A new energy vehicle charging service system and method

Abstract

The application relates to the field of automobile charging technology services, in particular to a new energy automobile charging service system and method; the system comprises a data access perception fusion unit used for collecting multi-source Internet of Things perception data and user charging demand data; a pile site adaptation graph judgment unit fuses a knowledge graph construction method, a visual recognition analysis method and a three-dimensional space reconstruction method, constructs a ternary pile vehicle relationship graph, and outputs a credible charging pile set and an unreachable pile site filtering list; a path strategy intelligent pile insertion unit constructs a path strategy graph based on the user charging demand data, generates a candidate charging path and a candidate charging pile through a pile insertion strategy decision mechanism; and a power load evolution judgment unit is used for screening to obtain a suitable charging pile and a charging pile path. The application realizes a new energy automobile charging service system with charging pile space adaptability identification, false pile site elimination, path level pile insertion recommendation and power response prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile charging technology services, in particular to a new energy automobile charging service system and method. BACKGROUND

[0002] With the continuous rise of the number of new energy vehicles, the density and intelligence level of urban charging infrastructure are gradually becoming key factors affecting the travel experience and charging efficiency of new energy vehicle users. At present, most smart city energy service platforms have supported basic charging pile distribution query and path navigation functions, and some platforms have realized charging state monitoring, charging reservation and other services through the Internet of Vehicles interface. However, in peak travel, unfamiliar city scenarios and urban heterogeneous pile network environments, users often face actual problems such as unavailability of charging piles, false failures, physical space incompatibility, high path time consumption or insufficient power of target piles, resulting in a decline in charging experience and even travel interruption, which seriously affects the promotion and use of new energy vehicles;

[0003] The existing technology mainly has the following key problems: first, the urban new energy vehicle charging system generally lacks an automatic evaluation mechanism for the space adaptability of charging piles, making it difficult to determine whether the vehicle can physically reach the target pile, especially in the presence of obstacles or insufficient size in front of the parking space, resulting in a large number of false available information; second, there is a lack of ability to combine user travel intentions with dynamic pile insertion strategies, resulting in recommended results that do not match actual travel needs; third, the power response capability of the target pile is mostly static labels or real-time information queries, which cannot be used for future power prediction and path power comprehensive analysis; the above problems limit the efficient and trustworthy intelligent charging path decision-making of new energy vehicle users in complex urban environments. SUMMARY

[0004] The purpose of the present application is to provide a new energy automobile charging service system and method to solve the three types of problems mentioned in the background.

[0005] To achieve the above purpose, the purpose of the present application is to provide a new energy automobile charging service system, comprising:

[0006] A data access perception fusion unit is used to collect multi-source Internet of Things perception data and user charging demand data, and to perform standard structured processing;

[0007] A pile site adaptation graph judgment unit is used to construct a ternary pile-vehicle relationship graph containing vehicle nodes, pile nodes and user behavior nodes based on multi-source Internet of Things perception data and user charging demand data, and to output a set of reliable charging piles and a list of unreachable pile sites by fusing knowledge graph construction methods, visual recognition analysis methods and three-dimensional space reconstruction methods.

[0008] The path strategy intelligent plug-in unit constructs a path strategy graph based on user charging demand data, generates a candidate charging path and a candidate charging pile through a plug-in strategy decision mechanism, and transmits the candidate charging path and the candidate charging pile to the power load evolution judgment unit.

[0009] The power load evolution judgment unit is used to construct a charging time evolution graph in a city area, predict the future power response rate of the candidate charging pile in each candidate path by using a graph neural network model, obtain a path power response matrix, and finally select a suitable charging pile and a charging pile path.

[0010] Preferably, in the data access perception fusion unit, the multi-source Internet of Things perception data includes vehicle terminal data, charging pile equipment data, geomagnetic sensor data, parking space image spatial layout data, regional power supply information data, and charging pile operating state; and the user charging demand data includes vehicle type parameters, remaining power, destination information, path delay tolerance, and on-the-way intention.

[0011] Preferably, the pile site adaptation graph judgment unit includes a three-dimensional space reconstruction module.

[0012] The three-dimensional space reconstruction module is used to extract parking edge features and ground guide line coordinates based on parking space image spatial layout data, construct a three-dimensional space point cloud model of the parking space, and judge whether a target vehicle can physically enter a target pile front parking space range by geometric matching operation in combination with vehicle type parameters to judge the parking space adaptability of the charging parking space.

[0013] Preferably, the ternary pile-vehicle relationship graph is used to depict the spatial adaptation relationship, state credibility relationship, and user preference behavior mode between new energy vehicles and city charging piles.

[0014] The nodes of the ternary pile-vehicle relationship graph include vehicle nodes, charging pile nodes, and user behavior nodes; the vehicle nodes include vehicle length, width, height, charging interface type, and minimum turning radius attributes; the charging pile nodes include charging pile size, cable reachable distance, fault state, and historical connection frequency attributes; and the user behavior nodes include user score records, complaint records, and successful charging times.

