m / s². According to a lane width of 3.75 meters and substituting into the conflict index formula, where = 25 meters, Divided by Is 5 seconds, = 5 km / h, lane change advance amount Is 0.5 seconds, jerk Is 0.1 m / s³, the yaw rate is 0.05 rad / m, slope of the deceleration curve Is -2 m / s², and the calculated first conflict index CTI is 0.85.

[0025] It should be noted that when constructing the cell transmission model, the above short-term motion characteristics are used as the initial car-following parameters. Specifically, if the vehicle acceleration is large, it indicates strong dynamic performance, and a relatively large initial car-following distance is set within the cell to avoid frequent adjustment of the distance due to strong acceleration ability. For vehicles with high jerk, it means that the acceleration changes violently, and the initial car-following distance is appropriately increased to ensure driving safety and comfort. If the heading angle gradient is large, it means that the vehicle's turning trend is significant, and the initial car-following parameters are adjusted, such as reducing the threshold of car-following distance change, so that it can adjust its position more smoothly when turning and reduce the conflict risk with adjacent vehicles. By integrating short-term motion characteristics into the initial car-following parameters, the cell transmission model better fits the actual vehicle motion characteristics, thereby more accurately simulating the transmission and change of traffic flow between cells.

[0026] In the actual evaluation, a 200-meter dynamic area is evenly divided into 4 cell units with a length of 50 meters. The starting point of the acceleration wave and the ending point of the deceleration wave define the cell boundaries. The lane width and the type of facility points are used as cell attributes. At this time, the traffic flow density is equal to the number of vehicles divided by the cell area, and 60 / (50×3.75) is 0.3 vehicles per meter. The vehicle position and speed are obtained according to the short-term motion data. The proportion of the first vehicle type is 30%. The conflict index of 0.85 is used as the cell risk coefficient, and the charging demand index of 0.65 is used as the cell behavior driving parameter. The inflow is determined based on the lane width, conflict index, and lane-changing advance amount at the exit of the upstream cell. , where is the lane width, and the outflow . The charging demand index of 0.65 is used to trigger the charging induction behavior for the first vehicle type, shortening the car-following distance of vehicles within a 5-meter cell. The cell state is updated with a 5-minute step, and the final output shows the traffic flow change data such as the traffic flow density (0.32 vehicles per meter), the proportion of the first vehicle type of 30%, the average speed of 78 km / h, and the queue length of 54.8 meters at different times for each cell, realizing the dynamic simulation and analysis of the congested traffic flow state on the highway.

[0027] Based on the motion data of the first vehicle type, the conflict index of the first vehicle type, the traffic flow change data, and the highway topology structure, the travel characteristics of the first vehicle type on the highway to be processed are obtained through graph neural network analysis.

[0028] In the actual evaluation, a 15-kilometer expressway section is selected, which contains 5 dynamic areas, 2 service areas, and 1 charging station. The expressway topology structure and graph structure data are extracted. One node is divided per kilometer, resulting in a total of 15 nodes. The dynamic areas are respectively mapped to nodes 3 - 5 (dynamic area 1), nodes 6 - 8 (dynamic area 2), and nodes 10 - 12 (dynamic area). The first motion data, the first conflict index, the traffic flow change data, and the position information of the expressway topology structure are assigned to the nodes. For example, the lane-changing lead time of node 4 in dynamic area 1 is 0.5 seconds, the deceleration curve slope is -2.5 m / s², the first conflict index is 0.7, the traffic flow change data is the traffic flow density of 0.8 vehicles / meter, the average speed is 85 km / h, and the position information is the longitude and latitude of node 4 (116.87, 39.92). Edges and data partitioning are determined according to the road connection relationships. For example, node 1 is connected to node 2, and node 2 is connected to node 3, etc., to construct the edges. The graph structure data is divided into a training set of 10 nodes, a validation set of 3 nodes, and a test set of 2 nodes according to a ratio of 7:2:1. Using the PyTorch Geometric library, a model with two graph convolutional layers is constructed. The first graph convolutional layer GCN fuses features such as lane-changing lead time and conflict index, and adopts multi-head attention to weighted aggregate the adjacent node value vectors. For example, 4-head attention is used to assign a weight of 0.6 to the traffic flow density of adjacent nodes and a weight of 0.4 to the speed. The time series graph convolutional layer collects data every hour to analyze the change in traffic flow density of node 4 at different time periods, and a graph neural network model is constructed using a convolutional layer with a one-dimensional convolutional kernel of 3; During training, through the cross-entropy loss function , where represents the value of the cross-entropy loss function, which is used to measure the difference between the model output and the true label, represents the number of samples, that is, how many data points are used to calculate the loss, represents the true label (actual value) of the th sample, represents the predicted label (predicted value) of the th sample, which is the probability prediction of the model for the sample i belonging to a certain category. Calculate the differences between the prediction and the actual image, traffic flow, and vehicle speed. During training, the validation set is input into the model. If the accuracy improvement is less than 0.7% for 5 consecutive rounds, the training ends. The model parameters are adjusted for each layer through the backpropagation algorithm, and the maximum number of iterations is set to 300 times. After the model training is completed, the test set is input into the model to calculate the model performance. Among them, the accuracy reaches 88%, and the recall rate reaches 83%. Finally, the first vehicle type travel characteristics are output through the trained graph neural network model. The first vehicle type travel characteristics include that at nodes 10 - 12 with a relatively high charging demand index, the average lane-changing lead time of the first vehicle type is shortened by 0.1 second; at node 4 with a relatively large conflict index, the average speed of the first vehicle type decreases by 6 km / h.

