A highway new energy charging service intelligent optimization analysis method and system

Abstract

The application discloses a kind of expressway new energy charging service intelligent optimization analysis method and system, method includes: constructing structured expressway travel big data set.Extract vehicle multidimensional feature to distinguish new energy vehicle and fuel vehicle, obtain basic information;Based on inverse reinforcement learning, construct Markov decision process model, combine the fitting optimal return function of kiosk data, solve trajectory strategy, reconstruct new energy vehicle space-time trajectory, and mine space-time characteristics of charging demand.Combining service area distribution, power supply and cost constraints, construct multi-objective optimization model, through reinforcement learning and traffic simulation training, iteration site selection and facility configuration, output full-cycle layout and flexible configuration scheme.Real-time acquisition vehicle and site state, construct "car-station" collaborative scheduling model, generate site matching, pile position allocation and traffic guidance strategy, realize charging demand and service ability adaptive scheduling.

Description

Technical Field

[0001] This invention belongs to the field of highway traffic operation and management technology, and more specifically, relates to a method and system for intelligent optimization analysis of highway new energy charging services. Background Technology

[0002] The construction and operation of the current highway charging service system for new energy vehicles are facing practical challenges such as lagging demand perception, unbalanced resource allocation, and inefficient dispatch management, which seriously restrict the long-distance travel experience of new energy vehicles and the development of the industry. With the continuous increase in the number of new energy vehicles, the charging demand at highway service areas exhibits uneven spatial and temporal distribution. Some sections experience excessively long charging queues during peak hours, while some stations operate at low capacity for extended periods, resulting in a contradiction between resource waste and supply-demand imbalance.

[0003] Traditional charging station planning relies heavily on experience-based judgments, lacking precise analysis of the spatiotemporal characteristics of actual charging demand. This results in site selection and facility configuration failing to match dynamically changing user needs, increasing construction and maintenance costs and failing to effectively meet users' fast charging demands. In vehicle identification and demand analysis, existing methods often rely on single-dimensional data, making it difficult to accurately distinguish between new energy vehicles and gasoline vehicles, and failing to accurately obtain vehicle driving trajectories and charging preferences, thus failing to provide a reliable basis for demand forecasting.

[0004] At the charging service scheduling level, the existing model mostly relies on vehicles to choose charging stations independently, lacking a global collaborative scheduling mechanism. When faced with fluctuations in station load within a region, it cannot promptly guide traffic flow diversion, easily leading to localized charging congestion. These problems not only reduce users' travel efficiency but also hinder the intelligent upgrading process of the highway charging service system. There is an urgent need for a solution that can achieve accurate demand perception, optimized resource allocation, and efficient scheduling collaboration to promote the high-quality development of highway new energy charging services. Summary of the Invention

[0005] This invention aims to address the problems of delayed demand perception, unbalanced resource allocation, and inefficient scheduling management for new energy charging services on highways. By integrating and analyzing multi-source data, accurately mining demand, optimizing multi-objective configuration, and coordinating scheduling, it achieves dynamic adaptation between charging station layout and service capabilities, thereby improving charging service efficiency and user experience.

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, as a first aspect of this invention, the present invention provides an intelligent optimization analysis method for highway new energy charging services, comprising: S1. Collect highway toll station data, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to construct a structured highway travel big data set; S2. Analyze the preprocessed mixed traffic flow data to distinguish between new energy vehicles and fuel vehicles, and obtain basic information about new energy vehicles; then, construct a Markov decision process model based on inverse reinforcement learning, combine the basic information and checkpoint data, fit the optimal reward function and solve the optimal trajectory strategy, reconstruct the complete spatiotemporal trajectory of new energy vehicles, and explore the spatiotemporal distribution characteristics of charging demand. S3. Based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, and combined with the distribution of highway service areas, power supply conditions, and construction and operation cost constraints, a multi-objective optimization model is constructed with the goals of minimizing the total charging time of users, maximizing facility utilization, and minimizing operating costs. The model is trained based on reinforcement learning and traffic simulation, and the planning and site selection of charging stations in undeveloped areas and the number and capacity levels of facilities at existing stations are dynamically iterated. The model outputs a full-cycle optimized layout scheme and flexible configuration strategy that adapts to the dynamic spatiotemporal changes of charging demand. S4. Real-time collection of the status of new energy vehicles and the facility operation status of each charging station; construction of a "vehicle-station" collaborative scheduling model based on reinforcement learning method; generation of optimal charging station matching recommendations for new energy vehicles, dynamic allocation schemes of charging piles within stations, and cross-station traffic flow guidance strategies for different regional charging demand and station service capacity scenarios.

[0007] Furthermore, the structured highway travel big data set in S1 is specifically a multi-dimensional associated dataset constructed according to a unified data standard, spatiotemporal coordinate system, and field format. Each data module achieves unique cross-module association through vehicle ID, timestamp, and geographic location coordinates. Its core structured data modules and corresponding fields are as follows: The vehicle passage basic data module includes the following fields: vehicle unique identifier ID, license plate number, passage time, passage location information, toll station number, roadside checkpoint number, corresponding latitude and longitude coordinates, vehicle initial type identifier, entrance and exit station number, and entrance and exit station name. The vehicle energy consumption characteristic data module includes the following fields: vehicle unique identifier ID, data collection time, remaining power value, energy consumption rate per unit mileage, cumulative mileage, historical charging records, charging start time, charging end time, charging amount, charging station number, and the remaining power value is represented by the SOC value. The road network topology and infrastructure basic data module includes the following fields: unique road segment number, actual road segment length, road segment design speed, road segment starting point latitude and longitude coordinates, road segment ending point latitude and longitude coordinates, service area unique number, service area geographical coordinates, existing charging station unique number, station geographical coordinates, existing number of charging piles, single pile capacity level, and station power supply capacity limit. The real-time traffic flow data module includes the following fields: unique road segment number, statistical time slice, traffic flow at the road segment cross section, average driving speed of the road segment, traffic density of the road segment, and congestion status indicator of the road segment. The statistical time slice is set in five minutes per slice, and the congestion status indicator is divided into three categories: smooth, slow, and congested. The roadside monitoring associated data module includes the following fields: unique ID of the monitoring device, video recording time, structured storage path of key frame screenshots, pixel coordinates of the vehicle target in the image, vehicle shape feature extraction results, vehicle length, vehicle width, vehicle height, and unique identification features.