[0015] The edges of the ternary pile-vehicle relationship graph include physical adaptation edges connecting vehicle nodes and charging pile nodes, used to express the reachability relationship between vehicles and piles; state feedback edges connecting charging pile nodes and user behavior nodes, used to express the pile site state credibility relationship; and use preference edges connecting vehicle nodes and user behavior nodes, used to express the user historical behavior constraint relationship.

[0016] Preferably, the pile site adaptation graph judgment unit comprises a charging pile screening module, which is configured to output a trusted charging pile set and an unreachable pile site filtering list;

[0017] The trusted charging pile set is a set of available charging piles that meet physical accessibility, functional availability, and use state credibility. The unreachable pile site filtering list is a set of charging piles that cannot be physically accessed by the current vehicle, have unavailable pile body states, and have false available states.

[0018] The output trusted charging pile set and the unreachable pile site filtering list are obtained by the following method:

[0019] S2.1, a trusted scoring function of the charging pile node is constructed based on graph reasoning and graph convolutional neural network, and the physical adaptation edge and the state feedback edge of the ternary pile-car relationship graph are jointly reasoned and calculated to evaluate the use state credibility of each charging pile. The trusted charging pile set is obtained from the set of charging pile nodes with a use state credibility higher than a preset threshold.

[0020] S2.2, the unreachable pile site filtering list is output from the set of charging pile nodes that are physically not accessible and have false available states.

[0021] Preferably, the path strategy intelligent pile insertion unit comprises a path strategy graph construction module.

[0022] The path strategy graph construction module is configured to construct a path strategy graph based on user charging demand data. The path strategy graph comprises a main path node, an on-route charging pile node, and a strategy control factor edge.

[0023] The main path node represents a node in the shortest path segment between the departure point and the destination. The on-route charging pile node represents a charging pile position that is reachable from the main path under the constraint of the user's on-route intention. The strategy control factor edge is used to represent the path extension adjustment relationship between the main path node and the on-route charging pile node, and includes a path time delay cost relationship edge, a charging waiting cost relationship edge, and a user preference matching relationship edge.

[0024] Preferably, the path strategy intelligent pile insertion unit further comprises a path sorting module. The path sorting module performs charging pile scoring and sorting on each candidate path in the path strategy graph through an insertion strategy decision mechanism to generate a candidate charging path and a candidate charging pile.

[0025] The insertion strategy decision mechanism is as follows:

[0026] The plug-in strategy decision mechanism calculates charging pile scores through a path score function, and sorts the charging piles according to the charging pile scores from large to small, and the charging piles with the top five charging pile scores and consistent with the en route charging pile node are taken as candidate charging piles; and the path from the main path node to the candidate charging pile is taken as a candidate charging path.

[0027] Preferably, the power load evolution judgment unit comprises a charging time evolution analysis module; the charging time evolution analysis module constructs a charging time evolution graph taking the charging pile node as a graph node and the power transmission edge as a graph edge, and the charging time evolution graph is used to describe the time sequence evolution relationship of the power load state between the charging piles in the urban area.

[0028] Preferably, the power transmission edge comprises a timestamp, a power change rate and a geographical proximity factor.

[0029] Preferably, the power load evolution judgment unit further comprises a power prediction path decision module; the power prediction path decision module calculates the future power response rate of all candidate charging piles in each candidate path based on the charging time evolution graph using a time sequence graph neural network model, constructs a path power response matrix according to the future power response rate of all pile sites in the candidate path, and finally determines the path charging pile and charging path set meeting the user demand based on the path power response matrix.

[0030] Preferably, the time sequence graph neural network model is a graph neural network model based on a time dynamic evolution graph, and is used to analyze and predict the charging power evolution trend of all candidate charging piles in each candidate path in each urban area.

[0031] Preferably, the future power response rate is an average available power value in a prediction time window, and the high and low of the future power response rate represents the strength of the power supply capacity of the charging pile in the future period; and the path power response matrix is a two-dimensional matrix structure composed of the future power response rates of all candidate charging pile nodes in all candidate paths, and is used to express the charging power supply level of all candidate paths as a whole.

[0032] On the other hand, the application provides a new energy vehicle charging service method, which is used for the new energy vehicle charging service system described above, and comprises the following steps:

[0033] S10.1, collecting multi-source Internet of Things sensing data and user charging demand data, and performing standard structured processing;

[0034] S10.2, based on the multi-source Internet of Things sensing data and the user charging demand data, fusing a knowledge graph construction method, a visual recognition analysis method and a three-dimensional space reconstruction method, constructing a ternary pile-car relationship graph containing vehicle nodes, pile site nodes and user behavior nodes, and outputting a trusted charging pile set and an unreachable pile site filtering list;

[0035] S10.3, constructing a path strategy graph based on user charging demand data, and generating a candidate charging path and a candidate charging pile through a plug-in strategy decision mechanism;

[0036] S10.4, constructing a charging time evolution graph in a city area, predicting a future power response rate of the candidate charging pile in each candidate path by using a graph neural network model, obtaining a path power response matrix, and finally screening a suitable charging pile and a path reaching the charging pile.

