A new energy vehicle trajectory estimation method and system based on multi-source data fusion

Abstract

The application provides a new energy vehicle trajectory estimation method and system based on multi-source data fusion, and belongs to the field of intelligent traffic data processing and vehicle trajectory mining. The method comprises the following steps: collecting and preprocessing gantry license plate recognition data and service area charging transaction data, and extracting vehicle space-time features; a charging pile queuing waiting time probability model is constructed; a space-time matching window is constructed based on the service area pile number position and the upstream and downstream gantry distance, and an initial candidate vehicle set is generated; a hard constraint layer is formed by the endurance feasibility constraint and the speed physical rationality constraint, a soft constraint layer is formed by combining the time closeness and the speed mode consistency score, and the comprehensive matching confidence of the candidate vehicle and the charging record is calculated; the Hungarian algorithm is used to solve the maximum weight matching of the bipartite graph, and the optimal trajectory estimation result is output. The application solves the problem of non-ETC new energy vehicle trajectory tracking failure during the holiday free period, and realizes accurate recovery and confidence quantification of the vehicle whole journey trajectory.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation data processing and vehicle trajectory mining technology, and particularly relates to a method and system for estimating the trajectory of new energy vehicles based on multi-source data fusion. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the deepening of global energy structure transformation and environmental protection policies, new energy vehicles have experienced explosive growth in ownership due to their core advantages of low emissions and high energy efficiency. This trend is particularly pronounced during holidays, when demand for cross-regional highway travel continues to rise, and new energy vehicles have become an important part of the highway network. As a key service node supporting long-distance travel for new energy vehicles, the charging infrastructure in highway service areas directly determines the overall service level of the road network through its scientific layout, capacity adaptability, and operational efficiency. It is also a core factor affecting user travel experience and alleviating range anxiety.

[0004] Currently, vehicle trajectory tracking, especially for new energy vehicles and travel behavior analysis, mainly relies on ETC gantry toll data and traditional toll records. Its core logic is to reconstruct the trajectory by associating vehicle entry and exit from highways and the tolls passed through. However, under the context of toll-free highways during statutory holidays, this traditional method faces a fundamental challenge: non-ETC vehicles can pass freely without a card, leading to a break in the toll data chain, rendering trajectory tracking of such vehicles completely ineffective. Simultaneously, although license plate recognition data collected by highway gantry systems and charging transaction data generated by service area charging piles exist independently, they lack a unified vehicle identity association mechanism and spatiotemporal alignment standard. Gantry data only reflects vehicle transit node information, and charging data only records vehicle charging behavior in service areas, failing to effectively connect the complete behavioral chain of "entering the highway - passing through road sections - charging in service areas - leaving the highway."

[0005] The aforementioned data fragmentation has led to a series of industry pain points: On the one hand, the spatiotemporal traceability of charging demand lacks a basis, making it impossible to accurately pinpoint which road sections and time periods have charging needs for new energy vehicles, and also making it difficult to analyze vehicle route selection preferences and charging decision-making logic; on the other hand, the lack of data support for assessing the service radius of charging facilities and predicting load pressure results in unreasonable planning and layout of charging networks in service areas, unbalanced capacity allocation, and prominent phenomena of charging queues and congestion during peak hours coexisting with some station resources being idle, seriously restricting the scientific formulation of dynamic scheduling strategies. In addition, existing technologies have not fully considered special scenarios such as surges in traffic during holidays, frequent traffic congestion, and long queues at charging piles, further reducing the accuracy of trajectory correlation and the reliability of charging behavior analysis. Summary of the Invention

[0006] To overcome the shortcomings of the existing technologies, this invention provides a method and system for estimating the trajectory of new energy vehicles based on multi-source data fusion. By fusing gantry license plate recognition data and service area charging transaction data, the complete trajectory chain of new energy vehicles passing through upstream and downstream gantries of service areas during free holiday periods is reconstructed, thereby realizing the full trajectory estimation and spatiotemporal feature mining of charging vehicles.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for estimating the trajectory of new energy vehicles based on multi-source data fusion; A method for estimating the trajectory of new energy vehicles based on multi-source data fusion includes: Collect and clean highway gantry license plate recognition data and service area charging transaction data, and extract the time series of vehicles passing through the gantries, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location. Construct a probability model of queuing time based on service area size, number of charging piles, time period characteristics, and holiday type, and generate a dynamic queuing time estimation function; Based on the service area station location and the distance between upstream and downstream gantry, an adaptive spatiotemporal matching window is constructed to generate an initial candidate vehicle set that meets time constraints; The candidate vehicles are subjected to dual verification with hard and soft constraints, and the comprehensive matching confidence of each vehicle with the charging record is calculated. Construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results.