[0029] In this embodiment, the method for collecting the data of the highway to be processed for preprocessing to obtain image data, traffic flow, vehicle speed, and facility points includes: Collect the image data, traffic flow and vehicle speed data of the highway to be processed, obtain the original coordinates of charging facilities, fueling facilities and service area facilities, and the highway topology structure, perform filtering and image enhancement on the image data in sequence, use the interpolation method to repair abnormal data for the traffic flow and vehicle speed data, and match and correct the original coordinates of the facility points with the high-precision map to obtain the preprocessed image data, traffic flow, vehicle speed, road data and facility points.

[0030] In this embodiment, the method for partitioning the sections where vehicles enter and exit the facilities based on the facility points, and obtaining the acceleration wave characteristics, deceleration wave characteristics and vehicle type images of the partitions according to the image data, traffic flow and vehicle speed of the partitions, and using the acceleration wave characteristics and deceleration wave characteristics to correct the partition range to obtain the dynamic area includes: Based on the transition zone design standard of highway facility points, divide the highway extension direction into a pre-order influence area and a subsequent influence area. Align the time stamps of the image data, traffic flow and vehicle speed according to the partitions, slice the image data, extract the vehicle type images and vehicle position changes in the slices, calculate the vehicle spacing using the vehicle position changes and vehicle speed, and use the vehicle spacing, traffic flow and vehicle speed to analyze the traffic flow speed sequence using the car-following model to obtain the acceleration wave characteristics and deceleration wave characteristics. If three wave superpositions occur in the acceleration wave and deceleration wave, expand the pre-order influence area and the subsequent influence area according to the multiple.

[0031] In this embodiment, the method for obtaining short-term motion characteristics using the vehicle type images of the dynamic area includes: Based on the vehicle type images and vehicle speed in the dynamic area, extract the vehicle positions and vehicle speeds of different vehicles at different times in the image data slices as short-term motion data, use the short-term motion data to construct a spatio-temporal propagation matrix, and calculate the wave energy density of multi-vehicle motion at different positions and times. The formula of the spatio-temporal propagation matrix is: where is the spatio-temporal propagation matrix, n is the total number of vehicles, representing the wave energy density at position and time , is the th vehicle speed at time , is the speed of the i-th vehicle at time , is the time interval determined by the time stamp scale, is the position coordinate of the i-th vehicle, is the Dirac delta function, which takes an infinite value only when and is 0 at other times. is the wave propagation speed and the average wave speed of the highway traffic flow. The wave propagates from to the target position The required time delay; Analyze the vehicle motion characteristics using the car-following model based on short-term motion data. Input the wave energy density into the car-following model as an additional stimulus term to correct the acceleration in the vehicle motion characteristics. Calculate the jerk based on the time change rate of the acceleration, perform an integration operation on the acceleration to obtain the speed, connect the vehicle positions at different times in chronological order to form a motion trajectory, and obtain the lateral offset and heading angle based on the lane markings and vehicle positions; Calculate the relative speed and distance of adjacent vehicles using the vehicle positions and vehicle speeds of the vehicle and adjacent vehicles to obtain short-term motion characteristics, and the short-term motion characteristics include jerk, acceleration, speed, motion trajectory, heading angle, lateral offset, relative speed of adjacent vehicles, and relative distance of adjacent vehicles.