[0037] Compared with the prior art, the above technical scheme of the present application has the following beneficial technical effects:

[0038] 1. In the present application, the ternary pile-car relationship graph constructed by fusing the knowledge graph construction method, the visual recognition analysis method and the three-dimensional space reconstruction method can automatically identify the physical adaptability of each charging pile and the target vehicle in space, and identify the false available state in combination with historical state feedback information, so as to output a credible charging pile set and an unreachable pile position filtering list, effectively improving the authenticity and availability of the charging pile recommendation information.

[0039] 2. In the present application, by constructing the path strategy graph and the plug-in strategy decision mechanism, and combining the charging time evolution graph and the path power response matrix, the graph neural network prediction of the future power response rate of each charging pile in the candidate path is realized, the path charging pile and the charging path set satisfying the user's on-the-way intention and the power sufficient condition can be screened, and the best charging demand of the user in the peak period is met. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The principle block diagram of an embodiment proposed by the present application is shown in the figure;

[0041] Reference signs: 1, data access perception fusion unit; 2, pile position adaptation graph judgment unit; 21, three-dimensional space reconstruction module; 22, charging pile screening module; 3, path strategy intelligent plug-in unit; 31, path strategy graph construction module; 32, path sorting module; 4, power load evolution judgment unit; 41, charging time evolution analysis module; 42, power prediction path decision module. DETAILED DESCRIPTION

[0042] Embodiment one, as shown in the figure, provides a new energy vehicle charging service system, which comprises: Figure 1

[0043] The data access perception fusion unit 1 is used for collecting multi-source Internet of Things perception data and user charging demand data, and performing standard structured processing.

[0044] ​In this embodiment, the multi-source Internet of Things perception data in the data access perception fusion unit 1 includes vehicle terminal data, charging pile equipment data, geomagnetic sensor data, parking space image spatial layout data, regional power supply information data, and charging pile operating state; and the user charging demand data includes vehicle type parameters, residual power, destination information, path delay tolerance, and on-the-way intention.

[0045] In this embodiment, the vehicle terminal data is obtained through a vehicle-mounted T-Box or OBD interface, and is uploaded to a service platform through a V2X protocol or a cellular network. The vehicle terminal data content includes current SOC, vehicle speed, residual endurance, VIN code, and real-time positioning of the vehicle. The charging pile equipment data is collected by a pile-side BMS module or OCPP communication interface, and is connected with the platform or edge gateway through Ethernet / Wi-Fi / NB-IoT, and is dynamically reported every 30 seconds. The charging pile equipment data content includes current pile voltage, current, output power, gun mouth idle state, communication state, and fault code information. The geomagnetic sensor data is monitored by a parking space ground embedded geomagnetic sensor to judge whether the parking space is occupied by monitoring the change of magnetic flux disturbance, and is reported in real time. The parking space image spatial layout data is obtained by shooting through a high-definition camera installed above and beside the charging pile. The parking space image spatial layout data content includes parking space image, obstacle position, ground guide line, and relative spatial scale point cloud. The regional power supply information data is collected by connecting with a city power grid monitoring system API, including transformer load rate, regional instantaneous power capacity, allowed load peak value, and time period electricity price information. The charging pile operating state is collected through platform OCPP interface interaction, including whether the charging pile is online, whether it is faulty, whether it is in maintenance, historical use frequency, and software version number. The destination information is input by the user in the APP or is obtained synchronously through navigation. The path delay tolerance is set by the user through platform preference, whether to accept detour or wait. The on-the-way intention is analyzed by the platform through historical behavior or instant inquiry of user intention, such as whether to accept on-the-way transfer charging, whether to prefer specific area charging, etc.

[0046] The pile site adaptation graph judgment unit 2 is based on multi-source Internet of Things perception data and user charging demand data, fuses knowledge graph construction method, visual recognition analysis method, and three-dimensional space reconstruction method, constructs a ternary pile-car relationship graph containing vehicle nodes, pile site nodes, and user behavior nodes, and outputs a credible charging pile set and an unreachable pile site filtering list.

[0047] In this embodiment, the pile site adaptation graph judgment unit 2 includes a three-dimensional space reconstruction module 21.

[0048] The three-dimensional space reconstruction module 21 is configured to extract a parking space edge feature and a ground guide line coordinate based on the parking space image space layout data, construct a three-dimensional space point cloud model of the parking space, and determine whether the target vehicle can physically enter the range of the target pile front parking space by geometric matching operation in combination with the vehicle type parameter, to determine the space adaptability of the charging parking space.