[0008] Furthermore, the process in S2 of analyzing the preprocessed mixed traffic flow data to distinguish between new energy vehicles and fuel vehicles, and obtaining basic information about new energy vehicles, is as follows: Based on the preprocessed mixed traffic flow data, two types of core distinguishing data are extracted. The first type is vehicle appearance identification data collected by roadside monitoring video, which directly identifies the vehicle's license plate format and exclusive new energy vehicle logo. The license plate format is determined according to the current motor vehicle license plate management regulations, and the exclusive new energy vehicle logo is confirmed according to the new energy exclusive symbol marked on the vehicle at the factory. The second category is vehicle energy replenishment record data collected at toll stations, which filters out vehicle data that shows charging behavior or energy replenishment. Cross-validate the two types of data, and determine that a vehicle is a new energy vehicle if it meets one of the following three conditions: the license plate format is a special format for new energy vehicles, it has a unique new energy logo, and its energy supply type is electric energy supply. Vehicles that do not meet all three conditions are classified as gasoline-powered vehicles. After classifying vehicles by type, basic information is extracted from the traffic flow data of new energy vehicles, including vehicle license plate information, travel time information, travel route information, and energy replenishment record information, forming a structured dataset of basic information of new energy vehicles, which provides data support for subsequent trajectory reconstruction and charging demand mining.

[0009] Furthermore, the process of constructing the complete spatiotemporal trajectory of the new energy vehicle in S2 is as follows: First, define the state space, action space, and state transition relationships of a Markov decision process: Let the state space be... Taking the highway segment as the core state unit, any state From the state vector Characterization, in which This is the only identifier for the road section. , The coordinates of the center of the road segment are latitude and longitude. This refers to the time it takes for a vehicle to pass through this section of road. The vehicle's speed on this section of road; Action space Defined as the set of path selection actions of a vehicle at a road segment node, i.e. ,in Indicates going straight. Indicates a left turn. Indicates a right turn; state transition probability Indicates the vehicle is in a certain state. Next action Then transferred to state The probability is determined based on the road network topology and the statistical characteristics of historical traffic flow paths. Secondly, construct and fit the optimal reward function, let the reward function be... Used to characterize the vehicle in state Execute action The subsequent benefits are constructed based on data including the passage time, location coordinates, and checkpoint capture records from the basic information of new energy vehicles, and are utilized through a deep neural network. Perform parameterized fitting on the reward function, i.e. ,in For neural network mapping functions, Network parameters; construct the objective function based on the maximum entropy criterion. ,in For parameterized strategies, It is the entropy function. Using entropy weights, the objective function is minimized using gradient descent. Iterative update parameters This continues until the objective function converges, yielding the optimal reward function. ; Finally, the optimal trajectory strategy is solved and the complete spatiotemporal trajectory is reconstructed using the optimal reward function. Based on this, an iterative strategy algorithm is used to solve for the optimal trajectory strategy. Combining the vehicle's unique identifier ID, travel time, and toll station and checkpoint capture records from the basic information of new energy vehicles, the optimal trajectory strategy is utilized. By performing path completion and temporal correlation on fragmented vehicle state nodes, a continuous state sequence of vehicles can be obtained. ,in This is the initial state. It is the final state and satisfies temporal continuity. This state sequence is the complete spatiotemporal trajectory of the new energy vehicle. By extracting the time of each state in the trajectory... with latitude and longitude coordinates This allows us to uncover the spatiotemporal distribution characteristics of charging demand.

[0010] Furthermore, the multi-objective optimization model in S3 is specifically as follows: First, define the set of decision variables as follows: ,in It can be 0 or 1, used to represent the first element in the undeveloped area. Whether charging stations should be planned and built at each of the candidate locations. Corresponding to the planning and construction, No corresponding planning or construction; For the first to land The number of charging piles configured at each charging station is a positive integer. For the first to land The capacity level of a single charging pile at each charging station is a discrete positive value; here A unique identifier for a candidate location or an existing site. Establish status indicators for the site. This is an identifier for the number of charging stations. For capacity level identification; Secondly, a multi-objective optimization system is constructed with the core objectives of minimizing the total charging time for users, maximizing facility utilization, and minimizing operating costs, and corresponding objective optimization functions are set for each objective. Finally, the constraints are set as follows: Service area distribution constraints are... ,in Taking 0 or 1 represents the first... Are the candidate locations within the service area? The corresponding location is within the service area. The corresponding location is not within the service area; power supply constraints are: ,in It is the first The upper limit of power supply capacity in the area where each site is located; construction and operation cost constraints are... ,in It is the preset upper limit of total construction and operation costs.

[0011] Furthermore, the objective optimization functions are as follows: The objective is to minimize the total user charging time. ,in This represents the total number of new energy vehicles. Assign vehicle number, The total number of charging stations. It is the first Vehicles and the The shortest driving distance to each station It is the first The average speed of the car It is the first The vehicle in The average waiting time at each site is determined by the decision variables. Decide; The goal of maximizing facility utilization is ,in The total number of time periods. Number the time period. It is the first The site at the The actual number of vehicles charging during the cycle. It is the first Duration of the cycle; The goal of minimizing operating costs is ,in It is the first The site construction cost for each candidate location, This refers to the purchase and installation cost of a single charging station. It is the energy consumption cost per unit capacity of a single pile. It is the site operation and maintenance cost, and the decision variables. Positive correlation.

[0012] Furthermore, the process of determining the full-cycle optimized layout scheme and flexible configuration strategy in S3 is as follows: First, a training environment is built based on reinforcement learning and traffic simulation: taking the spatiotemporal characteristics of charging demand as input, the decision variables of the multi-objective optimization model are used as the action space of the agent, and the comprehensive benefits of the user's total charging time, facility utilization rate, and operating cost are used as the reward function of the agent. At the same time, constraints such as service area distribution, power supply conditions, and cost ceiling are embedded to construct a simulation training environment. Secondly, conduct iterative training of the model: the agent outputs the initial values ​​of the decision variables in the simulation environment, simulates the charging service operation status under these values ​​through traffic simulation, and calculates the corresponding reward function value; based on the reward function value, the agent's strategy is updated using a deep reinforcement learning algorithm, and the values ​​of the decision variables are continuously iterated and adjusted until the comprehensive benefit reaches the preset threshold or the number of iterations meets the requirements; Finally, the full-cycle optimized layout scheme and flexible configuration strategy are output: From the decision variables after training convergence, the site construction identifiers of candidate locations in the unconstructed area are extracted to form a full-cycle optimized layout scheme, which clarifies the service area locations where charging stations need to be planned and constructed; The number of charging piles and the single pile capacity level of the existing sites are extracted, and combined with the fluctuation pattern of charging demand in the time and space, a flexible configuration strategy is formed to clarify the rules for adjusting the number of charging piles activated and the capacity of existing sites in different time periods.