[0008] As a further technical solution, the process of cleaning, calibrating, and extracting features from highway gantry signage data and service area charging data, while simultaneously constructing a probabilistic model for charging pile queuing time, includes: Collect gantry license plate recognition data, remove records with confidence levels below a set threshold, sort by license plate number and time to form an initial trajectory sequence, calibrate equipment timing deviation through historical data features, and extract the initial spatiotemporal features of each vehicle. Collect charging data from charging piles, remove abnormal charging data, and extract the spatiotemporal characteristics of the charging process for each vehicle. Based on historical holiday data, a probability model of charging pile queuing time is constructed by considering the size of the service area, the number of available charging piles, time period characteristics, and holiday type to infer the dwell time of vehicles in the service area.

[0009] As a further technical solution, the queuing time estimation function is: T_queue = f(Service Area ID, Time Period Characteristics, Holiday Type, Number of Available Charging Stations) Where T_queue is the queuing time, and f is a statistical distribution model based on historical data.

[0010] As a further technical solution, based on the service area's station location and the distance between upstream and downstream gantry points, an adaptive spatiotemporal matching window is constructed to generate an initial candidate vehicle set that meets time constraints, including: Calculate the distance ratio coefficients λ_up and λ_down based on the mileage difference between the service area center station number and the upstream and downstream gantry station numbers; The total travel time T_travel between gantry frames is divided into upstream travel time T_travel_up and downstream travel time T_travel_down according to the distance ratio; Construct an upstream alignment window and filter gantry records that satisfy T_charge_start - T_queue_max ≤ T_pass_up + T_travel_up ≤ T_charge_start - T_queue_min; where T_charge_start is the charging start time; T_queue_max is the maximum queuing time; T_queue_min is the minimum queuing time; T_pass_up is the upstream gantry passage time; and T_travel_up is the upstream segment travel time. Construct a downstream alignment window and filter gantry records that satisfy T_charge_end + T_depart_min ≤ T_pass_down - T_travel_down ≤ T_charge_end + T_depart_max; where T_charge_end is the charging end time; T_depart_min is the minimum departure preparation time; T_depart_max is the maximum departure preparation time; T_pass_down is the downstream gantry passage time; and T_travel_down is the downstream segment travel time. Vehicles that simultaneously meet the upstream and downstream window conditions are selected to form the initial candidate set C_initial.

[0011] As a further technical solution, hard constraint verification is performed on candidate vehicles, including range feasibility constraints and speed physical rationality constraints. The range feasibility constraints are as follows: the battery capacity is estimated based on the charging amount, and the vehicle type is inferred to be a hybrid vehicle, a small pure electric vehicle, or a medium-to-large pure electric vehicle based on the battery capacity range. For pure electric vehicles, the driving distance after charging is calculated, and it is verified whether the actual driving distance D_actual after charging satisfies D_actual ≤ D_max × α, where η is the charging efficiency, E_avg is the average energy consumption, and α is the safety factor. The physical rationality constraint for speed is: the average speed between the upstream and downstream gantry sections must be within a preset threshold range.

[0012] As a further technical solution, soft constraint verification is performed on the candidate vehicles, including: The time proximity score S_time is calculated based on the deviation between the theoretical time for the vehicle to pass through the gantry and the start and end times of charging. A speed pattern consistency score S_speed is calculated based on the similarity between the vehicle's speed within the service area and its historical baseline speed. The overall score is calculated as Score = Σ(w_i × S_i), where w_i is the weight of each soft constraint item, Σw_i=1; and S_i represents the normalized score of the i-th soft constraint item.

[0013] As a further technical solution, a bipartite graph of charging records and candidate vehicles is constructed, and a global optimization algorithm is used to solve for the maximum weight matching, outputting trajectory estimation results, including: Construct a bipartite graph of charging record-candidate trajectory, use the Hungarian algorithm to solve the maximum weight matching, calculate the comprehensive confidence score, and output the complete trajectory sequence and corresponding confidence value.

