CN-121996849-A - Charging station dynamic recommendation method and system based on multidimensional real-time data

CN121996849ACN 121996849 ACN121996849 ACN 121996849ACN-121996849-A

Abstract

The invention discloses a charging station dynamic recommendation method and a charging station dynamic recommendation system based on multidimensional real-time data, which relate to the technical field of big data analysis and comprise the steps of acquiring charging data of all users according to charging requests of the users, wherein the charging data comprises historical order data, real-time charging gun data, charging pile data and charging station data; and updating all charging station node characteristics based on adjacent order nodes, order node information of each user, charging gun node characteristics and charging pile node characteristics to acquire final all charging station node characteristics, and calculating recommendation indexes of all charging stations to recommend to the user, so that recommendation results can more meet the actual intention of the user and station operation cooperativity.

Inventors

LIN CHUANWEN
LIU RUI
WANG DINGMIN
HU ZHENG
CHEN FANGFANG
CUI HAIYING
LU SHENGGAN
XING JING

Assignees

合肥中安数据科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251205

Claims (10)

1. A charging station dynamic recommendation method based on multidimensional real-time data is characterized by comprising the following steps: s1, acquiring charging data of all users according to charging requests of the users, wherein the charging data comprises historical order data, real-time charging gun data, charging pile data and charging station data; s2, constructing a multi-source heterogeneous relation graph G= (V, E) of a related order, a charging gun, a charging pile and a charging station based on charging data of all users, wherein V is a node set, and E is an edge set; The nodes comprise order nodes, charging gun nodes, charging pile nodes and charging station nodes, wherein the sides comprise order-order single sides, order-charging gun sides, charging gun-charging pile sides and charging pile-charging station sides; s3, constructing a corresponding initial feature vector for each node in the multi-source heterogeneous relation graph; s4, updating the characteristics of each order node of all users based on adjacent order nodes; S5, updating all charging gun node characteristics based on the order node characteristics of each user; S6, updating all the node characteristics of the charging piles based on the node characteristics of the charging gun; S7, updating all charging station node characteristics based on all charging pile node characteristics; S8, carrying out loop iteration L times on the steps S4-S7, obtaining the node characteristics of all charging stations after final updating, and calculating recommendation indexes of all charging stations; And S9, recommending the charging stations to the users according to the queuing sequence of the charging stations from high to low according to the recommendation index of the charging stations.
2. The charging station dynamic recommendation method based on multidimensional real-time data according to claim 1, wherein the order data includes an order ID, a vehicle ID, a charging gun ID, a start charging time, an end charging time, a charging duration, a total electricity charge, a total service charge, a total charge amount, and a charge end reason; The charging gun data comprises a charging gun ID, a charging pile ID, a charging gun type, a charging gun rated power, a charging gun real-time power, a charging gun current state and a charging gun current state duration; the charging pile data comprises charging pile ID, belonging charging station ID, charging pile type, charging pile rated power, charging pile real-time power, charging pile state information and charging pile state duration; charging station data comprises charging station ID, charging station longitude and latitude, equipment online rate, equipment failure rate and star rating of user; one charging station includes a plurality of charging posts, and one charging post includes a plurality of charging guns.
3. The method for dynamically recommending charging stations based on multidimensional real-time data according to claim 1, wherein the specific construction process of the multi-source heterogeneous relationship graph G of the associated order-charging gun-charging pile-charging station is as follows: The node definition comprises creating an order node for each effective historical order or in-progress order, creating a charging gun node for each physical charging gun, creating a charging pile node for each physical charging pile, and creating a charging station node for each charging station; The method comprises the steps of defining an order-order side, an order-gun side, a gun-pile side, a charging pile node and a charging gun node, wherein the order-order side is used for establishing a bidirectional side between all order nodes placed by the same vehicle; Pile-station side, according to the belonged charging station ID in the charging pile data, the charging pile node points to the belonged charging station node, and a directed side is established.
4. The method of claim 1, wherein in step S3, the ith order node is characterized by Wherein, the method comprises the steps of, Represent the first The length of the charge time for each order, Represent the first The total electric charge of the individual orders is calculated, Represent the first The total service charge for the individual orders is, Represent the first The total charge amount of the individual orders, Represent the first The charging end reason of each order; jth charging gun node feature Wherein, the method comprises the steps of, Represent the first The type of charging gun of the individual charging guns, Represent the first The charging guns of the charging guns are rated for power, Represent the first The real-time power of the charging gun of each charging gun, Represent the first The charge gun status information of the individual charge guns, Represent the first A charge gun state duration feature of the charge guns; Node feature of the (r) th charging pile Wherein, the method comprises the steps of, Represent the first The type of charging stake of the individual charging stake, Represent the first The rated power of the charging piles, Represent the first The real-time power of the charging piles, Represent the first The charging pile status information of the individual charging piles, Represent the first The charging pile state duration of each charging pile; Kth charging station node feature ; Represent the first The charging station longitude of the individual charging stations, Represent the first The dimensions of the charging stations of the individual charging stations, Represent the first The device presence rate of the individual charging stations, Represent the first The equipment off-line rate of the individual charging stations, Represent the first And (5) evaluating star grade of each charging station.
