User feature-based charging station operation policy intelligent generation method and system

Abstract

This invention relates to the field of intelligent charging station operation optimization technology, and particularly to a method and system for intelligently generating charging station operation strategies based on user characteristics. It includes the following steps: S1: Dividing the charging station area into functional zones and virtual functional zones, and constructing a charging station operation feedback graph network for each user based on the functional zones and virtual functional zones; S2: Calculating the correlation index of each node based on the edge weights constructed according to the connection edge types between nodes in the charging station operation feedback graph network; S3: Constructing a strategy data sequence based on the charging station operation feedback graph network for each user. This invention constructs a charging station operation feedback graph network that treats physical functional zones and their corresponding virtual functional zones as peer nodes, and establishes connections based on five types of connection edges that contain clear business logic, profoundly revealing the behavioral hub attributes and potential preferences of users in the integrated space.

Description

Technical Field

[0001] This invention relates to the field of intelligent charging station operation optimization technology, and in particular to a method and system for intelligently generating charging station operation strategies based on user characteristics. Background Technology

[0002] With the continuous growth of electric vehicle ownership, charging stations are gradually evolving from single energy replenishment facilities into comprehensive service nodes integrating information services, user interaction, and diversified operational functions. How to achieve efficient allocation of charging resources and refined formulation of operational strategies, given the significant spatial and temporal differences in charging demand and user behavior, has become a key issue in the intelligent development of charging infrastructure.

[0003] Existing charging station operation and management solutions mostly rely on charging pile operating status, historical utilization rates, or simple user statistics, configuring strategies through preset rules or fixed models. These methods typically have the following limitations in practical applications: First, the internal space of a charging station is often treated as a single or limited functional area, lacking fine-grained characterization of user movement, dwelling, and behavioral differences between different physical areas, making it difficult to reflect true space usage patterns. Second, with the introduction of digital services such as pre-booked charging, online guidance, virtual queuing, and community interaction, user behavior now occurs simultaneously in both physical and virtual spaces, but existing technologies usually treat these two aspects separately, failing to establish an effective correlation model between virtual interactive behavior and subsequent physical charging behavior. Summary of the Invention

[0004] To overcome the shortcomings of existing charging station operations, such as the separation of physical and virtual data, superficial understanding of user behavior, and static and rigid strategy generation, this invention provides a method and system for intelligent generation of charging station operation strategies based on user characteristics.

[0005] Technical solution: A method for intelligently generating charging station operation strategies based on user characteristics, including the following steps: S1: Divide the area of ​​the charging station to obtain the charging station functional area and the virtual functional area, and construct the charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area; S2: Calculate the correlation index of each node based on the edge weights constructed according to the connection edge types between each node in the charging station operation feedback graph network. S3: Construct a strategy data sequence based on the charging station operation feedback graph network of each user; S4: Obtain a value feedback data sequence based on the strategy data sequence, and perform strategy evaluation based on the value feedback data sequence; S5: Build and apply dynamic operational knowledge graphs to optimize strategy generation.

[0006] Preferably, the step of dividing the charging station area into charging station functional areas and virtual functional areas, and constructing a charging station operation feedback graph network for each user based on the charging station functional areas and the virtual functional areas, includes: obtaining the virtual functional areas corresponding to each charging station functional area, where the virtual functional area is the area where the user uses the corresponding virtual space within the charging station functional area; analyzing the user's area usage; constructing a first type of connection edge: constructing bidirectional connection edges between the charging station functional areas and the corresponding virtual functional areas; constructing a second type of connection edge: constructing unidirectional or bidirectional connection edges based on the user's usage within the charging station functional areas; and constructing a third type of connection edge: based on... User interaction behavior between virtual functional areas establishes unidirectional or bidirectional connections between these areas; a fourth type of connection is constructed: based on historical policy execution records, a unidirectional connection is established from the virtual functional area to the charging station functional area; a fifth type of connection is constructed: based on physical space usage data feedback, a unidirectional connection is established from the charging station functional area to a non-corresponding virtual functional area; each charging station functional area and its corresponding virtual functional area are used as graph nodes; based on the bidirectional connections, the unidirectional connections, and the graph nodes, a charging station operation feedback graph network for each user is constructed; when constructing the charging station operation feedback graph network, edge weights are constructed according to the connection edge types between nodes.

[0007] Preferably, the virtual functional area is the area in the charging station functional area where the user uses the corresponding virtual space, including: the virtual functional area is an online meeting area where multiple users have a common destination or an online meeting area where multiple users have an unconfirmed destination.

[0008] Preferably, when constructing the charging station operation feedback graph network, edge weights are constructed according to the connection edge types between each node, including: setting the edge weight of the first type of connection edge to one to represent the basic mapping relationship; obtaining the edge weight of the second type of connection edge based on the number of times the user moves within the functional area of ​​the charging station, the average movement time, and the value generated during the movement; obtaining the edge weight of the third type of connection edge based on the frequency and complexity of the user's virtual interaction between virtual functional areas; obtaining the edge weight of the fourth type of connection edge based on the success rate and value of historical strategies; and obtaining the edge weight of the fifth type of connection edge based on the frequency and quality of data feedback.