[0014] The second aspect of this invention provides a trajectory estimation system for new energy vehicles based on multi-source data fusion.

[0015] A trajectory estimation system for new energy vehicles based on multi-source data fusion includes: The data acquisition and preprocessing module is configured to: collect and clean highway gantry license plate recognition data and service area charging transaction data, extract the time series of vehicles passing through the gantries, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location; The queuing time modeling module is configured to: construct a queuing time probability model based on service area size, number of charging piles, time period characteristics and holiday type, and generate a dynamic queuing time estimation function; The hierarchical spatiotemporal screening module is configured to: construct an adaptive spatiotemporal matching window based on the service area station location and the distance between upstream and downstream gantries, and generate an initial candidate vehicle set that meets time constraints; The confidence score module is configured to perform dual verification of candidate vehicles with hard and soft constraints, and calculate the comprehensive matching confidence of each vehicle with the charging record. The optimal matching module is configured to: construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results.

[0016] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a new energy vehicle trajectory estimation method based on multi-source data fusion as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the new energy vehicle trajectory estimation method based on multi-source data fusion as described in the first aspect of the present invention.

[0018] The above one or more technical solutions have the following beneficial effects: (1) This invention constructs a cross-system bridge linking vehicle identity and behavior by integrating highway gantry license plate recognition data and service area charging transaction data. During free passage periods on holidays when ETC toll data is invalid, this method can successfully restore the driving trajectory of non-ETC new energy vehicles by utilizing full-coverage license plate recognition data and detailed charging behavior data, filling the technical gap in vehicle trajectory tracking in this scenario and providing a key data source that was previously difficult to obtain for traffic management and service planning. By introducing dynamic spatiotemporal window construction and queuing time probability model, it can adapt to complex traffic conditions such as holiday traffic surges and road congestion, dynamically adjust the fault tolerance range of time matching, and more accurately depict the complete timeline of vehicle travel and service area stay. Combined with trajectory legality verification based on gantry topology, it strictly ensures the consistency of the restored trajectory in spatial logic and passage order, effectively filters out erroneous matches that violate physical laws, and greatly improves the overall credibility of trajectory reconstruction results.

[0019] (2) This invention identifies vehicle types by estimating battery capacity and establishes differentiated remaining driving range models. Starting from the vehicle's own physical characteristics, it can quickly eliminate candidate matches that cannot complete a specific journey in terms of driving range, thereby compressing the trajectory search space in the early stages of the algorithm, reducing unnecessary computation, and improving the overall processing efficiency and practicality of the method. Simultaneously, this invention not only performs basic spatiotemporal correlation but also delves into multi-dimensional behavioral characteristics such as driving speed patterns, service area dwell patterns, and charging behavior preferences to construct a behavioral profile of individual vehicles. By using these behavioral patterns as auxiliary criteria for cross-validation and consistency matching, the rationality of trajectory correlation can be further identified and confirmed from the perspective of driving habits, making the trajectory reconstruction results not only "logically sound" but also "behavioralally consistent," enhancing the depth and persuasiveness of the analytical conclusions.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of the method in the first embodiment.

[0023] Figure 2 This is a schematic diagram illustrating a complete travel trajectory example of a new energy vehicle in the first embodiment.

[0024] Figure 3 This is a system structure diagram of the second embodiment. Detailed Implementation

[0025] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0026] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0027] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0028] Example 1 This embodiment discloses a new energy vehicle trajectory estimation method based on multi-source data fusion. By fusing gantry identification and charging transaction data, and through dynamic spatiotemporal window construction, multi-level verification, and maximum weight matching solution, it can accurately restore the complete trajectory of new energy vehicles during free periods, solve the problem of trajectory tracking failure for non-ETC vehicles, and provide reliable data support for charging facility planning and road network service optimization.

[0029] Specifically, such as Figure 1 As shown, a method for estimating the trajectory of new energy vehicles based on multi-source data fusion includes: Step S1: Collect and clean highway gantry license plate recognition data and service area charging transaction data, and extract the time sequence of vehicles passing through the gantry, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location.