[0049] In the embodiment, the three-dimensional space reconstruction module 21 is connected to a high-definition camera device arranged around the charging pile. The camera device is arranged at a top view angle or a side view angle, and obtains the parking space image space layout data. The parking space image included in the parking space image space layout data is subjected to a pre-processing step such as distortion correction, illumination normalization, and image enhancement, to obtain a clear input image frame. An edge detection algorithm is applied to extract the contour information of the left and right boundary lines, the rear limit line, and the ground parking auxiliary line. Meanwhile, the object contour such as the shelter and the isolation pile in the image is detected. The image is subjected to perspective correction by using a preset camera parameter, and a transformation matrix between the camera image plane and the actual coordinate system of the ground is established. A homography matrix is used to map the two-dimensional coordinate points in the image to a three-dimensional point set in the real physical space. The three-dimensional space reconstruction module 21 reconstructs a three-dimensional point cloud model of the parking space region in the space coordinate system. The point cloud model includes the actual length, width, effective depth, obstacle position, and shape of the parking space, and other space information. The system calls the vehicle type parameter. The three-dimensional space reconstruction module 21 projects the vehicle shape into the parking space point cloud model, and executes a geometric matching algorithm to determine whether the vehicle model can smoothly enter the current parking space without collision and boundary crossing. If there is a collision-free trajectory solution for the parking path of the vehicle model in the point cloud model, and the remaining turning space is greater than a set safety tolerance of 30 cm, it is determined that the parking space is “parkable” for the vehicle type. Otherwise, the parking space is marked as “not suitable for the parking space”.

[0050] In the embodiment, the ternary pile-car relationship graph is used to depict the space adaptation relationship, state credibility relationship, and user preference behavior mode between the new energy vehicle and the urban charging pile.

[0051] The nodes of the ternary pile-car relationship graph include a vehicle node, a charging pile node, and a user behavior node. The vehicle node includes a vehicle length, a width, a height, a charging interface type, and a minimum turning radius attribute. The charging pile node includes a charging pile size, a cable reachable distance, a fault state, and a historical connection frequency attribute. The user behavior node includes a user score record, a complaint record, and a successful charging number.

[0052] The edges of the ternary pile-car relationship graph include a physical adaptation edge connecting the vehicle node and the charging pile node, which is used to express the reachability relationship between the vehicle and the pile. A state feedback edge connecting the charging pile node and the user behavior node, which is used to express the pile state credibility relationship. A use preference edge connecting the vehicle node and the user behavior node, which is used to express the user historical behavior constraint relationship.

[0053] In the embodiment, the pile site adaptation graph judgment unit 2 comprises a charging pile screening module 22, which is used to output a trusted charging pile set and an unreachable pile site filtering list;

[0054] The trusted charging pile set is a set of available charging piles that meet physical parking availability, functional availability, and use state credibility; the unreachable pile site filtering list is a set of charging piles that cannot be physically accessed by the current vehicle, whose pile body state is unavailable, and whose state is false;

[0055] The output trusted charging pile set and the unreachable pile site filtering list are specifically as follows:

[0056] S2.1, based on graph reasoning and graph convolutional neural network, a trusted scoring function of the charging pile node is constructed, and the physical adaptation edge and the state feedback edge of the ternary pile-car relationship graph are jointly inferred and calculated to evaluate the use state credibility of each charging pile, and the trusted charging pile set is obtained in the charging pile node set whose use state credibility is higher than a preset threshold;

[0057] S2.2, output the unreachable pile site filtering list in the charging pile node set that is physically not parkable and in a false available state.

[0058] In the embodiment, the physical parking availability means that after analyzing the charging pile parking space through a three-dimensional point cloud reconstruction model, it is confirmed that the target vehicle can be parked without collision and boundary crossing, and the plug-in charging gun operation has spatial freedom; the functional availability means that the current charging pile operating state is "online", and is not in maintenance, downtime or user occupancy state; the use state credibility is based on the charging success rate, failure rate and abnormal reporting in the historical access data to judge whether the pile site state credibility meets the standard;

[0059] In the embodiment S2.1, the trusted scoring function is used to obtain the structured edge features from the vehicle node and the user node based on the physical adaptation edge and the state feedback edge of the ternary pile-car relationship graph, and to calculate the use state credibility of the pile node in the spatial adaptability and the historical use state two dimensions by aggregating and weighting the fusion through the edge feature coding function, which is used to comprehensively evaluate whether the charging pile is currently truly available, and finally used as a screening trusted charging pile set; the trusted scoring function is specifically as follows:

[0060]