[0009] Preferably, the step of calculating the correlation index of each node based on the edge weights constructed according to the connection edge types between nodes in the charging station operation feedback graph network includes: when calculating the correlation index of each node, physical activity, virtual influence, user flow centrality, data feedback intensity, or virtual-real coupling degree; the physical activity is the sum of the in-degree weights and out-degree weights of the second type of connection edge; the virtual influence is the sum of the out-degree weights of the fourth type of connection edge; the user flow centrality is the node importance of the second type of connection edge and the third type of connection edge; the data feedback intensity is the sum of the in-degree weights of the fifth type of connection edge; and the virtual-real coupling degree is the sum of the connection weights of the node in the first type of connection edge.

[0010] Preferably, the user flow centrality is the node importance of the second type of connection edge and the third type of connection edge, including: obtaining the user flow centrality through a flow formula, wherein the flow formula is: ; In the formula For nodes The importance of nodes, i.e., the centrality of user flow; For the node To the node And passing through nodes The number of shortest paths; The number of all shortest paths; Path weight; ; In the formula, For normalized slave nodes To the node Number of users along the path; For normalized slave nodes To the node Average user value of the path; The normalized average dwell time is the average dwell time from the user's node. To the node The average total time spent at all nodes along the route; These are the weighting coefficients.

[0011] Preferably, the step of constructing a strategy data sequence based on the charging station operation feedback graph network of each user includes: sorting each node in the charging station operation feedback graph network of each user according to the correlation index to form a node sorting sequence; performing node analysis on each node in the node sorting sequence; the node analysis is an analysis of the strategy of each user at each node according to the historical fourth type of connection edge in chronological order, to obtain the strategy data sequence of each user at each node.

[0012] Preferably, the step of obtaining a value feedback data sequence based on a strategy data sequence and evaluating a strategy based on the value feedback data sequence includes: obtaining user conversion data based on the strategy data sequence; the conversion data is data obtained based on user conversion behavior, where the conversion behavior is the user's behavior of interacting with a virtual functional area and then charging a physical device in the node sorting sequence; the conversion data includes instantaneous conversion data and continuous conversion data; the instantaneous conversion data is the behavior of not subsequently interacting with a virtual functional area and then charging a physical device under the corresponding strategy; the continuous conversion data is the behavior of continuously interacting with a virtual functional area and then charging a physical device under the corresponding strategy; sorting the conversion data according to the order of the node sorting sequence to obtain a value feedback data sequence; calculating the position difference and frequency of the same strategy among the value feedback data sequences, and evaluating the effectiveness of the strategy based on the position difference and frequency.

[0013] Preferably, the construction and application of a dynamic operational knowledge graph to optimize strategy generation includes: based on the graph nodes, connecting edges, and correlation indicators in the charging station operation feedback graph network, integrating multi-source heterogeneous data such as charging pile operation status, user feedback text, and environmental parameters, extracting and defining knowledge nodes including operational entities, user entities, and strategy entities, as well as the relationships between knowledge nodes, to construct a dynamic operational knowledge graph; the dynamic operational knowledge graph is used for strategy reasoning, including reasoning and generating readable natural language strategy descriptions based on the correlations between entities in the graph and the current operational status; the quality of the dynamic operational knowledge graph is dynamically evaluated and updated through preset operational indicators; the strategy schemes generated by the dynamic operational knowledge graph reasoning will be matched and filtered in conjunction with historical user feature profiles to achieve differentiated strategy pushes to different user groups that are price-sensitive and time-sensitive.

[0014] Preferably, the intelligent generation system for charging station operation strategies based on user characteristics includes: The area division module is used to divide and construct the area of ​​the charging station, and obtain the functional area and virtual functional area of ​​the charging station; The network construction module is used to construct a charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area. The edge weight construction module is used to construct edge weights based on the connection edge types between nodes when building a charging station operation feedback graph network. The indicator calculation module is used to calculate the correlation index of each node based on the edge weights constructed by the connection edge types between each node in the charging station operation feedback graph network. The importance calculation module assigns the importance of nodes to the second and third types of connection edges based on user flow centrality. The strategy sequence construction module is used to construct strategy data sequences based on the charging station operation feedback graph network of each user. The strategy evaluation module is used to obtain a value feedback data sequence based on the strategy data sequence, and to perform strategy evaluation based on the value feedback data sequence. The strategy generation module is optimized to build and apply a dynamic operational knowledge graph to optimize strategy generation. Beneficial effects