[0013] Furthermore, the "vehicle-station" collaborative scheduling model in S4 is as follows: First, define the state space of the model, let the state space be... any state It is jointly represented by a vehicle status subset and a site status subset; the vehicle status subset includes the location, remaining battery power, and charging demand level of all new energy vehicles, and the site status subset includes the number of charging piles occupied, remaining available capacity, and current service load of all charging sites; the real-time update data of the status space comes from the collected new energy vehicle status and the facility operation status of each charging site. Secondly, define the action space of the model, let the action space be... any action It includes three core scheduling instructions: the optimal charging station matching recommendation instruction for new energy vehicles, the dynamic allocation instruction for charging piles within a station, and the cross-station traffic flow guidance instruction; the output of the actions needs to be adapted to the differences in charging demand in different areas and the scenarios of station load fluctuations. Finally, we define the model's reward function and training optimization method, assuming the reward function is... Its value is determined by three indicators: vehicle charging waiting time, station charging pile utilization rate, and regional traffic flow balance. The shorter the vehicle charging waiting time, the higher the station charging pile utilization rate, and the better the regional traffic flow balance, the larger the value of the return function. With the goal of maximizing the reward function, a reinforcement learning algorithm is used to train the model. Through continuous interaction between the agent and the real-time scheduling environment, the policy network parameters are iteratively updated until the scheduling instructions output by the model can match the vehicle charging demand with the station service capacity.

[0014] As a second aspect of the present invention, a smart optimization analysis system for highway new energy charging services is also provided, comprising: The multi-source data acquisition and database building unit is used to collect data from highway toll stations, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to build a structured highway travel big data set. The vehicle identification and demand mining unit is used to analyze the preprocessed mixed traffic flow data, extract multi-dimensional features of vehicles to distinguish new energy vehicles from fuel vehicles, and obtain basic information of new energy vehicles. Then, based on inverse reinforcement learning, a Markov decision process model is constructed. Combining the basic information and checkpoint data, the optimal reward function is fitted and the optimal trajectory strategy is solved to reconstruct the complete spatiotemporal trajectory of new energy vehicles and mine the spatiotemporal distribution characteristics of charging demand. The site optimization layout generation unit is used to construct a multi-objective optimization model based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, combined with the distribution of highway service areas, power supply conditions, and construction and operation cost constraints. The model aims to minimize the total charging time of users, maximize the utilization rate of facilities, and minimize the operating cost. The model is trained based on reinforcement learning and traffic simulation, and dynamically iterates the planning and site selection of charging sites in undeveloped areas, the number and capacity levels of facilities in existing sites, and outputs a full-cycle optimized layout scheme and flexible configuration strategy that adapts to the dynamic spatiotemporal changes of charging demand. The vehicle-to-station collaborative scheduling strategy generation unit is used to collect the status of new energy vehicles and the facility operation status of each charging station in real time, and to build a vehicle-to-station collaborative scheduling model based on reinforcement learning methods. In response to the differences in charging demand in different regions and the fluctuation of station load, it generates the optimal charging station matching recommendation for new energy vehicles, the dynamic allocation scheme of charging piles within the station, and the cross-station traffic flow guidance strategy to complete the adaptation scheduling of vehicle charging demand and station service capacity.

[0015] As a third aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which is executed by a processor as described in any one of the present invention, a method for intelligent optimization analysis of highway new energy charging services.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: 1. This invention provides an intelligent optimization analysis method for new energy charging services on highways. It collects multi-source data from highway toll stations, roadside monitoring videos, vehicle energy consumption, road network topology, and real-time traffic flow to construct a structured travel big data set, providing unified and correlated data support for subsequent analysis. After preprocessing mixed traffic flow data, it extracts the correlation features between vehicle appearance and energy consumption. Through feature mapping fusion and cosine similarity calculation, it accurately distinguishes between new energy vehicles and fuel vehicles, obtaining basic vehicle information. Then, based on inverse reinforcement learning, it constructs a Markov decision process model, combines checkpoint data to fit the optimal reward function, solves the optimal trajectory strategy, reconstructs the complete spatiotemporal trajectory of new energy vehicles, and mines the spatiotemporal distribution characteristics of charging demand. This solves the problems of low vehicle identification accuracy and inaccurate charging demand mining in traditional methods, providing a reliable demand basis for charging service optimization.

[0017] 2. This invention provides an intelligent optimization analysis method for new energy charging services on highways. By combining constraints such as highway service area distribution, power supply conditions, and construction and operation costs, a multi-objective optimization model is constructed with the objectives of minimizing total user charging time, maximizing facility utilization, and minimizing operating costs. The method uses candidate site identification, the number of charging piles, and the capacity level of a single pile as decision variables, and sets constraints such as service area distribution, power supply capacity, and cost ceilings. A training environment is built based on reinforcement learning and traffic simulation, using comprehensive benefits as the agent's reward function to continuously iterate and optimize the values ​​of decision variables. This process achieves dynamic adaptation between charging site planning and facility configuration, effectively balancing user experience, facility utilization efficiency, and construction and operation costs. The output full-cycle optimized layout scheme and flexible configuration strategy can accurately match the spatiotemporal dynamic changes in charging demand.

[0018] 3. This invention provides an intelligent optimization analysis method for highway new energy charging services. It collects real-time status data such as the location, remaining battery power, and charging demand level of new energy vehicles, as well as facility operation status data such as charging pile occupancy and service capacity at each charging station. Based on reinforcement learning, it constructs a "vehicle-station" collaborative scheduling model. This model uses vehicle and station states as the state space, station matching recommendation, dynamic charging pile allocation, and cross-station traffic flow guidance as the action space, and vehicle waiting time, station utilization rate, and traffic flow balance as the core reward function indicators. After training and optimization, it outputs a scheduling strategy. For scenarios involving differences in charging demand in different areas and fluctuations in station load, this model can generate precisely adapted scheduling schemes, achieving dynamic matching between vehicle charging demand and station service capacity, effectively alleviating charging congestion, and improving the overall efficiency and intelligence level of highway new energy charging services. Attached Figure Description