[0061] Wherein, T p is the use state credibility of the charging pile node v p , ranging from (0, 1), representing the "trusted availability degree" of the charging pile in the current environment, which is used for subsequent decision whether to recommend as a trusted candidate charging pile; vp is a charging pile node p; p is a charging pile node index; σ() is a Sigmoid activation function, used to normalize the comprehensive score to the interval (0, 1); W1 is a physical adaptation edge aggregation weight, representing the importance degree of the influence of the system on the vehicle space adaptation relationship on the credibility; W2 is a state feedback edge aggregation weight, representing the importance of the influence of the user state feedback on the credibility; b is a bias term, representing the initial offset of the system credibility score; is a charging pile node v p is a set of all vehicle nodes that exist physical adaptation edges; is a charging pile node v p is a set of all user nodes that exist state feedback edges; is a vehicle node v c and a charging pile node v p physical adaptation edge; is a physical adaptation edge feature function, used to map the spatial adaptation relationship edge between the vehicle node v c and the charging pile node v p to a numerical score; is a user node v p and a charging pile node v p state feedback edge; is a state feedback edge feature function, used to map the state feedback edge between the user node v p and the charging pile node v p to a numerical score.

[0062] The path strategy intelligent plug-in unit 3 constructs a path strategy graph based on user charging demand data, generates a candidate charging path and a candidate charging pile through a plug-in strategy decision mechanism, and transmits the candidate charging path and the candidate charging pile to the power load evolution judgment unit 4.

[0063] In this embodiment, the path strategy intelligent plug-in unit 3 includes a path strategy graph construction module 31.

[0064] The path strategy graph construction module 31 is configured to construct a path strategy graph based on user charging demand data; the path strategy graph includes a main path node, a route charging pile node, and a strategy control factor edge.

[0065] The main path node represents a node in the shortest path segment of the starting point and the destination; the route charging pile node represents a charging pile position reachable from the main path under the user route intention constraint; the strategy control factor edge is used to represent the path extension adjustment relationship between the main path node and the route charging pile node, and includes a path time delay cost relationship edge, a charging waiting cost relationship edge, and a user preference matching relationship edge.

[0066] In the embodiment, the user on-the-way intention constraint refers to a constraint condition that the user sets when proposing a charging demand, i.e., a maximum tolerance degree constraint condition for not deviating from the current navigation main path, which can be set by the user or automatically estimated and generated by the system according to the user departure time, estimated arrival time and available time window. The numerical form of the user on-the-way intention constraint is: maximum acceptable detour time, maximum acceptable detour distance and whether to accept a reverse direction charging pile travel path.

[0067] The strategy control factor is a set of multi-dimensional constraint parameters attached to the path insertion edge between the main path node and the on-the-way charging pile node, which is used to guide the system to comprehensively consider the time efficiency, waiting cost and personalized matching degree in the process of inserting the pile, and control whether the final candidate path meets the user preference and constraint. The strategy control factor includes:

[0068] The path time delay cost relationship edge is used to describe the additional time delay cost of the user's overall path after inserting a certain candidate charging pile into the path. The charging waiting cost relationship edge represents the estimated waiting time cost at the current candidate charging pile. The user preference matching relationship edge is used to reflect whether a certain candidate charging pile meets the user demand.

[0069] In the embodiment, the path strategy intelligent pile insertion unit 3 further includes a path sorting module 32. The path sorting module 32 performs charging pile scoring and sorting on each candidate path in the path strategy graph through a pile insertion strategy decision mechanism, and generates a candidate charging path and a candidate charging pile.

[0070] Specifically, the pile insertion strategy decision mechanism is as follows:

[0071] The pile insertion strategy decision mechanism calculates the charging pile score through a path scoring function, and sorts the charging piles in descending order of the charging pile score. The top five charging piles with the same on-the-way charging pile node are taken as candidate charging piles. The path from the main path node to the candidate charging pile is taken as a candidate charging path.

[0072] In the embodiment, the path scoring function is specifically defined as follows:

[0073] Γ(p)=α·T p +β·D p +γ·S p ;

[0074] Wherein, T p is the usage state credibility of the charging pile node v p ; α is the usage state credibility weight; D p is the path distance relative score, which represents the spatial distance mapping score of the charging pile node v p from the midpoint of the main path; β is the path distance relative weight; Sp is the on-way direction consistency score, indicating whether the charging pile is in the main path direction advancing area, and is penalized if the direction deviates; γ is the on-way direction consistency weight; and α+β+γ=1.

[0075] The power load evolution judgment unit 4 is used to construct a charging time evolution graph in an urban area, predict the future power response rate of the candidate charging pile in each candidate path by using a graph neural network model, obtain a path power response matrix, and finally screen to obtain a suitable charging pile and a charging pile path;

[0076] In the embodiment, the power load evolution judgment unit 4 includes a charging time evolution analysis module 41; the charging time evolution analysis module 41 constructs a charging time evolution graph with charging pile nodes as graph nodes and power transmission edges as graph edges, and the charging time evolution graph is used to describe the time sequence evolution relationship of the power load state between the charging piles in the urban area.