[0015] 1. This invention incorporates both the physical and virtual functional areas within a charging station into a unified modeling framework, and expresses the behavioral relationships of users in both virtual and physical spaces using a graph network. This allows for a more comprehensive reflection of the user's actual usage trajectory within the charging station scenario. This virtual-physical integrated modeling approach breaks through the limitations of traditional methods that only focus on physical space, explicitly expressing the correlation between online user interactions, path flows, and physical charging behaviors, thus providing a more information-dense data foundation for subsequent strategy analysis. 2. This invention introduces multiple types of connection edges and their corresponding edge weights to construct a unified mapping of information from different sources, such as user movement, virtual interaction, historical strategy execution effects, and data feedback, into a structured relationship for computation. Furthermore, it is no longer limited to isolated single-event effect judgments, but constructs a strategy data sequence based on node sorting sequences, and further forms a value feedback data sequence, thereby achieving a comprehensive analysis of the strategy's performance at different stages and along different paths. Attached Figure Description

[0016] Figure 1 This is a flowchart of the intelligent generation method for charging station operation strategies based on user characteristics according to the present invention. Figure 2 This is a schematic diagram of the charging station operation feedback network of the present invention; Figure 3 This is a schematic diagram of the structure of the strategy data sequence of the present invention; Figure 4 This is a schematic diagram of the value feedback data sequence structure of the present invention; Figure 5 This is a schematic diagram of the intelligent generation system for charging station operation strategies based on user characteristics according to the present invention. Detailed Implementation

[0017] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but this does not limit the scope of protection and application of the present invention.

[0018] Example 1: A method for intelligently generating charging station operation strategies based on user characteristics, such as... Figure 1 , Figure 3 , Figure 4 and Figure 5As shown, it includes the following steps: S1: Divide the area of ​​the charging station to obtain the charging station functional area and the virtual functional area, and construct the charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area; Based on the functional areas of each charging station, virtual functional areas are obtained corresponding to each charging station functional area. These virtual functional areas represent the areas where users utilize the corresponding virtual space within the charging station functional area. The user's area usage is analyzed, and the following connection edges are constructed: First type of connection edge: bidirectional connection edges are constructed between the charging station functional areas and their corresponding virtual functional areas. Second type of connection edge: unidirectional or bidirectional connection edges are constructed based on the user's usage within the charging station functional areas. Third type of connection edge: unidirectional or bidirectional connection edges are established between virtual functional areas based on the user's interaction behavior between them. Fourth type of connection edge: unidirectional connection edges are established from virtual functional areas to charging station functional areas based on historical policy execution records. Fifth type of connection edge: unidirectional connection edges are established from charging station functional areas to non-corresponding virtual functional areas based on physical space usage data feedback. Each charging station functional area and its corresponding virtual functional area are used as graph nodes. Based on the bidirectional connection edges, unidirectional connection edges, and graph nodes, a charging station operation feedback graph network for each user is constructed. When constructing the charging station operation feedback graph network, edge weights are constructed based on the connection edge types between nodes.

[0019] It needs to be explained in detail, such as Figure 3As shown, this embodiment takes a single charging station as the object and constructs an independent charging station operation feedback graph network for each user. The implementation process includes data collection and preprocessing, mapping the corresponding functional areas of the charging station to virtual functional areas, analyzing the usage of areas by user sequence, constructing five types of connection edges by type, using each charging station functional area and its corresponding virtual functional area as graph nodes to form a graph structure, and finally generating edge weights by connection edge type and writing them into edge attributes. Examples of data sources include: charging pile occupancy or release logs, charging start or end records, in-site positioning or heat map data, parking space detection sensors, monitoring or indoor positioning systems, user mobile application operation logs, historical policy execution records, and user structured feedback or evaluation data. The raw logs are sessionified by user ID and time window, and physical events are mapped to charging station functional areas, such as the entrance area, waiting area, and charging station. The system includes charging station areas, rest areas, and retail areas. Online / APP events are mapped to virtual functional areas, such as appointment sessions, online navigation pages, online meeting rooms, or destination confirmation pages. Noise is eliminated, and timestamps are standardized to the system time base. The mapping between charging station functional areas and virtual functional areas uses a combination of static and dynamic mapping: the static mapping defines regular correspondences, such as a rest area corresponding to an online chat room; the dynamic mapping automatically identifies based on usage behavior, for example, if a charging station functional area experiences a large number of visits to the same online meeting room within a short period, a temporary correspondence is established between that functional area and the corresponding virtual functional area. During the construction process, each charging station functional area and its corresponding virtual functional area are treated as one or more graph nodes. The first type of connection edge: When the same user performs physical behavior in a charging station functional area and online interaction in the corresponding virtual functional area within the same session or short window, a bidirectional connection edge is established between the two nodes, representing basic mapping and bidirectional perception. The second type of connection edge: Based on the user's movement trajectory in the functional area of ​​the charging station, obtained from the location record, the number of movements from functional area A to functional area B, the average movement time, and the business value generated along the way are counted. If the movement is a normal round trip, a bidirectional connection edge is established; otherwise, a unidirectional connection edge is established. In this embodiment, the business value generated along the way is the total consumption of the user on this path. The third type of connection edge is established based on the user's interaction behavior sequence between different virtual functional areas, such as online meeting rooms, destination confirmation pages, and navigation pages. If the user frequently jumps from virtual area X to virtual area Y, and the jump is in a stable pattern within the statistical window, then a one-way connection is established between X and Y. If the jump frequency between X and Y is symmetrical and significant, then a two-way connection is established. The fourth type of connection edge: When a historical strategy, such as a discount strategy or reminder strategy, is triggered in a virtual functional area, leading to a user arriving at and completing physical charging or generating a target behavior in a charging station functional area, a one-way connection between the triggering virtual node and the target physical node is recorded. This connection reflects the historical mapping between the strategy and the physical behavior; The fifth type of connection edge: When physical space usage data, such as abnormal passenger flow, long stay, and equipment failure reports, shows that the usage pattern of a certain charging station functional area is significantly related to that of a non-corresponding virtual functional area, for example, when the queuing information of a certain functional area occurs simultaneously with the increased interaction frequency of an online discussion room that is not its corresponding one, a one-way connection is established from the charging station functional area to the non-corresponding virtual functional area. This is used to capture the cross-mapping influence path and abnormal propagation link for anomaly detection, cross-area influence analysis, and to reveal hidden virtual-physical coupling paths.