[0019] Figure 1 This is a flowchart of an intelligent optimization analysis method for highway new energy charging services according to an embodiment of the present invention; Figure 2 This is a system unit diagram of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 Please refer to Figure 1 This embodiment 1 provides an intelligent optimization analysis method for highway new energy charging services, including: S1. Collect highway toll station data, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to construct a structured highway travel big data set; S2. Analyze the preprocessed mixed traffic flow data to distinguish between new energy vehicles and fuel vehicles, and obtain basic information about new energy vehicles; then, construct a Markov decision process model based on inverse reinforcement learning, combine the basic information and checkpoint data, fit the optimal reward function and solve the optimal trajectory strategy, reconstruct the complete spatiotemporal trajectory of new energy vehicles, and explore the spatiotemporal distribution characteristics of charging demand. S3. Based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, and combined with the distribution of highway service areas, power supply conditions, and construction and operation cost constraints, a multi-objective optimization model is constructed with the goals of minimizing the total charging time of users, maximizing facility utilization, and minimizing operating costs. The model is trained based on reinforcement learning and traffic simulation, and the planning and site selection of charging stations in undeveloped areas and the number and capacity levels of facilities at existing stations are dynamically iterated. The model outputs a full-cycle optimized layout scheme and flexible configuration strategy that adapts to the dynamic spatiotemporal changes of charging demand. S4. Real-time collection of the status of new energy vehicles and the facility operation status of each charging station; construction of a "vehicle-station" collaborative scheduling model based on reinforcement learning method; generation of optimal charging station matching recommendations for new energy vehicles, dynamic allocation schemes of charging piles within stations, and cross-station traffic flow guidance strategies for different regional charging demand and station service capacity scenarios.

[0022] This embodiment 1 further elaborates on the above steps.

[0023] (1) Multi-source data acquisition and database construction Given the current industry context of supply and demand imbalance in highway new energy charging services and the need to improve resource allocation efficiency, in order to break the status quo of scattered and isolated multi-source data and inconsistent standards, it is necessary to systematically collect highway toll station data, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data, and build a structured highway travel big data set based on this.

[0024] This structured big data dataset is a multi-dimensional, interconnected dataset constructed according to unified data standards, spatiotemporal coordinate systems, and field formats. Each data module achieves unique cross-module association through three core key fields: vehicle unique identifier (ID), timestamp, and geographic location coordinates, ensuring data consistency and traceability. The vehicle passage basic data module focuses on vehicle travel trajectory information within the highway network, including fields such as vehicle unique identifier (ID), license plate number, passage time, passage location information, toll station number, roadside checkpoint number, corresponding latitude and longitude coordinates, vehicle initial type identifier, entrance / exit station number, and entrance / exit station name. The data in this module primarily originates from highway toll station passage records and roadside checkpoint capture records, providing a clear view of vehicle entry / exit information and road network travel paths.

[0025] The vehicle energy consumption characteristic data module sets up fields around the energy consumption and replenishment status of new energy vehicles, covering the vehicle's unique identifier ID, data collection time, remaining battery power, energy consumption rate per unit mileage, cumulative mileage, historical charging records, charging start time, charging end time, charging amount, and charging station number. The remaining battery power is represented by the battery's state of charge (SOC). The data in this module is collected synchronously through onboard sensors and the charging station operation system, which can accurately reflect the vehicle's energy consumption level and charging demand characteristics.

[0026] The road network topology and infrastructure basic data module is designed with fields for the highway network structure and charging infrastructure distribution. These fields include the unique segment number, actual segment length, segment design speed, starting and ending coordinates of the segment, unique service area number, service area geographical coordinates, unique number of existing charging stations, geographical coordinates of the stations, number of existing charging piles, single pile capacity level, and maximum power supply capacity of the stations. The data in this module is integrated with highway network planning archives and charging station operation and maintenance data, providing basic geographical and facility parameter support for subsequent charging station layout optimization.

[0027] The real-time traffic flow data module focuses on the dynamic monitoring of the road network's operational status. It sets fields such as unique road segment number, statistical time slice, road segment cross-sectional traffic flow, average driving speed of the road segment, road segment traffic density, and road segment congestion status indicator. The statistical time slice is set at five minutes per slice, and the congestion status indicator is divided into three categories: smooth, slow, and congested. The data of this module is collected in real time through roadside radar, video detectors and other equipment, which can dynamically reflect the traffic flow change pattern of the road network and provide real-time traffic condition basis for vehicle trajectory reconstruction and spatiotemporal distribution analysis of charging demand.

[0028] The roadside monitoring associated data module is built based on the intelligent analysis results of roadside monitoring videos. It includes fields such as the unique number of the monitoring equipment, video shooting time, structured storage path of key frame screenshots, pixel coordinates of vehicle targets in the picture, vehicle shape feature extraction results, vehicle length, vehicle width, vehicle height, and unique identification features. The data in this module is obtained by parsing the monitoring video through computer vision technology, which can supplement vehicle appearance feature information and provide data support for the accurate differentiation between new energy vehicles and fuel vehicles.

[0029] This multi-dimensional and highly correlated structured dataset integrates static basic information and dynamic energy consumption information of vehicle traffic, and links road network infrastructure layout and real-time traffic operation status. It provides comprehensive, reliable and highly correlated data support for subsequent accurate differentiation of vehicle types, reconstruction of spatiotemporal trajectories of new energy vehicles, mining of charging demand characteristics, optimization of charging station layout and generation of "vehicle-station" collaborative scheduling strategies.

[0030] (2) Vehicle identification and demand mining Furthermore, based on the previously constructed structured highway travel big data set, in-depth analysis was conducted on the preprocessed mixed traffic flow data. By extracting multi-dimensional features of vehicles, the distinction between new energy vehicles and fuel vehicles was effectively achieved, and basic information of new energy vehicles was obtained simultaneously.

[0031] The specific process is as follows: based on the preprocessed mixed traffic flow data, two types of core distinguishing data are extracted. The first type is the vehicle appearance identification data collected by roadside monitoring video, which directly identifies the vehicle's license plate format and exclusive new energy vehicle identification. The license plate format is determined according to the current motor vehicle license plate management regulations, and the exclusive new energy vehicle identification is confirmed according to the new energy exclusive symbol marked on the vehicle at the factory. The second category is vehicle energy replenishment record data collected at toll stations. Data showing charging activity or electric energy replenishment is filtered out. The two types of data are cross-validated. Vehicles that simultaneously meet one of the following three conditions are classified as new energy vehicles: a license plate format specifically for new energy vehicles, a unique new energy vehicle identification mark, and an energy replenishment type of electric energy replenishment. Vehicles that do not meet any of the three conditions are classified as gasoline-powered vehicles.

[0032] After classifying vehicles by type, basic information is extracted from the traffic flow data of new energy vehicles, including vehicle license plate information, travel time information, travel route information, and energy replenishment record information, forming a structured dataset of basic information of new energy vehicles, which provides data support for subsequent trajectory reconstruction and charging demand mining.