5. The method for dynamically recommending charging stations based on multidimensional real-time data as recited in claim 4, wherein step S4 comprises: s41 for each order node All neighbor order nodes of the current order node are found through an order-order unilateral and recorded as an order set ; S42, aggregating all order node information pointing to the order node i, and capturing the attention coefficient of each order to the adjacent order nodes in the multi-source heterogeneous relation diagram : Wherein, the Representing an activation function; Representing an exponential function; Representing order node characteristics Is a transpose of (2); , the representation points to the first An order node set of individual order nodes; the representation points to the first Other p-th order node characteristics of the vehicle corresponding to the orders; Representation of The number of collections, i.e., the total amount of orders associated with the vehicle to which the current order corresponds; indicate the opposite direction to the first Attention coefficients of other p-th order nodes of the vehicle corresponding to the orders; s43, merging the attention coefficients of all the neighbor order nodes, and updating the characteristics of each order node of all the users: Wherein, the Representing a feed-forward neural network; the ith order node characteristic updated for each user; For the ith order node feature obtained in the previous cycle, in the first cycle I.e. 。
6. The method for dynamically recommending charging stations based on multidimensional real-time data according to claim 5, wherein step S5 comprises: S51, for each charging gun node j, finding all order nodes pointing to the charging gun node j through an order-gun side, and marking the charging gun node as a set M (j); s52, calculating the attention coefficient of each charging gun node j to each order node in the order combination M (j) associated with the charging gun node j : Representing charging gun node characteristics Transpose of the matrix; , representation pointing and the first An order node set associated with each charging gun node; the representation points to the first The first charging gun node Individual order node features; Representation of Number of sets; indicate the opposite direction to the first First of charging guns Attention coefficients of the individual order nodes; S53, aggregating all historical order information associated with the current charging gun node, and updating all charging gun node characteristics: Wherein, the The node characteristics of the j-th charging gun after updating are obtained; the j-th charging gun node characteristic obtained for the previous cycle is obtained in the first cycle I.e. 。
7. The method for dynamically recommending charging stations based on multidimensional real-time data as recited in claim 6, wherein step S6 comprises: S61, for each charging pile node r, finding all subordinate charging gun nodes pointing to the charging pile node r through gun-pile edges, and marking the charging gun nodes as a set S (r); s62, calculating each charging gun node in each charging pile node r pair and each charging gun node in the charging gun node set S (r) associated with each charging pile node r Attention coefficient of (a) : Wherein, the Representing charging pile node characteristics Is a transpose of (2); the representation points to the first A charging gun node set of the charging pile nodes; the representation points to the first The first charging pile node Node characteristics of the individual charging guns; Representation of Number of sets; indicate the opposite direction to the first First of charging piles Attention coefficients of the individual charging gun nodes; S63, aggregating subordinate charging gun information, and updating charging pile node characteristics: Wherein, the The node characteristics of the r-th charging pile after updating are obtained; The node characteristic of the (r) charging pile obtained for the previous cycle is that in the first cycle I.e. 。
8. The method for dynamically recommending charging stations based on multidimensional real-time data according to claim 7, the method is characterized in that the step 7 comprises the following steps: S71, for each charging station node k, finding all subordinate charging pile nodes pointing to the charging station node through the pile-station edge, and marking the charging pile nodes as a set D (k); S72, calculating the attention coefficient of each charging station node k to each charging pile node in the charging pile node set D (k) associated with the charging station node k : Wherein, the Representing charging station node characteristics Is a transpose of (2); , the representation points to the first A set of charging stake nodes of the individual charging station nodes; the representation points to the first F-th charging pile node characteristic of each charging station node, F represents Number of sets; indicate the opposite direction to the first The attention coefficient of the f-th charging pile node of the charging stations; S73, integrating the node characteristics of the charging piles with all attention coefficients, and updating the first node Characterization of individual charging station nodes: Wherein, the To the updated first A plurality of charging station node features; The node characteristic of the kth charging pile obtained for the previous cycle is that in the first cycle I.e. 。
9. The method for dynamically recommending charging stations based on multidimensional real-time data according to claim 8, wherein in step S8, a recommendation index calculation formula of the charging stations is: Wherein, the A transpose representing the final updated kth charging station node characteristics; Representing a matrix of learnable parameters; representing a learnable bias; Represent the first Recommendation index for each charging station.
10. The charging station dynamic recommendation system based on the multi-dimensional real-time data is characterized by comprising a feature selection acquisition module, a graph construction module, a charging station node feature acquisition module and a charging station recommendation module, wherein the feature selection acquisition module is used for selecting and acquiring charging parameters, the graph construction module is used for constructing a multi-source heterogeneous relation graph of the charging parameters, the charging station node feature acquisition module is used for acquiring charging station node features based on order node features, charging gun node features and charging pile node features, and the charging station recommendation module is used for recommending charging stations for charging requests of current users based on charging station recommendation indexes.