[0020] Virtual function areas are online meeting areas where multiple users share a common destination or where multiple users are waiting to confirm their destination.

[0021] It needs to be explained in detail that a group of users refers to multiple passengers in the same vehicle queue, or multiple independent users, who interact in a virtual meeting / group chat about the same destination. This could include organizing time-sharing charging, aggregating merchant discounts, or discussing carpooling destinations. Ultimately, they form a collective intention to reach the same destination through online conversations. The determination rule is as follows: By parsing the metadata and conversation content of the online meeting room, such as the meeting title, intention tags, shared map coordinates, or destination confirmation actions, the common destination attribute is determined. When the following conditions are met within the same meeting room, that meeting room is identified as an online meeting area with a common destination for multiple users. In online meetings, users haven't finalized their destinations, and multiple candidate destinations await group confirmation. For example, in a group chat, discussions might involve choosing between shopping mall A and mall B, requiring a vote or final confirmation. In this case, the meeting room is recorded as a virtual function area for destinations awaiting confirmation. Judgment rule: If multiple candidate destination tags appear in the meeting room without a clear final confirmation action, but multiple interactions or voting behaviors occur, it's classified as a virtual function area for destinations awaiting confirmation. The list of candidate destinations, voting / discussion status, and participating user list are recorded, and corresponding charging discount strategies are provided based on the destination modification.

[0022] The weight of the first type of connection edge is set to one to represent the basic mapping relationship; the weight of the second type of connection edge is obtained based on the number of times the user moves within the functional area of ​​the charging station, the average movement time, and the value generated during the movement; the weight of the third type of connection edge is obtained based on the frequency and complexity of the user's virtual interaction between virtual functional areas; the weight of the fourth type of connection edge is obtained based on the success rate and value of historical strategies; and the weight of the fifth type of connection edge is obtained based on the frequency and quality of data feedback.

[0023] It should be explained in detail that the first type of connection edge is used to represent the basic mapping relationship between the charging station functional area and its corresponding virtual functional area. In this embodiment, the weight of this type of edge is directly set to a constant 1, indicating the existence of the basic mapping and the equal-weight mapping, so as to ensure that the virtual and physical nodes can be connected and participate in subsequent calculations. The second type of connection edge reflects the actual movement behavior of users in the charging station functional area. Its edge weight comprehensively considers three original indicators: the number of movements, the average movement time, and the value generated during the movement. The edge weight of the second type of connection edge is obtained by normalization and summation. The third type of connection edge measures the online interaction intensity between virtual functional areas. The indicators include interaction frequency and interaction complexity. The interaction complexity is synthesized by message length, number of participants, session duration, and number of interaction trigger events. The edge weight of the third type of connection edge is obtained by normalization and summation of interaction frequency and interaction complexity. The fourth type of connection edge is derived from the historical policy execution record and is used to represent the policy influence direction from the virtual functional area to the charging station functional area. The edge weights of this type are based on the success rate and value of historical strategies. The strategy value is obtained by summing the average revenue or conversion rate increase per trigger. The edge weights of the fourth type of connection edges are obtained by normalizing and summing the success rate and strategy value of historical strategies. The fifth type of connection edges reflects the impact of physical space usage data feedback on non-corresponding virtual functional areas. The edge weights are determined by the data feedback frequency and data quality. The data feedback frequency is the number of reports per unit time. The data quality is synthesized by normalizing and summing the packet loss rate, timestamp deviation, and anomaly ratio. The edge weights of the fifth type of connection edges are obtained by normalizing and summing the data feedback frequency and data quality. The normalization method used is the min-max normalization method. To ensure the comparability, stability, and interpretability of edge weights of different categories, a full graph normalization is performed on all edge weights after the final calculation.