[0032] In this embodiment, the method for obtaining the first vehicle type in the front and rear intervals of the charging facility according to the vehicle type image, taking the remaining vehicle types as the second vehicle type, and calculating the charging demand index according to the first vehicle type entering and not entering the charging facility includes: Extract the license plate color as vehicle type information through the license plate color recognition algorithm based on the vehicle type image in the dynamic area of the charging facility. Use the vehicle type information to divide the green background with white characters or gradually changing green background license plates into the first vehicle type, and divide other vehicles into the second vehicle type; Divide the first vehicle type into two categories: entering and not entering the charging facility. Extract the lane-changing selection data for the first vehicle type entering and not entering the charging facility through the motion trajectory, obtain the queuing selection data for the first vehicle type based on the lane-changing selection data according to the two categories of entering and not entering the charging facility, and calculate the charging demand index according to the first vehicle type entering and not entering the charging facility. The charging demand index calculation formula is: Where is the number of charging vehicles in the first vehicle type, is the number of the first vehicle type in the dynamic area of the charging facility, is the vehicle 's charging identifier, is the charging vehicle in the first vehicle type 's wave energy density, is the wave energy density of the first vehicle type in the dynamic area of the charging facility, is the charging vehicle in the first vehicle type 's heading angle change amount, is the heading angle change amount of the first vehicle type in the dynamic area of the charging facility, is the average value of the change in the heading angle of the first vehicle type in the dynamic area of the charging facility, is the charging vehicle among the first vehicle types for lane change selection, is the lane change selection of the first vehicle type in the dynamic area of the charging facility, is the lane change mean value of the first vehicle type, expressed as a lane change ratio, is the charging vehicle among the first vehicle types for the change in speed, is the change in speed of the first vehicle type in the dynamic area of the charging facility, is the average value of the change in speed of the first vehicle type in the dynamic area of the charging facility, is the charging vehicle among the first vehicle types for queuing selection, is the queuing selection of the first vehicle type in the dynamic area of the charging facility, is the average value of the queuing selection of the first vehicle type in the dynamic area of the charging facility, expressed as a queuing ratio.

[0033] In this embodiment, the method of obtaining the motion data of the first vehicle type and the second vehicle type by fitting the short-term motion characteristics of the first vehicle type and the second vehicle type according to the charging demand index and traffic flow in the dynamic area, and calculating the conflict index of the first vehicle type and the conflict index of the second vehicle type using the motion data of the first vehicle type, the motion data of the second vehicle type and the partition road data, includes: Align the time stamps of the charging demand index, traffic flow, and the short-term motion characteristics of the first vehicle type and the second vehicle type in the dynamic area. Use the input charging demand index and traffic flow as input features, and use the short-term motion characteristics of the first vehicle type and the second vehicle type as output labels. Perform non-linear fitting through a random forest based on the input features and output labels to obtain the motion data of the first vehicle type and the motion data of the second vehicle type. The motion data of the first vehicle type and the motion data of the second vehicle type include the lane change advance amount and the deceleration curve slope. Obtain the partition road data of the dynamic area according to the road data. The partition road data includes the lane width and the curve radius. Calculate the conflict index of the first vehicle type and the conflict index of the second vehicle type through the conflict index formula using the motion data of the first vehicle type, the motion data of the second vehicle type, the corresponding short-term motion characteristics, and the partition road data. The conflict index formula is: where is the relative distance between adjacent vehicles, is the time conflict point, indicating the time for the two vehicles to maintain their current motion states until they collide, is the jerk absolute value of, is the heading angle gradient, Δv is the vehicle speed change, and CDI is the charging demand index. Δl is the lane change advance. k is the deceleration curve slope.