[0033] After obtaining basic information about new energy vehicles, a Markov decision process model is further constructed based on inverse reinforcement learning. Combining the basic information of new energy vehicles with data collected from highway roadside checkpoints, the optimal reward function is fitted and the optimal trajectory strategy is solved. This allows for the reconstruction of the complete spatiotemporal trajectory of new energy vehicles, ultimately revealing the spatiotemporal distribution characteristics of charging demand. The specific trajectory construction process is as follows: First, define the state space, action space, and state transition relationships of a Markov decision process: Let the state space be... Taking the highway segment as the core state unit, any state From the state vector Characterization, in which This is the only identifier for the road section. , The coordinates of the center of the road segment are latitude and longitude. This refers to the time it takes for a vehicle to pass through this section of road. The vehicle's speed on this section of road; Action space Defined as the set of path selection actions of a vehicle at a road segment node, i.e. ,in Indicates going straight. Indicates a left turn. Indicates a right turn; state transition probability Indicates the vehicle is in a certain state. Next action Then transferred to state The probability is determined based on the road network topology and the statistical characteristics of historical traffic flow paths. Secondly, construct and fit the optimal reward function, let the reward function be... Used to characterize the vehicle in state Execute action The subsequent benefits are constructed based on data including the passage time, location coordinates, and checkpoint capture records from the basic information of new energy vehicles, and are utilized through a deep neural network. Perform parameterized fitting on the reward function, i.e. ,in For neural network mapping functions, Network parameters; construct the objective function based on the maximum entropy criterion. ,in For parameterized strategies, It is the entropy function. Using entropy weights, the objective function is minimized using gradient descent. Iterative update parameters This continues until the objective function converges, yielding the optimal reward function. ; Finally, the optimal trajectory strategy is solved and the complete spatiotemporal trajectory is reconstructed using the optimal reward function. Based on this, an iterative strategy algorithm is used to solve for the optimal trajectory strategy. Combining the vehicle's unique identifier ID, travel time, and toll station and checkpoint capture records from the basic information of new energy vehicles, the optimal trajectory strategy is utilized. By performing path completion and temporal correlation on fragmented vehicle state nodes, a continuous state sequence of vehicles can be obtained. ,in This is the initial state. It is the final state and satisfies temporal continuity. This state sequence is the complete spatiotemporal trajectory of the new energy vehicle. By extracting the time of each state in the trajectory... with latitude and longitude coordinates This allows us to uncover the spatiotemporal distribution characteristics of charging demand.

[0034] (3) Generation of site optimization layout scheme After clarifying the spatiotemporal distribution characteristics of charging demand for new energy vehicles, the key to improving the quality of charging services and optimizing resource allocation efficiency lies in how to combine the existing infrastructure conditions and cost constraints of highways to formulate a charging station layout and facility configuration plan that adapts to the dynamic changes in demand.

[0035] Based on this, it is necessary to take the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction as the core basis, comprehensively consider the real constraints such as the distribution of highway service areas, power supply conditions and construction and operation costs, construct a multi-objective optimization model, and carry out model training through reinforcement learning and traffic simulation, dynamically iteratively optimize the planning and configuration parameters of charging stations, and finally output a full-cycle optimized layout scheme and flexible configuration strategy.

[0036] When constructing a multi-objective optimization model, the first step is to define the set of decision variables, which includes three core types of variables: First, the site construction identifier variable for candidate locations in undeveloped areas, used to determine whether each candidate location is planned for charging station construction; second, the number of charging piles configured at existing charging stations, taking positive integer values ​​corresponding to the specific number of charging piles installed; and third, the single-pile capacity level variable for existing charging stations, taking discrete positive values ​​corresponding to the power level specifications of the charging piles. Specifically, these correspond as follows: First, define the set of decision variables as follows: ,in It can be 0 or 1, used to represent the first element in the undeveloped area. Whether charging stations should be planned and built at each of the candidate locations. Corresponding to the planning and construction, No corresponding planning or construction; For the first to land The number of charging piles configured at each charging station is a positive integer. For the first to land The capacity level of a single charging pile at each charging station is a discrete positive value; here A unique identifier for a candidate location or an existing site. Establish status indicators for the site. This is an identifier for the number of charging stations. This is a capacity level identifier.

[0037] Secondly, focusing on the core requirements of optimizing charging services, a multi-objective optimization system is constructed with the goals of minimizing total user charging time, maximizing facility utilization, and minimizing operating costs as the core, and corresponding optimization functions are set for each objective.

[0038] The objective of minimizing the total user charging time is: ,in This represents the total number of new energy vehicles. Assign vehicle number, The total number of charging stations. It is the first Vehicles and the The shortest driving distance to each station It is the first The average speed of the car It is the first The vehicle in The average waiting time at each site is determined by the decision variables. Decide; The goal of maximizing facility utilization is ,in The total number of time periods. Number the time period. It is the first The site at the The actual number of vehicles charging during the cycle. It is the first Duration of the cycle; The goal of minimizing operating costs is ,in It is the first The site construction cost for each candidate location, This refers to the purchase and installation cost of a single charging station. It is the energy consumption cost per unit capacity of a single pile. It is the site operation and maintenance cost, and the decision variables. Positive correlation.

[0039] Finally, based on practical conditions, three types of constraints are set: service area distribution constraints clearly stipulate that charging stations can only be planned and built within the service area, ensuring the accessibility and compliance of the stations; power supply constraints limit the total power load of each station to no more than the upper limit of the power supply capacity of the area, ensuring the stable operation of the power system; and construction and operation cost constraints clearly stipulate that the total investment cost cannot exceed a preset upper limit, achieving a balance between economic benefits and service quality. In other words, the service area distribution constraints are... ,in Taking 0 or 1 represents the first... Are the candidate locations within the service area? The corresponding location is within the service area. The corresponding location is not within the service area; power supply constraints are: ,in It is the first The upper limit of power supply capacity in the area where each site is located; construction and operation cost constraints are... ,in It is the preset upper limit of total construction and operation costs.

[0040] The process of determining the full-cycle optimization layout scheme and flexible configuration strategy requires the construction of a complete training system based on reinforcement learning and traffic simulation. First, a training environment is built based on reinforcement learning and traffic simulation: taking the spatiotemporal characteristics of charging demand as input, the decision variables of the multi-objective optimization model are used as the action space of the agent, and the comprehensive benefits of the user's total charging time, facility utilization rate, and operating cost are used as the reward function of the agent. At the same time, constraints such as service area distribution, power supply conditions, and cost limits are embedded to build a simulation training environment. Secondly, conduct iterative training of the model: the agent outputs the initial values ​​of the decision variables in the simulation environment, simulates the charging service operation status under these values ​​through traffic simulation, and calculates the corresponding reward function value; based on the reward function value, the agent's strategy is updated using a deep reinforcement learning algorithm, and the values ​​of the decision variables are continuously iterated and adjusted until the comprehensive benefit reaches the preset threshold or the number of iterations meets the requirements; Finally, the full-cycle optimized layout scheme and flexible configuration strategy are output: From the decision variables after training convergence, the site construction identifiers of candidate locations in the unconstructed area are extracted to form a full-cycle optimized layout scheme, which clarifies the service area locations where charging stations need to be planned and constructed; The number of charging piles and the single pile capacity level of the existing sites are extracted, and combined with the fluctuation pattern of charging demand in the time and space, a flexible configuration strategy is formed to clarify the rules for adjusting the number of charging piles activated and the capacity of existing sites in different time periods.