[0077] The power transmission edge includes a timestamp, a power change rate, and a geographical proximity factor.

[0078] In the embodiment, the structure of the power transmission edge is as follows:

[0079]

[0080] Wherein, m and n are both charging pile node indexes; v m is the charging pile node m; v n is the charging pile node n.

[0081] is the power transmission edge between the charging pile node m and the charging pile node n. is the power transmission weight between the charging pile node m and the charging pile node n at time τ.

[0082] Wherein, the power transmission weight between the charging pile node m and the charging pile node n at time τ is calculated as follows:

[0083]

[0084] Wherein, λ d is a distance attenuation factor for controlling the propagation range; d mn is the geographical distance between the charging pile node m and the charging pile node n; η is a standard distance constant. represents the power change rate of the charging pile node m; γ mn is the power supply coupling coefficient in the regional power grid, indicating whether there is a physical power supply channel or transformer co-association relationship between two pile sites, and the value range is [0, 1].

[0085] In the embodiment, the power load evolution judgment unit 4 further comprises a power prediction path decision module 42; the power prediction path decision module 42 calculates the future power response rate of all candidate charging piles in each candidate path based on the charging time evolution graph using a time series graph neural network model, constructs a path power response matrix according to the future power response rate of all pile positions in the candidate path, and finally determines the path charging pile and charging path set meeting the user demand based on the path power response matrix;

[0086] The time series graph neural network model is a graph neural network model based on a time dynamic evolution graph, and is used for analyzing and predicting the charging power evolution trend of all candidate charging piles in each candidate path in each city area.

[0087] The future power response rate is an average available power value in a prediction time window, and the high and low of the future power response rate represents the strength of the power supply capacity of the charging pile in the future period; the path power response matrix is a two-dimensional matrix structure composed of the future power response rates of all candidate charging pile nodes in all candidate paths, and is used for expressing the charging power supply level of all candidate paths as a whole.

[0088] In the embodiment, the specific modeling of the time series graph neural network model is as follows:

[0089] Based on the time dynamic evolution graph, a time dynamic evolution graph sequence is constructed which represents the evolution of the graph structure in the past T w time steps, and the input feature vector of each charging pile node includes the current power, the power change rate, the city area density and the user demand;

[0090] A propagation function is defined based on a graph attention mechanism:

[0091]

[0092] Wherein, t is time; AGGREGATE(·) is an aggregation operation in the graph neural network; k and j are both charging pile node indexes; is a neighbor node set of the charging pile node k; is the input feature vector of the charging pile node j at time t; W is a node feature transformation matrix; is the attention weight coefficient of the charging pile node j to the charging pile node k; is the feature representation of the aggregated charging pile node k;

[0093] A time recurrent structure GRU is introduced to update the state of the charging pile node in the time dimension:

[0094]

[0095] Wherein, is the state vector of the charging pile node k at time t; is the state vector of the charging pile node k at time t-1; is the node feature of the charging pile node k for the graph attention mechanism; GRU(·) is a gated recurrent unit;

[0096] The available power prediction value at time t is calculated using a multi-layer perception

[0097]

[0098] wherein, is the available power prediction value of the charging pile node k at time t; MLP(·) is a multi-layer perception.

[0099] In this embodiment, the mathematical definition of the future power response rate is:

[0100]

[0101] wherein, k is the charging pile node index; R k is the future power response rate of the charging pile node k; ΔT is the prediction time window; t is the time; t0 is the prediction starting time; is the available power prediction value of the charging pile node k at time t, which is generated by a time series graph neural network model.

[0102] In embodiment two, the application proposes a new energy vehicle charging service method, which is used for the new energy vehicle charging service system in the above embodiment one, and includes the following steps:

[0103] S10.1, collect multi-source Internet of Things perception data and user charging demand data, and perform standard structured processing;

[0104] S10.2, based on the multi-source Internet of Things perception data and the user charging demand data, fuse a knowledge graph construction method, a visual recognition analysis method, and a three-dimensional space reconstruction method, construct a ternary pile-car relationship graph containing vehicle nodes, pile nodes, and user behavior nodes, and output a credible charging pile set and an unreachable pile site filtering list;

[0105] S10.3, construct a path strategy graph based on the user charging demand data, and generate a candidate charging path and a candidate charging pile through a pile insertion strategy decision mechanism;

[0106] S10.4, construct a charging time evolution graph in a city area, predict the future power response rate of the candidate charging pile in each candidate path using a graph neural network model, obtain a path power response matrix, and finally screen to obtain a suitable charging pile and a charging pile path.

[0107] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the embodiments, and various changes can be made by those skilled in the art within the scope of knowledge acquired from the present disclosure, without departing from the spirit of the present application.