[0034] In this embodiment, the method for obtaining traffic flow change data through the cellular transmission model based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type movement data, the second vehicle type movement data, and the sectional road data includes: Uniformly divide based on the dynamic zone length to obtain cellular units, use the sections of the acceleration wave starting point and the deceleration wave ending point as the cellular boundaries, use the lane width, the curve radius, and the facility point type as cellular attributes, obtain the traffic flow density based on the divisor of the number of vehicles and the cellular section area, obtain the vehicle position and vehicle speed according to the short-term movement data, obtain the vehicle type ratio using the first vehicle type proportion, use the first vehicle type conflict index and the second vehicle type conflict index as the cellular risk coefficients, use the charging demand index as the cellular behavior driving parameter, determine the inflow based on the lane width, conflict index, and lane change advance of the upstream cellular exit, determine the outflow according to the lane width and deceleration curve slope of the downstream cell, trigger the charging induction behavior for the first vehicle type using the charging demand index, and use the vehicle acceleration, jerk, and heading angle gradient in the short-term movement characteristics as the initial car-following parameters to obtain the cellular transmission model; Based on the cellular transmission model, use the traffic flow fluctuation time scale as the time step, update the states of each cell in the order of the upstream cell, the current cell, and the downstream cell, calculate the inflow, outflow, change in the internal vehicle type ratio, and speed decay in sequence, and obtain the traffic flow change data according to the output of the cellular transmission model of the traffic flow density, the proportion of the first vehicle type, the average speed, and the queue length of each cell at different times; In this embodiment, the method for analyzing and obtaining the travel characteristics of the first vehicle type on the highway to be processed through a graph neural network based on the first vehicle type movement data, the first vehicle type conflict index, the traffic flow change data, and the highway topology structure includes: Extract the highway topology structure according to the road data and facility points, obtain the graph structure based on the highway topology structure, map the dynamic zone to the nodes in the graph structure, endow the nodes with the first movement data, the first conflict index, the traffic flow change data, and the position information of the highway topology structure, and use the connection relationship of the road data as the edges of the nodes to obtain the graph structure data, which is divided into a training set, a validation set, and a test set. Input the training set into the graph neural network. The model aggregates node features through at least two layers of graph convolutional layers, uses the attention mechanism layer to aggregate the value vectors of adjacent nodes through weighted coefficients, and uses the time series graph convolutional layer to process the temporal evolution of node states. During training, calculate the differences between the prediction results and the image data, traffic flow, and vehicle speed through the cross-entropy loss function. Input the validation set into the model, and verify the overfitting state based on the changes in the loss function value and accuracy. If the accuracy improvement of the validation set is less than 0.7% for 5 consecutive rounds, end the training. Use the backpropagation algorithm to feedback the differences to each layer of the model to adjust the parameters until the maximum number of iterations is reached to complete the training. Input the test set into the trained graph neural network, and evaluate the model performance using accuracy and recall. Output the travel characteristics of the first vehicle type based on the trained model.

[0035] The second aspect of the present invention also provides a highway new energy vehicle travel characteristic analysis system, including: Data acquisition module: used to collect the highway data to be processed for preprocessing to obtain image data, traffic flow, vehicle speed, road data, and facility points. Data processing module: partition the sections where vehicles enter and exit the facilities based on the facility points, obtain the acceleration wave characteristics, deceleration wave characteristics, and vehicle type images of the partitions according to the image data, traffic flow, and vehicle speed of the partitions, and use the acceleration wave characteristics and deceleration wave characteristics to correct the partition range to obtain the dynamic area. Traffic flow change simulation module: used to obtain the short-term motion characteristics using the vehicle type images in the dynamic area, obtain the first vehicle type in the intervals before and after the charging facilities according to the vehicle type images, regard the remaining vehicle types as the second vehicle type, calculate the charging demand index according to the first vehicle types that enter and do not enter the charging facilities, fit the short-term motion characteristics of the first vehicle type and the second vehicle type according to the charging demand index and traffic flow in the dynamic area to obtain the first vehicle type motion data and the second vehicle type motion data, calculate the first vehicle type conflict index and the second vehicle type conflict index using the first vehicle type motion data, the second vehicle type motion data, and the partition road data, and obtain the traffic flow change data through the cell transmission model based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type motion data, the second vehicle type motion data, and the partition road data. Travel characteristic output module: analyze and obtain the travel characteristics of the first vehicle type on the highway to be processed through the graph neural network based on the first vehicle type motion data, the first vehicle type conflict index, the traffic flow change data, and the highway topology structure.

[0036] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by this claim book, they should all belong to the protection scope of the present invention.