[0041] (4) Generation of train-station collaborative scheduling strategy After completing the full-cycle optimization and flexible configuration of charging stations, in the face of real-time fluctuations in the charging demand of new energy vehicles in the highway network and dynamic changes in the service capacity of each charging station, relying solely on static layout schemes can no longer fully guarantee the efficient operation of charging services. Therefore, it is necessary to establish a dynamic "vehicle-station" collaborative scheduling mechanism.

[0042] Based on this, it is necessary to collect the status of new energy vehicles and the facility operation status of each charging station in real time, and build a "vehicle-station" collaborative scheduling model based on reinforcement learning methods. In response to the differences in charging demand in different regions and the fluctuation of station load, the model generates the optimal charging station matching recommendation for new energy vehicles, the dynamic allocation scheme of charging piles within the station, and the cross-station traffic flow guidance strategy, and finally completes the adaptation and scheduling of vehicle charging demand and station service capacity.

[0043] When constructing a "vehicle-station" collaborative scheduling model, the first step is to define the model's state space, denoted as . any state It is jointly represented by a vehicle status subset and a site status subset; the vehicle status subset includes the location, remaining battery power, and charging demand level of all new energy vehicles, and the site status subset includes the number of charging piles occupied, remaining available capacity, and current service load of all charging sites; the real-time update data of the status space comes from the collected new energy vehicle status and the facility operation status of each charging site. Secondly, define the action space of the model, let the action space be... any action It includes three core scheduling instructions: the optimal charging station matching recommendation instruction for new energy vehicles, the dynamic allocation instruction for charging piles within a station, and the cross-station traffic flow guidance instruction; the output of the actions needs to be adapted to the differences in charging demand in different areas and the scenarios of station load fluctuations. Finally, we define the model's reward function and training optimization method, assuming the reward function is... Its value is determined by three indicators: vehicle charging waiting time, station charging pile utilization rate, and regional traffic flow balance. The shorter the vehicle charging waiting time, the higher the station charging pile utilization rate, and the better the regional traffic flow balance, the larger the value of the return function. With the goal of maximizing the reward function, a reinforcement learning algorithm is used to train the model. During the training process, the intelligent system outputs scheduling actions based on the current state space, and then adjusts the strategy based on the actual effect after the actions are executed. After multiple iterations of optimization, the scheduling instructions output by the model can accurately match the vehicle charging demand with the station service capacity, ensuring the efficient and stable operation of new energy charging services on highways.

[0044] The intelligent optimization analysis method for new energy charging services proposed in this embodiment can effectively address pain points in the current industry, such as lagging charging demand perception, unreasonable site layout, and unbalanced resource allocation. It constructs a structured dataset through multi-source data fusion, enabling accurate identification and trajectory reconstruction of new energy vehicles, thereby uncovering the spatiotemporal distribution patterns of charging demand. Combined with a multi-objective optimization model and reinforcement learning simulation, the output charging site full-cycle layout scheme and flexible configuration strategy can fully adapt to the dynamic changes in highway traffic flow, providing traffic management departments and charging operators with a scientific basis for decision-making, helping to achieve precise planning and efficient operation and maintenance of charging infrastructure, and significantly reducing construction and operation costs.

[0045] In practical applications, the "vehicle-station" collaborative scheduling model constructed by this method can generate optimal station matching recommendations, dynamic allocation of charging piles, and cross-station traffic guidance strategies in real time, taking into account differences in charging demand and station load fluctuations in different regions. This effectively shortens the charging waiting time for new energy vehicles, improves the utilization efficiency of charging piles, and alleviates charging congestion during peak hours. With the continuous growth of the number of new energy vehicles, this method can be further extended to the national highway network, providing efficient and convenient charging services for long-distance travel of new energy vehicles, promoting the upgrading of the highway charging service system towards intelligence and refinement, and contributing to the green and low-carbon transformation and development of the transportation sector.

[0046] Example 2 Please refer to Figure 2 This embodiment 2 provides an intelligent optimization and analysis system for highway new energy charging services, including: The multi-source data acquisition and database building unit is used to collect data from highway toll stations, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to build a structured highway travel big data set. The vehicle identification and demand mining unit is used to analyze the preprocessed mixed traffic flow data, extract multi-dimensional features of vehicles to distinguish new energy vehicles from fuel vehicles, and obtain basic information of new energy vehicles. Then, based on inverse reinforcement learning, a Markov decision process model is constructed. Combining the basic information and checkpoint data, the optimal reward function is fitted and the optimal trajectory strategy is solved to reconstruct the complete spatiotemporal trajectory of new energy vehicles and mine the spatiotemporal distribution characteristics of charging demand. The site optimization layout generation unit is used to construct a multi-objective optimization model based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, combined with the distribution of highway service areas, power supply conditions, and construction and operation cost constraints. The model aims to minimize the total charging time of users, maximize the utilization rate of facilities, and minimize the operating cost. The model is trained based on reinforcement learning and traffic simulation, and dynamically iterates the planning and site selection of charging sites in undeveloped areas, the number and capacity levels of facilities in existing sites, and outputs a full-cycle optimized layout scheme and flexible configuration strategy that adapts to the dynamic spatiotemporal changes of charging demand. The vehicle-to-station collaborative scheduling strategy generation unit is used to collect the status of new energy vehicles and the facility operation status of each charging station in real time, and to build a vehicle-to-station collaborative scheduling model based on reinforcement learning methods. In response to the differences in charging demand in different regions and the fluctuation of station load, it generates the optimal charging station matching recommendation for new energy vehicles, the dynamic allocation scheme of charging piles within the station, and the cross-station traffic flow guidance strategy to complete the adaptation scheduling of vehicle charging demand and station service capacity.

[0047] Example 3 This embodiment 3 also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement any step of a smart optimization analysis method for highway new energy charging services.