[0030] Step S11 involves using a gantry recognition system deployed along the highway to collect license plate recognition data from passing vehicles in real time. The collected data includes core fields such as license plate number, recognition confidence level, gantry number, passage time, and gantry marker number. A confidence level threshold of 85% is set, and an automated program filters and removes fuzzy recognition records with confidence levels below this threshold to prevent low-quality data from interfering with subsequent analysis. Clearly abnormal data, such as empty license plate numbers and invalid gantry numbers, are also removed.

[0031] The vehicles are grouped by license plate number, and the valid records of the same vehicle are sorted in ascending order of passage time to form an initial trajectory sequence for each vehicle. To address the time discrepancies caused by inconsistent time synchronization among different gantry equipment, the time feature library of historical gantry passage data from the same period is called up, and the passage time of each record is calibrated using linear interpolation to ensure consistency in the time dimension.

[0032] From the calibrated trajectory sequence, the initial spatiotemporal features of each vehicle are extracted, including the gantry passage time sequence, station position sequence, and travel time between adjacent gantries.

[0033] Step S12: Collect vehicle charging process data through the highway service area charging pile management system, including information such as license plate number, charging start time, charging end time, charging amount, charging power, service area number, and charging pile number.

[0034] Set up abnormal data filtering rules to automatically remove invalid charging records with a charging time of less than 5 minutes and abnormal lingering records with a charging time of more than 8 hours; at the same time, remove logically contradictory data such as charging amount of 0 and charging power that is always zero to ensure data authenticity.

[0035] From the cleaned and valid data, the spatiotemporal characteristics of the charging process of each vehicle are extracted, including the charging start time, charging end time, charging duration, charging amount, and service area location.

[0036] Step S2: Construct a probability model of queuing time based on service area size, number of charging piles, time period characteristics, and holiday type, and generate a dynamic queuing time estimation function.

[0037] Historical charging data from service areas during holidays over the past three years were collected and categorized by service area size, holiday type, and time period. Based on historical charging queuing data, a probabilistic model of queuing time was constructed, considering service area heterogeneity, time period characteristics, and holiday effects, to estimate the probability distribution of potential vehicle queuing duration. Specifically, a multivariate statistical model was constructed using service area ID, time period, holiday type, and number of available charging piles as input variables, and queuing time as the output variable. The model was trained using a multiple linear regression algorithm, and the model parameters were optimized by dividing historical data into a training set (70%) and a validation set (30%). The final model is a queuing time probability model T_queue = f (service area ID, time period, holiday type, number of available charging piles), which can accurately infer the potential queuing time of vehicles in service areas based on real-time input variables, providing data support for subsequent dwell time estimation.

[0038] Step S3: Based on the service area station location and the distance between upstream and downstream gantry, construct an adaptive spatiotemporal matching window to generate an initial candidate vehicle set that meets the time constraints; In this embodiment, T_queue_min and T_depart_min are both set to 0 by default, T_queue_max is calculated to be 60 minutes by the queuing waiting time probability model, and T_depart_max is set to 20 minutes according to the service area operation specifications.

[0039] First, the distance ratio coefficients are calculated. The center station of the highway service area, the station of its upstream gantry, and the station of the nearest downstream gantry are obtained. Based on these station numbers, the mileage difference L_upstream from the upstream gantry to the service area and the mileage difference L_downstream from the downstream gantry to the service area are calculated. According to the distance ratio coefficient calculation formula, the upstream distance ratio coefficient λ_up = L_upstream / (L_upstream + L_downstream) and the downstream distance ratio coefficient λ_down = L_downstream / (L_upstream + L_downstream) are calculated respectively.

[0040] After calculating the coefficients, the total travel time between gantries is proportionally divided. Through the gantry identification data preprocessing stage, the total travel time T_travel between the upstream and downstream gantries is extracted. Based on the calculated distance ratio coefficients, the total travel time is divided into upstream and downstream segments. The travel time from the upstream gantry to the service area is T_travel_up = T_travel × λ_up, and the travel time from the service area to the downstream gantry is T_travel_down = T_travel × λ_down.