Claims

1. A method for analyzing the travel characteristics of new energy vehicles on highways, characterized in that: The following steps are involved: Collect the highway data to be processed for preprocessing to obtain image data, traffic volume, vehicle speed, road data and facility locations; The road sections where vehicles enter and exit the facilities are divided into zones based on the facility points. The acceleration wave characteristics, deceleration wave characteristics and vehicle model images of the zones are obtained according to the image data, traffic flow and vehicle speed of the zones. The acceleration wave characteristics and deceleration wave characteristics are used to correct the zone range to obtain the dynamic zone. Use the vehicle type image in the dynamic area to obtain short-term motion characteristics, obtain the first vehicle type in the interval before and after the charging facility based on the vehicle type image, take the remaining vehicles as the second vehicle type, and calculate the charging demand index based on the first vehicle type that enters and does not enter the charging facility; According to the charging demand index and the traffic flow in the dynamic area, the short-term motion characteristics of the first vehicle model and the second vehicle model are fitted to obtain the motion data of the first vehicle model and the second vehicle model, and the conflict index of the first vehicle model and the conflict index of the second vehicle model are calculated using the motion data of the first vehicle model, the motion data of the second vehicle model and the partitioned road data. Based on the conflict index of the first vehicle model, the conflict index of the second vehicle model, the motion data of the first vehicle model, the motion data of the second vehicle model and the partitioned road data, the traffic flow change data is obtained through a cellular transmission model; Based on the movement data of the first vehicle model, the conflict index of the first vehicle model, the traffic flow change data and the highway topology structure, the travel characteristics of the first vehicle model on the highway to be processed are obtained through graph neural network analysis.

2. A method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1, characterized in that: The method of collecting and preprocessing the highway data to be processed to obtain image data, vehicle flow, vehicle speed and facility location includes: Collect image data, traffic flow and speed data of the highway to be processed, obtain the original coordinates and highway topology of the charging facilities, refueling facilities and service area facilities, filter and enhance the image data in turn, use interpolation method to repair abnormal data of traffic flow and speed data, match and correct the original coordinates of the facility points with the high-precision map, and obtain the pre-processed image data, traffic flow, speed, road data and facility points.

3. A method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1, characterized in that: The method of partitioning the road sections where vehicles enter and exit the facility based on the facility points, obtaining the partition acceleration wave characteristics, deceleration wave characteristics and vehicle model images according to the partition image data, vehicle flow and vehicle speed, and using the acceleration wave characteristics and deceleration wave characteristics to correct the partition range to obtain the dynamic zone includes: Based on the design standard of transition zones of highway facility points, the extension direction of the highway is divided into preceding influence zones and subsequent influence zones. The image data, traffic flow and vehicle speed are time-stamp aligned according to the partitions, the image data is sliced, the vehicle model images and vehicle position changes in the slices are extracted, the vehicle distance is calculated using the vehicle position change and vehicle speed, and the following model is used to analyze the traffic speed sequence based on the vehicle distance, traffic flow and vehicle speed to obtain the acceleration wave characteristics and deceleration wave characteristics. If three waves of acceleration and deceleration waves are superimposed, the preceding influence zone and subsequent influence zone will be expanded by multiples.

4. A method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1, characterized in that: The method for obtaining short-term motion characteristics using a vehicle model image in a dynamic area comprises: Based on the dynamic area vehicle model image and vehicle speed, the vehicle positions and vehicle speeds of different vehicles at different times in the image data slice are extracted as short-term motion data. The short-term motion data is used to construct a spatiotemporal propagation matrix to calculate the fluctuation energy density of multi-vehicle motion at different positions and times. The spatiotemporal propagation matrix formula is: in is the space-time propagation matrix, n is the total number of vehicles, indicating the position and time The fluctuating energy density at For the Car in time speed, is the i-th vehicle at time speed, The time interval is determined by the timestamp scale. is the position coordinate of the i-th vehicle, is the Dirac delta function, only in The value is infinite at the time and 0 at other times. is the wave propagation speed, is the average wave speed of highway traffic, For fluctuations from Propagate to target location the time delay required; A car-following model is used to analyze vehicle motion characteristics based on short-term motion data, and the fluctuating energy density is input into the car-following model as an additional stimulus item to correct the acceleration in the vehicle motion characteristics. The jerk is calculated based on the time rate of change of the acceleration, and the speed is obtained by integrating the acceleration. The vehicle positions at different times are connected in chronological order to form a motion trajectory, and the lateral offset and heading angle are obtained based on the lane markings and the vehicle position. The relative speed and distance of the adjacent vehicle are calculated using the vehicle position and vehicle speed of the vehicle and the adjacent vehicle to obtain short-term motion characteristics, and the short-term motion characteristics include jerk, acceleration, speed, motion trajectory, heading angle, lateral offset, relative speed of adjacent vehicles, and relative distance of adjacent vehicles.