[0024] S2: Calculate the correlation index of each node based on the edge weights constructed according to the connection edge types between each node in the charging station operation feedback graph network. When calculating the correlation index of each node, physical activity, virtual influence, user flow centrality, data feedback intensity, or virtual-real coupling degree are included; the physical activity is the sum of the in-degree weights and out-degree weights of the second type of connection edges; the virtual influence is the sum of the out-degree weights of the fourth type of connection edges; the user flow centrality is the node importance of the second and third type of connection edges; the data feedback intensity is the sum of the in-degree weights of the fifth type of connection edges; and the virtual-real coupling degree is the sum of the connection weights of the node in the first type of connection edges.

[0025] It should be explained in detail that physical activity is defined as the sum of the in-degree weights and out-degree weights of the second type of connecting edges, and the summation range is all second type of in-degree and out-degree edges related to the nodes in the graph; virtual influence is the sum of the out-degree weights of the fourth type of connecting edges; data feedback intensity is the sum of the in-degree weights of the fifth type of connecting edges, representing the intensity of the influence of physical data from the functional area on the node; virtual-real coupling is the sum of the connection weights of the node in the first type of connecting edges. Since the weight of the first type of connecting edges is always 1 in this embodiment, virtual-real coupling is represented as the sum of the number or existence of first type of connecting edges between the node and the corresponding virtual or real node; the above five types of indicators are normalized and summed according to the operational focus to synthesize the node correlation degree. If a certain type of indicator does not exist, that type of indicator is not calculated.

[0026] User flow centrality is obtained through the flow formula, where the flow formula is: ; In the formula For nodes The importance of nodes, i.e., the centrality of user flow; For the node To the node And passing through nodes The number of shortest paths; The number of all shortest paths; Path weight; ; In the formula, For normalized slave nodes To the node Number of users along the path; For normalized slave nodes To the node Average user value of the path; The normalized average dwell time is the average dwell time from the user's node. To the node The average total time spent at all nodes along the route; These are the weighting coefficients.

[0027] It should be explained in detail that this embodiment illustrates how to calculate the user flow centrality of each graph node based on the topology and weight information of the second and third type of connection edges in the charging station operation feedback graph network. This centrality is used to measure the importance of a node in user flow, i.e., the movement or interaction of users between functional areas and virtual functional areas. The calculation process includes: graph preprocessing, path weight construction, shortest path definition and counting, and UFC aggregation based on the shortest path count. For each source node Perform the extended Dijkstra algorithm and run the dependency loopback step of Brandes' algorithm to obtain the results for all objectives. of The cumulative contribution value will be added to each pair during the feedback process. Shortest path share and corresponding path weights Multiply and add to Above, after the calculation is completed, for Normalization is performed using the z-score method to facilitate comparisons between different time points or different sites. For nodes The importance of a node, i.e., the centrality of user flow, is measured in terms of its role in the flow of all node pairs. The importance of intervention; For normalized slave nodes To the node The number of users along the path, within the observation window, such as the past 30 days or the statistical period set by the business, is the number of independent users who pass through / reach the path. If the path consists of multiple edges, the number of users reached along any node or all nodes on the path is counted. For normalized slave nodes To the node The average value of users along the path, the average business value of these users, such as the average amount of money a user spends per charge at a charging station or the estimated contribution. The weighting coefficients are set to be non-negative and normalized, i.e. Based on data fitting and regression analysis of historical operational goals, this involves collecting actual observations on different paths (s,t) within historical time periods: the actual user flow efficiency of each path, and the corresponding... , , The data was analyzed using statistical methods such as multiple linear regression, with three normalized variables as independent variables and the target variable Y as the dependent variable. The standardized regression coefficients obtained after the fitting were used as the data. , , The reference values. This process reflects the actual contribution of various factors to circulation efficiency based on historical data mining; S3: Construct a strategy data sequence based on the charging station operation feedback graph network of each user; Each node in the user charging station operation feedback graph network is sorted according to the correlation index to form a node sorting sequence; node analysis is performed on each node in the node sorting sequence; the node analysis is the analysis of the user's strategy at each node according to the strategy corresponding to the historical fourth type of connection edge in time order, to obtain the strategy data sequence of the user at each node.

[0028] It needs to be explained in detail, such as Figure 4 As shown, the nodes in the operation feedback graph network of each user charging station are first sorted according to the correlation index to form a node sorting sequence. The correlation index here is a collective term, encompassing multiple predefined and calculated quantitative values. These indices include at least: physical activity, reflecting the frequency and value of user movement in physical space; virtual influence, characterizing the ability of a node to influence other nodes through historical strategies; data feedback intensity, measuring the intensity of data feedback received by a node from physical space; virtual-physical coupling, depicting the degree of binding between a node and its corresponding virtual or physical space; and user flow centrality, as detailed above, evaluating the pivotal role of a node in the process of users moving across functional areas. Each index is calculated based on a specific type of connection edge and its weight in the graph network using a corresponding algorithm model. For each node in the node sorting sequence, the following node analysis is performed to obtain the node's strategy data sequence: query the historical record library for all fourth-type connection edge historical records related to the node, where each record contains execution time and corresponding strategy information, such as strategy ID, execution parameters, and the identifier and value of the current conversion event; arrange the historical record set in ascending order of time to obtain the strategy data sequence, which is a sequence of strategies related to the charging station, such as the charging station coffee shop optimization strategy and the charging discount strategy.