[0041] Subsequently, a bidirectional time alignment window is constructed and gantry record filtering is completed. For a single charging record in a service area, the charging start time T_charge_start and charging end time T_charge_end are obtained. When constructing the upstream alignment window, the effective range of the upstream gantry passage time T_pass_up is determined according to the time constraint formula T_charge_start - T_queue_max ≤ T_pass_up + T_travel_up ≤ T_charge_start - T_queue_min, and the upstream gantry sign records within this time period are filtered out. The upstream alignment window is constructed to filter gantry records that satisfy T_charge_start - T_queue_max ≤ T_pass_up + T_travel_up ≤ T_charge_start - T_queue_min; where T_charge_start is the charging start time; T_queue_max is the maximum queuing time; T_queue_min is the minimum queuing time; T_pass_up is the upstream gantry passage time; and T_travel_up is the upstream segment travel time.

[0042] When constructing the downstream alignment window, the effective range of the downstream gantry passage time T_pass_down is determined based on the constraint formula T_charge_end + T_depart_min ≤ T_pass_down - T_travel_down ≤ T_charge_end + T_depart_max, and the downstream gantry license plate records within this time period are filtered out. Where T_charge_end is the charging end time; T_depart_min is the minimum departure preparation time; T_depart_max is the maximum departure preparation time; T_pass_down is the downstream gantry passage time; and T_travel_down is the downstream segment travel time.

[0043] Finally, an initial candidate vehicle set is generated. License plate number matching is performed on the gantry license plate recognition records filtered by the upstream and downstream alignment windows. New energy vehicles appearing simultaneously in both upstream and downstream valid records are extracted. These vehicles are then integrated to form the initial candidate set C_initial, providing basic vehicle samples for subsequent candidate set disambiguation and confidence scoring. This embodiment, through the above steps, accurately selects vehicles that meet spatiotemporal constraints from massive gantry license plate recognition data, effectively compressing the search space for trajectory matching and improving the efficiency and accuracy of subsequent trajectory estimation.

[0044] Step S4: Perform dual verification of candidate vehicles using both hard and soft constraints, and calculate the comprehensive matching confidence of each vehicle with the charging record.

[0045] The process involves hard constraint verification for candidate vehicles, including range feasibility constraints and speed physical rationality constraints. For range feasibility constraints, charging data for candidate vehicles is extracted, and battery capacity is estimated using the formula C_battery=Q_charge / ΔSOC. Vehicle types are categorized by capacity: <20kWh for hybrid vehicles, 20-60kWh for small pure electric vehicles, and >60kWh for mid-to-large pure electric vehicles. Hybrid vehicles directly pass this constraint; pure electric vehicles are calculated using D_max=Q_charge×η / E_avg for the driving distance after charging, and E_avg is adjusted seasonally to verify that the actual driving distance D_actual ≤ D_max×α. For example, candidate vehicle A has a charging capacity of 50kWh, ΔSOC=0.8, and an estimated battery capacity of 62.5kWh, classifying it as a mid-to-large pure electric vehicle. The calculated D_max=45km, and the verified D_actual=40km≤45×1.2=54km, passes the constraint.

[0046] To ensure the physical rationality of the speed, the difference between the time and distance between the upstream and downstream gantries of the candidate vehicle is extracted, the average speed of the interval is calculated, and it is verified whether it is within the range of [20km / h, 120km / h].

[0047] After completing the hard constraint verification, soft constraint verification is carried out on qualified vehicles and a comprehensive score is calculated. The score results are all normalized to the [0,1] interval.

[0048] The time deviation is calculated using Δt_up = |(T_upstream_pass + T_travel_upstream) - T_start| and Δt_down = |(T_downstream_pass - T_travel_downstream) - T_end|. This time deviation is then substituted into the formula S_time = max(0, 1- (Δt_up + Δt_down) / (2 × δ_allow)) to calculate the time proximity score S_time. Here, Δt_up is the upstream time deviation; T_upstream_pass is the actual time it takes for the vehicle to pass the nearest upstream gantry of the service area; T_travel_upstream is the theoretical travel time from the upstream gantry to the service area, calculated by multiplying the total travel time between upstream and downstream gantry by the upstream distance proportionality coefficient λ_up; and T_start is the actual start time of charging at the service area's charging station. Δt_down represents the downstream time deviation, which is the absolute difference between the theoretical departure time of the vehicle from the service area to the downstream gantry and the actual charging end time. T_downstream_pass represents the actual time it takes for the vehicle to pass the nearest downstream gantry of the service area. T_travel_downstream represents the theoretical travel time of the vehicle from the service area to the downstream gantry, calculated by multiplying the total travel time between upstream and downstream gantry by the downstream distance ratio coefficient λ_down. T_end represents the actual charging end time of the vehicle at the charging station in the service area. δ_allow represents the maximum allowed total time deviation threshold.