5. The method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1 is characterized in that: The method of obtaining the first vehicle model in the interval before and after the charging facility according to the vehicle model image, taking the remaining vehicle models as the second vehicle models, and calculating the charging demand index according to the first vehicle models that have entered and have not entered the charging facility includes: Based on the vehicle image in the dynamic area of ​​the charging facility, the license plate color is extracted as the vehicle model information through the license plate color recognition algorithm. The vehicle model information is used to classify the license plate with white characters on a green background or a gradient green background as the first vehicle model, and the other vehicles are classified as the second vehicle model; the first vehicle model is divided into two categories: entering and not entering the charging facility, and the lane change selection data of the first vehicle model entering and not entering the charging facility is extracted through the motion trajectory. Based on the lane change selection data, the queue selection data of the first vehicle model is obtained according to the two categories of entering and not entering the charging facility, and the charging demand index is calculated according to the first vehicle model entering and not entering the charging facility. The charging demand index calculation formula is: in is the number of charging vehicles in the first model, is the number of first vehicle types in the dynamic area of ​​the charging facility, For vehicles Charging logo, Charging vehicles for the first model The fluctuating energy density, is the fluctuating energy density of the first vehicle type within the dynamic area of ​​the charging facility, Charging vehicles for the first model The heading angle change, is the heading angle change of the first vehicle type in the dynamic area of ​​the charging facility, is the average value of the heading angle change of the first vehicle type in the dynamic area of ​​the charging facility, Charging vehicles for the first model Lane change options, Lane change selection for the first vehicle type within the dynamic area of ​​the charging facility, is the mean lane change value of the first vehicle type, expressed as the lane change ratio, Charging vehicles for the first model The speed change, is the speed change of the first vehicle type in the dynamic area of ​​the charging facility, is the average speed change of the first vehicle type in the dynamic area of ​​the charging facility, Charging vehicles for the first model The queue selection, For the first vehicle type queuing selection within the dynamic area of ​​the charging facility, It is the mean of the queue selection of the first vehicle type in the dynamic area of ​​the charging facility, expressed as the queue proportion.

6. A method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1, characterized in that: The method of fitting the short-term motion characteristics of the first vehicle model and the second vehicle model according to the charging demand index and the vehicle flow in the dynamic area to obtain the motion data of the first vehicle model and the second vehicle model, and calculating the conflict index of the first vehicle model and the second vehicle model using the motion data of the first vehicle model, the motion data of the second vehicle model and the partitioned road data includes: The charging demand index, vehicle flow, and short-term motion characteristics of the first vehicle model and the second vehicle model in the dynamic area are timestamped, and the input charging demand index and vehicle flow are used as input features, and the short-term motion characteristics of the first vehicle model and the second vehicle model are used as output labels. Nonlinear fitting is performed through random forest according to the input features and the output labels to obtain the motion data of the first vehicle model and the second vehicle model, and the motion data of the first vehicle model and the second vehicle model include the lane change advance amount and the deceleration curve slope; the partitioned road data of the dynamic area is obtained according to the road data, and the partitioned road data includes the lane width and the curve radius. The first vehicle model motion data and the second vehicle model motion data and the corresponding short-term motion characteristics are combined with the partitioned road data to calculate the first vehicle model conflict index and the second vehicle model conflict index through the conflict index formula, and the conflict index formula is: in For the The relative distance between neighboring vehicles, is the time conflict point, which means the time from the two vehicles maintaining the current motion state to the collision. Jerk The absolute value of is the heading angle gradient, is the speed change of the vehicle, CDI is the charging demand index, is the lane change advance amount, is the slope of the deceleration curve.