Claims

1. A new energy vehicle charging service system, characterized in that, include: Data access sensing fusion unit (1) is used to collect multi-source IoT sensing data and user charging demand data, and perform standard structured processing. The pile location adaptation graph judgment unit (2) is based on multi-source Internet of Things sensing data and user charging demand data. It integrates knowledge graph construction method, visual recognition analysis method and three-dimensional space reconstruction method to construct a three-element pile-vehicle relationship graph containing vehicle nodes, pile location nodes and user behavior nodes, and outputs a set of reliable charging piles and an unreachable pile location filter list. The specific method for outputting the set of reliable charging stations and the filter list of unreachable charging stations is as follows: S2.1 Construct a reliable scoring function for charging pile nodes based on graph reasoning and graph convolutional neural network, and perform joint reasoning calculation on physical adaptation edges and state feedback edges in the three-dimensional charging pile-vehicle relationship graph to evaluate the reliability of the usage status of each charging pile, and obtain a set of reliable charging piles from the set of charging pile nodes whose usage status reliability is higher than a preset threshold. S2.2 outputs an unreachable charging station filter list from the set of charging station nodes that are physically unavailable or falsely available; The intelligent charging pile insertion unit (3) constructs a path strategy map based on user charging demand data, generates candidate charging paths and candidate charging piles through the charging pile strategy decision mechanism, and transmits the candidate charging paths and candidate charging piles to the power load evolution judgment unit (4). The specific decision-making mechanism for the charging pile insertion strategy is as follows: The mechanism calculates the charging pile score using a path scoring function, sorts the charging piles from highest to lowest score, and selects the top five charging piles with scores consistent with roadside charging pile nodes as candidate charging piles. The path from the main path node to the candidate charging pile is then selected as the candidate charging path. The path scoring function is defined as follows: ； in, For charging pile nodes The reliability of the usage status; To use state credibility weights; The path distance relative score represents the charging pile node. Spatial distance mapping score from the midpoint of the main path; The relative weights of path distances; The score for consistency with the direction of travel indicates whether the charging station is in the area along the main path. If it deviates from the direction, a penalty is imposed. Weights are assigned to those traveling in the same direction; and ; The power load evolution judgment unit (4) is used to construct a charging time evolution map in urban areas, use a graph neural network model to predict the future power response rate of candidate charging piles in each candidate path, obtain the path power response matrix, and finally select suitable charging piles and paths to reach the charging piles.

2. The new energy vehicle charging service system according to claim 1, characterized in that, In the data access sensing fusion unit (1), the multi-source IoT sensing data includes vehicle terminal data, charging pile equipment data, geomagnetic sensor data, parking space image spatial layout data, regional power supply information data, and charging pile operation status; user charging demand data includes vehicle type parameters, remaining power, destination information, path delay tolerance, and following the route intention.

3. The new energy vehicle charging service system according to claim 2, characterized in that, The pile location adaptation map judgment unit (2) includes a three-dimensional space reconstruction module (21). Among them, the three-dimensional space reconstruction module (21) is used to extract the edge features of the parking space and the coordinates of the ground guide line based on the spatial layout data of the parking space image, construct the three-dimensional spatial point cloud model of the parking space, and combine the vehicle type parameters to determine whether the target vehicle can physically enter the parking space range in front of the target pile through geometric matching calculation, and determine the parking space adaptability of the charging parking space.

4. The new energy vehicle charging service system according to claim 3, characterized in that, The three-dimensional charging pile-vehicle relationship map is used to depict the spatial adaptation relationship, state reliability relationship, and user preference behavior pattern between new energy vehicles and urban charging piles. The nodes in the three-dimensional charging pile-vehicle relationship graph include: vehicle nodes, charging pile nodes, and user behavior nodes; vehicle nodes include attributes such as vehicle length, width, height, charging interface type, and minimum turning radius; charging pile nodes include attributes such as charging pile size, cable reachability, fault status, and historical connection frequency; user behavior nodes include user rating records, complaint records, and number of successful charging attempts. The edges of the three-dimensional charging pile-vehicle relationship graph include: physical adaptation edges connecting vehicle nodes and charging pile nodes, used to express the reachability relationship between vehicles and charging piles; state feedback edges connecting charging pile nodes and user behavior nodes, used to express the reliability relationship of charging pile status; and usage preference edges connecting vehicle nodes and user behavior nodes, used to express the user's historical behavior constraint relationship.

5. The new energy vehicle charging service system according to claim 4, characterized in that, The pile location adaptation map judgment unit (2) includes a charging pile screening module (22), which is used to output a set of reliable charging piles and a list of unreachable pile locations. The trusted charging pile set is a set of available charging piles that meet the requirements of physical parking accessibility, functional availability, and reliable usage status; the unreachable charging pile filter list is a set of charging piles that the current vehicle cannot physically enter, that are unavailable, or that are in a false available state.