[0029] S4: Obtain a value feedback data sequence based on the strategy data sequence, and perform strategy evaluation based on the value feedback data sequence; User conversion data is obtained based on the strategy data sequence; the conversion data is data obtained based on user conversion behavior, which is the user's behavior of interacting with the virtual functional area and then charging the physical device in the node sorting sequence; the conversion data includes instantaneous conversion data and continuous conversion data; the instantaneous conversion data is the behavior of not subsequently interacting with the virtual functional area and then charging the physical device under the corresponding strategy; the continuous conversion data is the behavior of continuously interacting with the virtual functional area and then charging the physical device under the corresponding strategy; the value feedback data sequence is obtained by sorting the conversion data according to the order of the node sorting sequence; the position difference and frequency of the same strategy in each value feedback data sequence are calculated, and the effectiveness of the strategy is evaluated based on the position difference and frequency.

[0030] It needs to be explained in detail, such as Figure 5 As shown, the conversion data is defined as follows: In this embodiment, conversion data refers to the observation information from user interaction in the virtual functional area to the occurrence of physical charging behavior after the execution of the fourth type of connection edge. The conversion data includes: conversion occurrence identifier, conversion time interval, i.e., the time interval from virtual interaction to physical charging, conversion value, such as actual charging amount or estimated contribution, and user identifier and session context; Instantaneous conversion: If the corresponding user occurs physical charging behavior within a preset short window after the strategy is executed, such as within 24 hours, and it is the first occurrence, it is marked as a preset instantaneous conversion; if the behavior does not occur again after the strategy, it is marked as instantaneous conversion data; Continuous conversion: When several consecutive strategies In the execution window, for example, if a user repeatedly engages in physical charging behavior after virtual interaction within at least three consecutive cycles or multiple triggers within 30 days, this is marked as continuous conversion data, and the continuous conversion rate and average continuous value are calculated. The instantaneous conversion data and continuous conversion data related to real-time charging in the strategy data sequence are sorted and concatenated according to the node sorting sequence to obtain a value feedback data sequence. This value feedback data sequence reflects the effect of each strategy on promoting user consumption. The value feedback data sequence is used to calculate the location difference and the number of occurrences. This embodiment provides an example method for evaluating the effectiveness of strategies based on the location difference and the number of occurrences: ;in Rate the effectiveness of the strategy; The normalized position difference of the same strategy in the value feedback data sequence; This represents the number of times the same strategy is normalized.

[0031] S5: Build and apply dynamic operational knowledge graphs to optimize strategy generation.

[0032] Based on the graph nodes, connecting edges, and correlation indicators in the charging station operation feedback graph network, and by integrating multi-source heterogeneous data such as charging pile operation status, user feedback text, and environmental parameters, a dynamic operation knowledge graph is constructed by extracting and defining knowledge nodes including operation entities, user entities, and strategy entities, as well as the relationships between these knowledge nodes. This dynamic operation knowledge graph is used for strategy reasoning, including generating readable natural language strategy descriptions based on the relationships between entities in the graph and the current operation status. The quality of the dynamic operation knowledge graph is dynamically evaluated and updated using preset operation indicators. The strategy schemes generated by the dynamic operation knowledge graph reasoning are matched and filtered in conjunction with historical user feature profiles to achieve differentiated strategy pushes to different user groups, such as price-sensitive and time-sensitive users.

[0033] It needs to be explained in detail that the data sources include, but are not limited to: charging pile operation status: structured time-series data of pile occupancy, charging power, charging duration, and fault logs; user feedback text: unstructured text of user reviews, customer service records, and social media comments; environmental parameters: meteorological data, events around the station, and external context of the time period; standardized preprocessing is performed on various types of data: time synchronization, missing value handling, unit conversion, text segmentation and entity candidate extraction, using domain dictionaries and rules, and mapping the data to nodes and edges in the charging station operation feedback graph network to supplement node and edge attributes; the following knowledge node types are extracted and defined from the graph network and fused data: operation entities: charging piles, functional areas, maintenance teams, and operation strategies; user entities: user profiles, price-sensitive, time-sensitive, or loyal users, user behavior patterns, and conversion history; strategy entities: strategy types, discounts, guidance, queuing optimization, and strategy effect evaluation results; and node definitions are defined. Attribute examples: Charging pile node attributes include current occupancy rate and failure rate; functional area node attributes include physical activity and virtual-physical coupling degree; user entities include historical average single charging value and preference tags; Extraction relationship type examples: Trigger: strategy entity to conversion event; Location: charging pile to functional area; Preference: user entity to strategy entity; Impact: environmental parameters to conversion probability; The above entities and relationships are stored in the form of triples (h, r, t, a), where h and t are the head and tail entity identifiers, r is the relationship type, and a is a relationship attribute vector containing timestamps and confidence levels; Support for time dimension expansion, i.e., recording the effective time interval for each triple for dynamic querying and time-series reasoning; Establishing indexing and caching mechanisms to ensure real-time reasoning performance, such as establishing memory indexes for frequently accessed entities; Rule-based reasoning: Based on predefined rules, if the physical activity of a functional area is greater than X and the virtual-physical coupling degree is greater than Y, it is recommended to add a guiding strategy to generate candidate strategies.