7. A method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1, characterized in that: The method for obtaining traffic flow change data through a cellular transmission model based on the first vehicle type conflict index, the second vehicle type conflict index, the first vehicle type motion data, the second vehicle type motion data and the partitioned road data includes: The cell unit is obtained by uniformly dividing the length of the dynamic area, and the cell boundary is obtained by using the road section of the acceleration wave starting point and the deceleration wave ending point. The lane width, curve radius, and facility point type are used as cell attributes. The traffic density is obtained based on the divisor of the number of vehicles and the cell section area. The vehicle position and vehicle speed are obtained according to the short-term motion data. The vehicle model ratio is obtained by using the proportion of the first vehicle model. The conflict index of the first vehicle model and the conflict index of the second vehicle model are used as the cell risk coefficient. The charging demand index is used as the cell behavior driving parameter. The inflow is determined based on the lane width, conflict index, and lane change advance of the upstream cell exit. The outflow is determined based on the lane width and deceleration curve slope of the downstream cell. The charging demand index is used to trigger the charging induced behavior for the first vehicle model. The vehicle acceleration, jerk, and heading angle gradient in the short-term motion characteristics are used as the initial following parameters to obtain the cell transmission model. Based on the cellular transmission model, the traffic flow fluctuation time scale is used as the time step, and the state of each cell is updated in the order of upstream cells, current cells, and downstream cells. The inflow, outflow, internal vehicle type ratio change and speed attenuation are calculated in turn. According to the cellular transmission model, the vehicle flow density, the proportion of the first vehicle type, the average speed and the queue length of each cell at different times are output to obtain the traffic flow change data.

8. The method for analyzing the travel characteristics of new energy vehicles on highways according to claim 1 is characterized in that: The method for obtaining the travel characteristics of the first vehicle type on the to-be-processed highway through graph neural network analysis based on the first vehicle type motion data, the first vehicle type conflict index, the traffic flow change data and the highway topology structure includes: The highway topology is extracted based on the road data and facility points, and the graph structure is obtained based on the highway topology. The dynamic area is mapped to the node in the graph structure, and the first motion data, the first conflict index, the traffic flow change data and the location information of the highway topology structure are used to assign the node. The connection relationship of the road data is used as the edge of the node to obtain the graph structure data, which will be divided into training set, verification set and test set. The training set is input into the graph neural network. The model aggregates node features through at least two layers of graph convolution layers. The attention mechanism layer is used to aggregate the value vectors of adjacent nodes through weighted coefficients. The time series graph convolution layer is used to process the temporal evolution of node states. During training, the cross entropy loss function is used to calculate the difference between the prediction results and the image data, traffic flow, and vehicle speed. The validation set is input into the model, and the overfitting state is verified according to the change in loss function value and accuracy. If the accuracy of the validation set is improved by less than 0.7% for five consecutive rounds, the training is terminated. The difference is fed back to each layer of the model to adjust the parameters using the back propagation algorithm until the training is completed when the maximum number of iterations is reached. The test set is input into the trained graph neural network, and the model performance is evaluated using accuracy and recall. The travel characteristics of the first vehicle model are output based on the trained model.

9. A highway new energy vehicle travel characteristics analysis system, used to execute a highway new energy vehicle travel characteristics analysis method according to any one of claims 1 to 8, characterized in that: The system comprises: Data acquisition module: used to collect the highway data to be processed for pre-processing, and obtain image data, traffic flow, vehicle speed, road data and facility locations; Data processing module: divide the road sections where vehicles enter and exit the facilities into zones based on the facility locations, obtain the zone acceleration wave characteristics, deceleration wave characteristics and vehicle model images based on the zone image data, vehicle flow and vehicle speed, and use the acceleration wave characteristics and deceleration wave characteristics to correct the zone range and obtain the dynamic zone; Traffic flow change simulation module: used to obtain short-term motion characteristics using vehicle model images in dynamic areas, obtain the first vehicle model in the interval before and after the charging facility based on the vehicle model images, and use the remaining vehicle models as the second vehicle models, calculate the charging demand index based on the first vehicle model that enters and does not enter the charging facility, fit the short-term motion characteristics of the first vehicle model and the second vehicle model based on the charging demand index and traffic flow in the dynamic area to obtain the first vehicle model motion data and the second vehicle model motion data, calculate the first vehicle model conflict index and the second vehicle model conflict index using the first vehicle model motion data, the second vehicle model motion data and the partitioned road data, and obtain the traffic flow change data through the cellular transmission model based on the first vehicle model conflict index, the second vehicle model conflict index, the first vehicle model motion data, the second vehicle model motion data and the partitioned road data; Travel feature output module: Based on the motion data of the first vehicle model, the conflict index of the first vehicle model, the traffic flow change data and the highway topology structure, the travel characteristics of the first vehicle model on the highway to be processed are obtained through graph neural network analysis.

Images