Description

Charging station dynamic recommendation method and system based on multidimensional real-time data Technical Field The invention relates to the technical field of big data analysis, in particular to a charging station dynamic recommendation method and system based on multidimensional real-time data. Background Along with the popularization of new energy automobiles, an efficient and intelligent charging station recommendation system is important for improving user experience and station operation efficiency. The existing charging station recommendation technology mainly comprises two types, namely a recommendation based on a location service and a static rule, such as a charging station with the closest recommendation distance, the lowest price or the largest number of idle guns, and a personalized recommendation model based on collaborative filtering or matrix decomposition, wherein the recommendation model is used for predicting preference by using historical order data of a user. However, the prior art has significant drawbacks. First, they generally treat charging devices as static resources, using only historical average data or instantaneous idle states, lack fine-grained modeling of charging guns, charging pile real-time operating states (e.g., faults, power fluctuations), and dynamic reliability, resulting in a poor experience of "arrive-at-fail" where users may be recommended to "surface idle, practically unavailable" devices. Secondly, the existing method cannot effectively fuse multi-source heterogeneous data, particularly cannot explicitly model the physical hierarchy relation of a charging facility 'gun-pile-station', and cannot model complex association between user charging preference and station service capability, so that the recommendation result has defects in equipment availability, user real intention matching degree and station operation cooperativity. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides a charging station dynamic recommendation method and a charging station dynamic recommendation system based on multidimensional real-time data. In order to achieve the above purpose, the present invention adopts the following technical scheme, including: A charging station dynamic recommendation method based on multidimensional real-time data comprises the following steps: s1, acquiring charging data of all users according to a charging request of a current user, wherein the charging data comprises historical order data, real-time charging gun data, charging pile data and charging station data; s2, constructing a multi-source heterogeneous relation graph G= (V, E) of a related order, a charging gun, a charging pile and a charging station based on charging data of all users, wherein V is a node set, and E is an edge set; The nodes comprise order nodes, charging gun nodes, charging pile nodes and charging station nodes, wherein the sides comprise order-order single sides, order-charging gun sides, charging gun-charging pile sides and charging pile-charging station sides; s3, constructing a corresponding initial feature vector for each node in the multi-source heterogeneous relation graph; s4, updating the characteristics of each order node of all users based on adjacent order nodes; S5, updating all charging gun node characteristics based on the order node characteristics of each user; S6, updating all the node characteristics of the charging piles based on the node characteristics of the charging gun; S7, updating all charging station node characteristics based on all charging pile node characteristics; S8, carrying out loop iteration L times on the steps S4-S7, obtaining the node characteristics of all charging stations after final updating, and calculating recommendation indexes of all charging stations; And S9, recommending the charging stations to the current user according to the queuing sequence of the charging stations from high to low according to the recommendation index of the charging stations. Preferably, the order data includes an order ID, a vehicle ID, a charging gun ID, a start charging time, an end charging time, a charging duration, a total electricity charge, a total service charge, a total charge amount, and a charge end reason; The charging gun data comprises a charging gun ID, a charging pile ID, a charging gun type, a charging gun rated power, a charging gun real-time power, a charging gun current state and a charging gun current state duration; the charging pile data comprises charging pile ID, belonging charging station ID, charging pile type, charging pile rated power, charging pile real-time power, charging pile state information and charging pile state duration; charging station data comprises charging station ID, charging station longitude and latitude, equipment online rate, equipment failure rate and star rating of user; one charging station includes a plurality of charging posts, and one charging post includes