6. The new energy vehicle charging service system according to claim 5, characterized in that, The path strategy intelligent instrumentation unit (3) includes a path strategy graph construction module (31). The path strategy graph construction module (31) is used to construct a path strategy graph based on user charging demand data; the path strategy graph includes: main path nodes, charging pile nodes along the route, and strategy control factor edges; Among them, the main path node represents the node in the shortest path segment between the origin and the destination; the charging pile node along the way represents the location of the charging pile that can be reached from the main path under the constraint of the user's intention to follow the route; the policy control factor edge is used to represent the path extension adjustment relationship between the main path node and the charging pile node along the route, including the path delay cost relationship edge, the charging waiting cost relationship edge and the user preference matching relationship edge.

7. The new energy vehicle charging service system according to claim 6, characterized in that, The intelligent charging pile insertion unit (3) of the path strategy also includes a path sorting module (32); the path sorting module (32) sorts the candidate paths in the path strategy map by the charging pile scoring mechanism, and generates candidate charging paths and candidate charging piles.

8. The new energy vehicle charging service system according to claim 7, characterized in that, The power load evolution judgment unit (4) includes a charging time evolution analysis module (41); the charging time evolution analysis module (41) constructs a charging time evolution graph with charging pile nodes as graph nodes and power transmission edges as graph edges. The charging time evolution graph is used to describe the temporal evolution relationship of the power load status between charging piles in the urban area. The power transfer edge includes a timestamp, a power change rate, and a geographical proximity factor.

9. The new energy vehicle charging service system according to claim 8, characterized in that, The power load evolution judgment unit (4) also includes a power prediction path decision module (42); the power prediction path decision module (42) calculates the future power response rate of all candidate charging piles in each candidate path using a time-series neural network model based on the charging time evolution map, and constructs a path power response matrix based on the future power response rate of all piles in the candidate path, and finally determines the set of path charging piles and charging paths that meet the user's needs based on the path power response matrix; Among them, the time-series graph neural network model is a graph neural network model based on the time dynamic evolution graph, which is used to analyze and predict the evolution trend of charging power of all candidate charging piles in each candidate path in various urban areas; The future power response rate is the average available power value within the prediction time window. The level of the future power response rate represents the strength of the power supply capability of the charging pile in the future period. The path power response matrix is ​​a two-dimensional matrix structure composed of the future power response rates of each candidate charging pile node in all candidate paths, used to express the overall charging power supply level of all candidate paths.

10. A method for providing charging services for new energy vehicles, used in a new energy vehicle charging service system as described in any one of claims 1-9, characterized in that: Includes the following steps: S10.1 Collect multi-source IoT sensing data and user charging demand data, and perform standard structured processing; S10.2 Based on multi-source IoT sensing data and user charging demand data, integrate knowledge graph construction method, visual recognition analysis method and three-dimensional space reconstruction method to construct a three-element charging pile-vehicle relationship graph containing vehicle nodes, charging pile nodes and user behavior nodes, and output a set of reliable charging piles and a list of unreachable charging piles. The specific method for outputting the set of reliable charging stations and the filter list of unreachable charging stations is as follows: S2.1 Construct a reliable scoring function for charging pile nodes based on graph reasoning and graph convolutional neural network, and perform joint reasoning calculation on physical adaptation edges and state feedback edges in the three-dimensional charging pile-vehicle relationship graph to evaluate the reliability of the usage status of each charging pile, and obtain a set of reliable charging piles from the set of charging pile nodes whose usage status reliability is higher than a preset threshold. S2.2 outputs an unreachable charging station filter list from the set of charging station nodes that are physically unavailable or falsely available; S10.3 Construct a path strategy map based on user charging demand data, and generate candidate charging paths and candidate charging piles through a charging pile strategy decision-making mechanism; The specific decision-making mechanism for the charging pile insertion strategy is as follows: The mechanism calculates the charging pile score using a path scoring function, sorts the charging piles from highest to lowest score, and selects the top five charging piles with scores consistent with roadside charging pile nodes as candidate charging piles. The path from the main path node to the candidate charging pile is then selected as the candidate charging path. The path scoring function is defined as follows: ； in, For charging pile nodes The reliability of the usage status; To use state credibility weights; The path distance relative score represents the charging pile node. Spatial distance mapping score from the midpoint of the main path; The relative weights of path distances; The score for consistency with the direction of travel indicates whether the charging station is in the area along the main path. If it deviates from the direction, a penalty is imposed. Weights are assigned to those traveling in the same direction; and ; S10.4 Construct a charging time evolution map in urban areas, use a graph neural network model to predict the future power response rate of candidate charging piles in each candidate path, obtain the path power response matrix, and finally select suitable charging piles and paths to reach the charging piles.

Images