[0048] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0049] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0050] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent optimization analysis of highway new energy charging services, characterized in that, include: S1. Collect highway toll station data, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to construct a structured highway travel big data set; S2. Analyze the preprocessed mixed traffic flow data to distinguish between new energy vehicles and fuel vehicles, and obtain basic information about new energy vehicles; then, construct a Markov decision process model based on inverse reinforcement learning, combine the basic information and checkpoint data, fit the optimal reward function and solve the optimal trajectory strategy, reconstruct the complete spatiotemporal trajectory of new energy vehicles, and explore the spatiotemporal distribution characteristics of charging demand. S3. Based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, and combined with the distribution of highway service areas, power supply conditions and construction and operation cost constraints, a multi-objective optimization model is constructed with the objectives of minimizing the total charging time of users, maximizing facility utilization, and minimizing operating costs. Model training is conducted based on reinforcement learning and traffic simulation. The planning and site selection of charging stations in undeveloped areas, the number and capacity levels of facilities at existing stations are dynamically iterated, and the output of full-cycle optimized layout schemes and flexible configuration strategies that adapt to the spatiotemporal dynamic changes in charging demand are generated. S4. Real-time collection of the status of new energy vehicles and the facility operation status of each charging station; construction of a "vehicle-station" collaborative scheduling model based on reinforcement learning method; generation of optimal charging station matching recommendations for new energy vehicles, dynamic allocation schemes of charging piles within stations, and cross-station traffic flow guidance strategies for different regional charging demand and station service capacity scenarios.

2. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The structured highway travel big data set in S1 is specifically a multi-dimensional associated dataset constructed according to a unified data standard, spatiotemporal coordinate system, and field format. Each data module achieves unique cross-module association through vehicle ID, timestamp, and geographic location coordinates. Its core structured data modules and corresponding fields are as follows: The vehicle passage basic data module includes the following fields: vehicle unique identifier ID, license plate number, passage time, passage location information, toll station number, roadside checkpoint number, corresponding latitude and longitude coordinates, vehicle initial type identifier, entrance and exit station number, and entrance and exit station name. The vehicle energy consumption characteristic data module includes the following fields: vehicle unique identifier ID, data collection time, remaining power value, energy consumption rate per unit mileage, cumulative mileage, historical charging records, charging start time, charging end time, charging amount, charging station number, and the remaining power value is represented by the SOC value. The road network topology and infrastructure basic data module includes the following fields: unique road segment number, actual road segment length, road segment design speed, road segment starting point latitude and longitude coordinates, road segment ending point latitude and longitude coordinates, service area unique number, service area geographical coordinates, existing charging station unique number, station geographical coordinates, existing number of charging piles, single pile capacity level, and station power supply capacity limit. The real-time traffic flow data module includes the following fields: unique road segment number, statistical time slice, traffic flow at the road segment cross section, average driving speed of the road segment, traffic density of the road segment, and congestion status indicator of the road segment. The statistical time slice is set in five minutes per slice, and the congestion status indicator is divided into three categories: smooth, slow, and congested. The roadside monitoring associated data module includes the following fields: unique ID of the monitoring device, video recording time, structured storage path of key frame screenshots, pixel coordinates of the vehicle target in the image, vehicle shape feature extraction results, vehicle length, vehicle width, vehicle height, and unique identification features.

3. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The process in S2 of analyzing the preprocessed mixed traffic flow data to distinguish between new energy vehicles and fuel vehicles, and obtaining basic information about new energy vehicles, is as follows: Based on the preprocessed mixed traffic flow data, two types of core distinguishing data are extracted. The first type is vehicle appearance identification data collected by roadside monitoring video, which directly identifies the vehicle's license plate format and exclusive new energy vehicle logo. The license plate format is determined according to the current motor vehicle license plate management regulations, and the exclusive new energy vehicle logo is confirmed according to the new energy exclusive symbol marked on the vehicle at the factory. The second category is vehicle energy replenishment record data collected at toll stations, which filters out vehicle data that shows charging behavior or energy replenishment. Cross-validate the two types of data, and determine that a vehicle is a new energy vehicle if it meets one of the following three conditions: the license plate format is a special format for new energy vehicles, it has a unique new energy logo, and its energy supply type is electric energy supply. Vehicles that do not meet all three conditions are classified as gasoline-powered vehicles. After classifying vehicles by type, basic information is extracted from the traffic flow data of new energy vehicles, including vehicle license plate information, travel time information, travel route information, and energy replenishment record information, forming a structured dataset of basic information of new energy vehicles, which provides data support for subsequent trajectory reconstruction and charging demand mining.

4. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The process of constructing the complete spatiotemporal trajectory of the new energy vehicle in S2 is as follows: First, define the state space, action space, and state transition relationships of a Markov decision process: Let the state space be... Taking the highway segment as the core state unit, any state From the state vector Characterization, in which This is the only identifier for the road section. , The coordinates of the center of the road segment are latitude and longitude. This refers to the time it takes for a vehicle to pass through this section of road. The vehicle's speed on this section of road; Action space Defined as the set of path selection actions of a vehicle at a road segment node, i.e. ,in Indicates going straight. Indicates a left turn. Indicates a right turn; state transition probability Indicates the vehicle is in a certain state. Next action Then transferred to state The probability is determined based on the road network topology and the statistical characteristics of historical traffic flow paths. Secondly, construct and fit the optimal reward function, let the reward function be... Used to characterize the vehicle in state Execute action The subsequent benefits are constructed based on data including the passage time, location coordinates, and checkpoint capture records from the basic information of new energy vehicles, and are utilized through a deep neural network. Perform parameterized fitting on the reward function, i.e. ,in For neural network mapping functions, For network parameters; Constructing an objective function based on the maximum entropy criterion ,in For parameterized strategies, It is the entropy function. Using entropy weights, the objective function is minimized using gradient descent. Iterative update parameters This continues until the objective function converges, yielding the optimal reward function. ; Finally, the optimal trajectory strategy is solved and the complete spatiotemporal trajectory is reconstructed using the optimal reward function. Based on this, an iterative strategy algorithm is used to solve for the optimal trajectory strategy. Combining the vehicle's unique identifier ID, travel time, and toll station and checkpoint capture records from the basic information of new energy vehicles, the optimal trajectory strategy is utilized. By performing path completion and temporal correlation on fragmented vehicle state nodes, a continuous state sequence of vehicles can be obtained. ,in This is the initial state. It is the final state and satisfies temporal continuity. This state sequence is the complete spatiotemporal trajectory of the new energy vehicle. By extracting the time of each state in the trajectory... with latitude and longitude coordinates This allows us to uncover the spatiotemporal distribution characteristics of charging demand.

5. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The multi-objective optimization model in S3 is specifically as follows: First, define the set of decision variables as follows: ,in It can be 0 or 1, used to represent the first element in the undeveloped area. Whether charging stations should be planned and built at each of the candidate locations. Corresponding to the planning and construction, No corresponding planning or construction; For the first to land The number of charging piles configured at each charging station is a positive integer. For the first to land The capacity level of a single charging pile at each charging station is a discrete positive value; here A unique identifier for a candidate location or an existing site. Establish status indicators for the site. This is an identifier for the number of charging stations. For capacity level identification; Secondly, a multi-objective optimization system is constructed with the core objectives of minimizing the total charging time for users, maximizing facility utilization, and minimizing operating costs, and corresponding objective optimization functions are set for each objective. Finally, the constraints are set as follows: Service area distribution constraints are... ,in Taking 0 or 1 represents the first... Are the candidate locations within the service area? The corresponding location is within the service area. The corresponding location is not within the service area; power supply constraints are: ,in It is the first The upper limit of power supply capacity in the area where each site is located; construction and operation cost constraints are... ,in It is the preset upper limit of total construction and operation costs.

6. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The objective optimization functions are as follows: The objective is to minimize the total user charging time. ,in This represents the total number of new energy vehicles. Assign vehicle number, The total number of charging stations. It is the first Vehicles and the The shortest driving distance to each station It is the first The average speed of the car It is the first The vehicle in The average waiting time at each site is determined by the decision variables. Decide; The goal of maximizing facility utilization is ,in The total number of time periods. Number the time period. It is the first The site at the The actual number of vehicles charging during the cycle. It is the first Duration of the cycle; The goal of minimizing operating costs is ,in It is the first The site construction cost for each candidate location, This refers to the purchase and installation cost of a single charging station. It is the energy consumption cost per unit capacity of a single pile. It is the site operation and maintenance cost, and the decision variables. Positive correlation.

7. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The process for determining the full-cycle optimized layout scheme and flexible configuration strategy in S3 is as follows: First, a training environment is built based on reinforcement learning and traffic simulation: taking the spatiotemporal characteristics of charging demand as input, the decision variables of the multi-objective optimization model are used as the action space of the agent, and the comprehensive benefits of the user's total charging time, facility utilization rate, and operating cost are used as the reward function of the agent. At the same time, constraints such as service area distribution, power supply conditions, and cost ceiling are embedded to construct a simulation training environment. Secondly, conduct iterative training of the model: the agent outputs the initial values ​​of the decision variables in the simulation environment, simulates the charging service operation status under these values ​​through traffic simulation, and calculates the corresponding reward function value; Based on the reward function value, a deep reinforcement learning algorithm is used to update the agent's strategy and continuously iterate to adjust the values ​​of decision variables until the overall benefit reaches a preset threshold or the number of iterations meets the requirements. Finally, the full-cycle optimized layout scheme and flexible configuration strategy are output: From the decision variables after training convergence, the site construction identifiers of candidate locations in the unconstructed area are extracted to form a full-cycle optimized layout scheme, which clarifies the service area locations where charging stations need to be planned and constructed; The number of charging piles and the single pile capacity level of the existing sites are extracted, and combined with the fluctuation pattern of charging demand in the time and space, a flexible configuration strategy is formed to clarify the rules for adjusting the number of charging piles activated and the capacity of existing sites in different time periods.

8. The intelligent optimization analysis method for highway new energy charging services according to claim 1, characterized in that, The "vehicle-station" collaborative scheduling model in S4 is as follows: First, define the state space of the model, let the state space be... any state It is jointly represented by a vehicle status subset and a site status subset; the vehicle status subset includes the location, remaining battery power, and charging demand level of all new energy vehicles, and the site status subset includes the number of charging piles occupied, remaining available capacity, and current service load of all charging sites; the real-time update data of the status space comes from the collected new energy vehicle status and the facility operation status of each charging site. Secondly, define the action space of the model, let the action space be... any action It includes three core scheduling instructions: the optimal charging station matching recommendation instruction for new energy vehicles, the dynamic allocation instruction for charging piles within a station, and the cross-station traffic flow guidance instruction; the output of the actions needs to be adapted to the differences in charging demand in different areas and the scenarios of station load fluctuations. Finally, we define the model's reward function and training optimization method, assuming the reward function is... Its value is determined by three indicators: vehicle charging waiting time, station charging pile utilization rate, and regional traffic flow balance. The shorter the vehicle charging waiting time, the higher the station charging pile utilization rate, and the better the regional traffic flow balance, the larger the value of the return function. With the goal of maximizing the reward function, a reinforcement learning algorithm is used to train the model. Through continuous interaction between the agent and the real-time scheduling environment, the policy network parameters are iteratively updated until the scheduling instructions output by the model can match the vehicle charging demand with the station service capacity.

9. A smart optimization analysis system for highway new energy charging services, characterized in that, include: The multi-source data acquisition and database building unit is used to collect data from highway toll stations, roadside monitoring video data, vehicle energy consumption data, road network topology data, and real-time traffic flow data to build a structured highway travel big data set. The vehicle identification and demand mining unit is used to analyze the preprocessed mixed traffic flow data, extract multi-dimensional features of vehicles to distinguish new energy vehicles from fuel vehicles, and obtain basic information of new energy vehicles. Then, based on inverse reinforcement learning, a Markov decision process model is constructed. Combining the basic information and checkpoint data, the optimal reward function is fitted and the optimal trajectory strategy is solved to reconstruct the complete spatiotemporal trajectory of new energy vehicles and mine the spatiotemporal distribution characteristics of charging demand. The site optimization layout scheme generation unit is used to construct a multi-objective optimization model based on the spatiotemporal characteristics of charging demand obtained from trajectory reconstruction, combined with the distribution of highway service areas, power supply conditions and construction and operation cost constraints, with the objectives of minimizing the total charging time of users, maximizing facility utilization, and minimizing operating costs. Model training is conducted based on reinforcement learning and traffic simulation. The planning and site selection of charging stations in undeveloped areas, the number and capacity levels of facilities at existing stations are dynamically iterated, and the output of full-cycle optimized layout schemes and flexible configuration strategies that adapt to the spatiotemporal dynamic changes in charging demand are generated. The vehicle-to-station collaborative scheduling strategy generation unit is used to collect the status of new energy vehicles and the facility operation status of each charging station in real time, and to build a vehicle-to-station collaborative scheduling model based on reinforcement learning methods. In response to the differences in charging demand in different regions and the fluctuation of station load, it generates the optimal charging station matching recommendation for new energy vehicles, the dynamic allocation scheme of charging piles within the station, and the cross-station traffic flow guidance strategy to complete the adaptation scheduling of vehicle charging demand and station service capacity.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor according to any one of claims 1-8, which describes a method for intelligent optimization analysis of new energy charging services on highways.

Images