[0049] The speed pattern consistency score S_speed is obtained by calculating the similarity between the service area speed V_charge and the vehicle's historical baseline speed V_base using S_speed = max(0, 1 |(V_charge - V_base) / V_base|).

[0050] Finally, the overall matching confidence score is calculated by weighted summation according to the formula Score=Σ(w_i×S_i), where w_i is the weight of each soft constraint item, Σw_i=1; and S_i represents the normalized score result of the i-th soft constraint indicator.

[0051] This embodiment uses hard constraints to quickly eliminate invalid candidate vehicles and soft constraints to achieve refined matching scores, providing accurate confidence basis for subsequent global optimal matching and effectively improving the accuracy and efficiency of trajectory matching.

[0052] Step S5: Construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for maximum weight matching, and output the trajectory estimation results.

[0053] The filtered charging records and candidate vehicles in the service area are respectively used as two types of nodes in a bipartite graph. The left node set consists of valid charging records r_i in the service area, with each record corresponding to the charging behavior of one new energy vehicle; the right node set consists of candidate vehicles c_j after double verification, with each node corresponding to one new energy vehicle that meets the trajectory matching conditions. Based on the comprehensive matching confidence score Score_ij of each candidate vehicle and each charging record obtained in step S4, the edge weight w_ij = Score_ij × λ_conflict connecting charging record r_i and candidate vehicle c_j in the bipartite graph is calculated. This completes the assignment of edge weights between all valid nodes, forming a complete bipartite graph model. Subsequently, the Hungarian algorithm is used to solve the global maximum weight matching, which is different from the local greedy matching method.

[0054] Specifically, the maximum weight matching problem with edge weights w_ij is transformed into a minimum cost problem. A cost matrix D is constructed, where D_ij = W_max - w_ij, and W_max is the maximum value among all edge weights w_ij. Then, the Hungarian algorithm is executed: for each charging record node r_i in the left node set, a label L(r_i) is assigned, representing the minimum value of the corresponding row in the cost matrix D; for each candidate vehicle node c_j in the right node set, a label L(c_j) = 0 is assigned. An equivalent subgraph G_L is constructed, retaining only edges that satisfy L(r_i) + L(c_j) = D_ij.

[0055] A depth-first search strategy is employed to find augmenting paths starting from the currently unmatched left node. If a perfect match is found, i.e., all charging records are uniquely matched or explicitly marked as unmatched, the optimal matching result is output; otherwise, the vertex labels are adjusted. Let S be the set of left nodes traversed in the current search process, and T be the set of right nodes traversed. The adjustment amount delta = min{L(r_i) + L(c_j) - D_ij | r_i belongs to S, c_j does not belong to T}; for all nodes in S, L(r_i) is updated to L(r_i) - delta, and for all nodes in T, L(c_j) is updated to L(c_j) + delta; after updating the equivalent subgraph G_L, an augmenting path is searched again.

[0056] The above process outputs the optimal matching result, yielding the optimal matching matrix X. Here, x_ij=1 indicates a successful match between charging record r_i and candidate vehicle c_j, while x_ij=0 indicates a mismatch. Adjacency list storage is used to optimize the algorithm's time complexity, meeting the real-time processing requirements of large-scale data during holidays.

[0057] Finally, for the charging record-candidate vehicle combinations obtained by the Hungarian algorithm, the original comprehensive confidence score corresponding to the edge weights is restored, and confidence level is completed according to a preset threshold. Simultaneously, combining gantry identification data and charging transaction data, the complete trajectory sequence of each vehicle is reconstructed. The trajectory information includes core spatiotemporal features such as the vehicle's upstream and downstream gantry passage time, service area arrival time, queuing time, charging start and end times, departure time, and driving parameters for each road segment. The complete trajectory sequence is correlated with the corresponding comprehensive confidence value and output. High-confidence results are directly adopted, medium-confidence results are marked as suspicious and suggested for sampling review, and low-confidence results trigger anomaly alarms and require manual intervention.