[0034] The candidate strategy set generated by inference is matched with user groups: using historical user characteristic profiles, such as price sensitivity, time sensitivity, or high loyalty, the matching degree of the candidate strategies is calculated. The matching degree is calculated based on a comprehensive analysis of historical success rate, user preference tag similarity, and strategy attributes. Strategies are then graded and screened according to the matching degree: high-matching-degree strategies are directly entered into the distribution queue; medium-matching-degree strategies are entered into small-scale trials; low-matching-degree strategies are only recorded offline. When a strategy is distributed, the knowledge graph simultaneously records the distribution context time, user group, and environmental parameters, so that it can be used as historical data input in the next round of evaluation, forming a closed loop.

[0035] Example 2: Intelligent generation system for charging station operation strategies based on user characteristics, such as Figure 2 As shown, it includes: The area division module is used to divide and construct the area of ​​the charging station, and obtain the functional area and virtual functional area of ​​the charging station; The network construction module is used to construct a charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area. The edge weight construction module is used to construct edge weights based on the connection edge types between nodes when building a charging station operation feedback graph network. The indicator calculation module is used to calculate the correlation index of each node based on the edge weights constructed by the connection edge types between each node in the charging station operation feedback graph network. The importance calculation module assigns the importance of nodes to the second and third types of connection edges based on user flow centrality. The strategy sequence construction module is used to construct strategy data sequences based on the charging station operation feedback graph network of each user. The strategy evaluation module is used to obtain a value feedback data sequence based on the strategy data sequence, and to perform strategy evaluation based on the value feedback data sequence. The strategy generation module is optimized to build and apply a dynamic operational knowledge graph to optimize strategy generation.

[0036] It should be understood that this embodiment is for illustrative purposes only and is not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

Claims

1. A method for intelligently generating charging station operation strategies based on user characteristics, characterized by: Includes the following steps: S1: Divide the area of ​​the charging station to obtain the charging station functional area and the virtual functional area, and construct the charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area; S2: Calculate the correlation index of each node based on the edge weights constructed according to the connection edge types between each node in the charging station operation feedback graph network. S3: Construct a strategy data sequence based on the charging station operation feedback graph network of each user; S4: Obtain a value feedback data sequence based on the strategy data sequence, and perform strategy evaluation based on the value feedback data sequence; S5: Build and apply dynamic operational knowledge graphs to optimize strategy generation.

2. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 1, characterized in that, The process involves dividing the charging station area into functional zones and virtual functional zones, and constructing a charging station operation feedback network for each user based on these zones. This includes: obtaining virtual functional zones corresponding to each functional zone, where each virtual functional zone represents the area where a user uses a virtual space within a functional zone; analyzing user usage patterns and constructing a first type of connection edge: constructing bidirectional connections between functional zones and corresponding virtual functional zones; constructing a second type of connection edge: constructing unidirectional or bidirectional connections based on user usage patterns within functional zones; and constructing a third type of connection edge: based on user... For interactions between virtual functional areas, establish unidirectional or bidirectional connections between virtual functional areas; construct a fourth type of connection: based on historical policy execution records, establish unidirectional connections from virtual functional areas to charging station functional areas; construct a fifth type of connection: based on physical space usage data feedback, establish unidirectional connections from charging station functional areas to non-corresponding virtual functional areas; treat each charging station functional area and its corresponding virtual functional area as graph nodes; construct a charging station operation feedback graph network for each user based on the bidirectional connections, the unidirectional connections, and the graph nodes; when constructing the charging station operation feedback graph network, construct edge weights according to the connection types between nodes.

3. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 2, characterized in that, The virtual functional area refers to the area in the charging station functional area where users use the corresponding virtual space, including: the virtual functional area is an online meeting area where multiple users have a common destination or an online meeting area where multiple users have an unconfirmed destination.

4. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 2, characterized in that, When constructing the charging station operation feedback graph network, edge weights are constructed based on the connection edge types between each node, including: the edge weight of the first type of connection edge is set to one, used to represent the basic mapping relationship; the edge weight of the second type of connection edge is obtained based on the number of times the user moves within the functional area of ​​the charging station, the average movement time, and the value generated during the movement; the edge weight of the third type of connection edge is obtained based on the frequency and complexity of the user's virtual interaction between virtual functional areas; the edge weight of the fourth type of connection edge is obtained based on the success rate and value of historical strategies; and the edge weight of the fifth type of connection edge is obtained based on the frequency and quality of data feedback.

5. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 1, characterized in that, The method of calculating the correlation index of each node based on the edge weights constructed according to the connection edge types between nodes in the charging station operation feedback graph network includes: when calculating the correlation index of each node, physical activity, virtual influence, user flow centrality, data feedback intensity, or virtual-real coupling degree; the physical activity is the sum of the in-degree weights and out-degree weights of the second type of connection edge; the virtual influence is the sum of the out-degree weights of the fourth type of connection edge; the user flow centrality is the node importance of the second type of connection edge and the third type of connection edge; the data feedback intensity is the sum of the in-degree weights of the fifth type of connection edge; and the virtual-real coupling degree is the sum of the connection weights of the node in the first type of connection edge.

6. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 5, characterized in that, The user flow centrality is the node importance of the second and third type of connection edges, including: obtaining the user flow centrality through the flow formula, wherein the flow formula is: ； In the formula For nodes The importance of nodes, i.e., the centrality of user flow; For the node To the node And passing through nodes The number of shortest paths; The number of all shortest paths; Path weight; ； In the formula, For normalized slave nodes To the node Number of users along the path; For normalized slave nodes To the node Average user value of the path; The normalized average dwell time is the average dwell time from the user's node. To the node The average total time spent at all nodes along the route; These are the weighting coefficients.

7. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 1, characterized in that, The step of constructing a strategy data sequence based on the charging station operation feedback graph network of each user includes: sorting each node in the charging station operation feedback graph network of each user according to the correlation index to form a node sorting sequence; performing node analysis on each node in the node sorting sequence; the node analysis is an analysis of the strategy of each user at each node according to the historical fourth type of connection edge in chronological order, to obtain the strategy data sequence of each user at each node.

8. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 1, characterized in that, The step of obtaining a value feedback data sequence based on a strategy data sequence and evaluating the strategy based on the value feedback data sequence includes: obtaining user conversion data based on the strategy data sequence; the conversion data is data obtained based on user conversion behavior, where the conversion behavior is the user's behavior of interacting with a virtual functional area and then charging a physical device in the node sorting sequence; the conversion data includes instantaneous conversion data and continuous conversion data; the instantaneous conversion data is the absence of subsequent behavior of interacting with a virtual functional area and then charging a physical device under the corresponding strategy; the continuous conversion data is the continuous occurrence of subsequent behavior of interacting with a virtual functional area and then charging a physical device under the corresponding strategy; sorting the conversion data according to the order of the node sorting sequence to obtain a value feedback data sequence; calculating the position difference and frequency of the same strategy among the value feedback data sequences, and evaluating the effectiveness of the strategy based on the position difference and frequency.

9. The intelligent generation method for charging station operation strategies based on user characteristics as described in claim 1, characterized in that, The construction and application of a dynamic operational knowledge graph to optimize strategy generation includes: based on the graph nodes, connecting edges, and correlation indicators in the charging station operation feedback graph network, integrating multi-source heterogeneous data such as charging pile operation status, user feedback text, and environmental parameters, extracting and defining knowledge nodes including operational entities, user entities, and strategy entities, as well as the relationships between knowledge nodes, to construct a dynamic operational knowledge graph; the dynamic operational knowledge graph is used for strategy reasoning, including reasoning and generating readable natural language strategy descriptions based on the relationships between entities in the graph and the current operational status; the quality of the dynamic operational knowledge graph is dynamically evaluated and updated through preset operational indicators; the strategy schemes generated by the dynamic operational knowledge graph reasoning will be matched and filtered in conjunction with historical user feature profiles to achieve differentiated strategy push to different user groups that are price-sensitive and time-sensitive.

10. A user-feature-based intelligent generation system for charging station operation strategies, used to implement the user-feature-based intelligent generation method for charging station operation strategies as described in any one of claims 1-9, characterized in that, include: The area division module is used to divide and construct the area of ​​the charging station, and obtain the functional area and virtual functional area of ​​the charging station; The network construction module is used to construct a charging station operation feedback graph network for each user based on the charging station functional area and the virtual functional area. The edge weight construction module is used to construct edge weights based on the connection edge types between nodes when building a charging station operation feedback graph network. The indicator calculation module is used to calculate the correlation index of each node based on the edge weights constructed by the connection edge types between each node in the charging station operation feedback graph network. The importance calculation module assigns the importance of nodes to the second and third types of connection edges based on user flow centrality. The strategy sequence construction module is used to construct strategy data sequences based on the charging station operation feedback graph network of each user. The strategy evaluation module is used to obtain a value feedback data sequence based on the strategy data sequence, and to perform strategy evaluation based on the value feedback data sequence. The strategy generation module is optimized to build and apply a dynamic operational knowledge graph to optimize strategy generation.

Images