[0058] Figure 2 This example demonstrates the complete travel trajectory of a new energy vehicle reconstructed using the method of this invention. It intuitively presents the entire spatiotemporal information of the vehicle from entering the highway, passing through a service area for charging, to exiting the highway. It records the time of the vehicle entering and exiting a certain highway, as well as the information on the stop at a service area, and then estimates the time to enter the service area and the actual charging period. The charging amount is 30.01 kWh, and the battery level is charged from 47% to 80%. This trajectory is generated by fusing gantry license plate recognition data and charging transaction data. The estimated entry time to the service area is calculated by the queuing time probability model of this invention. The entry and exit times and the charging time meet the adaptive spatiotemporal matching window constraints. It is verified by hard constraints such as range feasibility and speed rationality, and soft constraints such as time proximity and speed consistency. It is a high-confidence trajectory result output after maximum weight matching using a bipartite graph and the Hungarian algorithm. It clearly verifies the technical effect of this invention in accurately reconstructing the complete trajectory of non-ETC new energy vehicles in scenarios such as free highways during holidays and ETC data failure. It provides intuitive and reliable data support for charging facility planning and road network service optimization.

[0059] Example 2 This embodiment discloses a trajectory estimation system for new energy vehicles based on multi-source data fusion; like Figure 3 As shown, a trajectory estimation system for new energy vehicles based on multi-source data fusion includes: The data acquisition and preprocessing module is configured to: collect and clean highway gantry license plate recognition data and service area charging transaction data, extract the time series of vehicles passing through the gantries, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location; The queuing time modeling module is configured to: construct a queuing time probability model based on service area size, number of charging piles, time period characteristics and holiday type, and generate a dynamic queuing time estimation function; The hierarchical spatiotemporal screening module is configured to: construct an adaptive spatiotemporal matching window based on the service area station location and the distance between upstream and downstream gantries, and generate an initial candidate vehicle set that meets time constraints; The confidence score module is configured to perform dual verification of candidate vehicles with hard and soft constraints, and calculate the comprehensive matching confidence of each vehicle with the charging record. The optimal matching module is configured to: construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results.

[0060] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0061] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a new energy vehicle trajectory estimation method based on multi-source data fusion as described in Example 1.

[0062] Example 4 The purpose of this embodiment is to provide an electronic device.

[0063] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the new energy vehicle trajectory estimation method based on multi-source data fusion as described in Embodiment 1.

[0064] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0065] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0066] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for estimating the trajectory of new energy vehicles based on multi-source data fusion, characterized in that, In the absence of ETC billing data, including: Collect and clean highway gantry license plate recognition data and service area charging transaction data, and extract the time series of vehicles passing through the gantries, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location. Construct a probability model of queuing time based on service area size, number of charging piles, time period characteristics, and holiday type, and generate a dynamic queuing time estimation function; Based on the service area station location and the distance between upstream and downstream gantry stations, an adaptive spatiotemporal matching window is constructed to generate an initial candidate vehicle set that meets time constraints; specifically: based on the mileage difference between the service area center station and the upstream and downstream gantry station, the distance ratio coefficients λ_up and λ_down are calculated. The total travel time T_travel between gantry frames is divided into upstream travel time T_travel_up and downstream travel time T_travel_down according to the distance ratio; Construct an upstream alignment window and filter gantry records that satisfy T_charge_start - T_queue_max ≤ T_pass_up + T_travel_up ≤ T_charge_start - T_queue_min; where T_charge_start is the charging start time; T_queue_max is the maximum queuing time; T_queue_min is the minimum queuing time; T_pass_up is the upstream gantry passage time; and T_travel_up is the upstream segment travel time. Construct a downstream alignment window and filter gantry records that satisfy T_charge_end + T_depart_min ≤ T_pass_down - T_travel_down ≤ T_charge_end + T_depart_max; where T_charge_end is the charging end time; T_depart_min is the minimum departure preparation time; T_depart_max is the maximum departure preparation time; T_pass_down is the downstream gantry passage time; and T_travel_down is the downstream segment travel time. Vehicles that simultaneously satisfy both upstream and downstream window conditions are selected to form the initial candidate set C_initial; Candidate vehicles are subjected to both hard and soft constraints for verification, and the comprehensive matching confidence of each vehicle with the charging record is calculated. Hard constraint verification for candidate vehicles includes range feasibility constraints and speed physical rationality constraints. The range feasibility constraint is as follows: battery capacity is estimated based on the charging charge, and the vehicle type is inferred to be a hybrid vehicle, a small pure electric vehicle, or a medium-to-large pure electric vehicle based on the battery capacity range. For pure electric vehicles, the driving distance after charging is calculated, and it is verified whether the actual driving distance D_actual after charging satisfies D_actual ≤ D_max × α, where η is the charging efficiency, D_max is the driving distance after charging, and α is the safety factor. The speed physical rationality constraint is: the average speed between the upstream and downstream gantry sections is within a preset threshold range. Soft constraint verification is performed on candidate vehicles, including calculating the time closeness score S_time based on the deviation between the theoretical time of the vehicle passing through the gantry and the charging start and end time; calculating the speed pattern consistency score S_speed based on the similarity between the vehicle's speed in the service area and the historical benchmark speed; and calculating the comprehensive score Score = Σ(w_i × S_i), where w_i is the weight of each soft constraint item, Σw_i=1; and S_i represents the normalized score result of the i-th soft constraint indicator. Construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results.

2. The new energy vehicle trajectory estimation method based on multi-source data fusion as described in claim 1, characterized in that, The highway gantry signage data and service area charging data were cleaned, calibrated, and feature extracted. Simultaneously, a probabilistic model of charging pile queuing time was constructed, including: Collect gantry license plate recognition data, remove records with confidence levels below a set threshold, sort by license plate number and time to form an initial trajectory sequence, calibrate equipment timing deviation through historical data features, and extract the initial spatiotemporal features of each vehicle. Collect charging data from charging piles, remove abnormal charging data, and extract the spatiotemporal characteristics of the charging process for each vehicle. Based on historical holiday data, a probability model of charging pile queuing time is constructed by considering the size of the service area, the number of available charging piles, time period characteristics, and holiday type to infer the dwell time of vehicles in the service area.

3. The new energy vehicle trajectory estimation method based on multi-source data fusion as described in claim 1, characterized in that, The queuing time estimation function is: T_queue = f(Service Area ID, Time Period Characteristics, Holiday Type, Number of Available Charging Stations) Where T_queue is the queuing time, and f is a statistical distribution model based on historical data.

4. The new energy vehicle trajectory estimation method based on multi-source data fusion as described in claim 1, characterized in that, Construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results, including: Construct a bipartite graph of charging records and candidate trajectories, use the Hungarian algorithm to solve for maximum weight matching, calculate the comprehensive confidence score, and output the complete trajectory sequence and corresponding confidence value.

5. A trajectory estimation system for new energy vehicles based on multi-source data fusion, employing the trajectory estimation method for new energy vehicles based on multi-source data fusion as described in any one of claims 1-4, characterized in that, include: The data acquisition and preprocessing module is configured to: collect and clean highway gantry license plate recognition data and service area charging transaction data, extract the time series of vehicles passing through the gantries, the chainage sequence, the travel time between adjacent gantries, as well as the charging start time, end time, charging duration, charging amount, and service area location; The queuing time modeling module is configured to: construct a queuing time probability model based on service area size, number of charging piles, time period characteristics, and holiday type, and generate a dynamic queuing time estimation function; The hierarchical spatiotemporal screening module is configured to: construct an adaptive spatiotemporal matching window based on the service area station location and the distance between upstream and downstream gantries, and generate an initial candidate vehicle set that meets time constraints; The confidence score module is configured to perform dual verification of candidate vehicles with hard and soft constraints, and calculate the comprehensive matching confidence of each vehicle with the charging record. The optimal matching module is configured to: construct a bipartite graph of charging records and candidate vehicles, use a global optimization algorithm to solve for the maximum weight matching, and output the trajectory estimation results.

6. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the new energy vehicle trajectory estimation method based on multi-source data fusion as described in any one of claims 1-4.

7. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the new energy vehicle trajectory estimation method based on multi-source data fusion as described in any one of claims 1-4.

Images