Charging station comprehensive operation management method and system based on multi-dimensional user clustering

Abstract

The present application relates to the technical field of charging station operation optimization and resource allocation, and particularly relates to a charging station comprehensive operation management method and system based on multi-dimensional user grouping. The charging station comprehensive operation management method based on multi-dimensional user grouping comprises the following steps: S1: obtaining user charging demand data and user value-added demand data, and constructing a mutual information universe set according to the user charging demand data and the user value-added demand data; S2: determining a first mutual information set and a second mutual information set according to the mutual information universe set, and performing multi-dimensional user grouping and constructing a charging station operation dimension combination library based on the first mutual information set and the second mutual information set. The present application fuses user charging and value-added demand data, constructs a mutual information universe set for multi-dimensional grouping, introduces a directional demand conversion index, dynamically matches strategies, and realizes adaptive adjustment of resource allocation and closed-loop continuous optimization of a strategy library.

Description

Technical Field

[0001] This invention relates to the field of charging station operation optimization and resource allocation technology, and in particular to a comprehensive operation management method and system for charging stations based on multi-dimensional user segmentation. Background Technology

[0002] In existing charging station operation and management methods, refined operation strategies based on multi-dimensional user segmentation have become a key technological path for improving efficiency. Multi-dimensional user segmentation refers to dividing users into groups with different needs by analyzing multi-source data such as charging behavior, service preferences, and spatiotemporal characteristics. Based on this, comprehensive charging station operation and management methods aim to match differentiated charging services and value-added services to different user groups, thereby optimizing resource allocation and improving operational revenue. This development has effectively improved charging pile utilization and user satisfaction, laying the foundation for intelligent management of charging infrastructure.

[0003] However, existing user segmentation-based operational methods often fail quickly when faced with external disturbances such as sudden extreme weather events. Their static user profiles and optimal operational strategies, built upon historical patterns, often become ineffective. This is because special conditions cause collective, structural "mutations" rather than "fluctuations" in user behavior patterns, rendering the original segmentation results indistinguishable. The stable correlation between charging demand and value-added services demand is broken, ultimately leading to a severe mismatch between pre-set operational plans and real-time needs, and resource scheduling failure. This exposes a core deficiency in the adaptability of existing technologies to dynamic scenarios.

[0004] Therefore, there is an urgent need for a method that can sense sudden changes in the external environment, quantify structural changes in user demand, and dynamically adjust the optimal operation strategy accordingly to enhance the operational resilience of charging stations under uncertain conditions. Summary of the Invention

[0005] To overcome the shortcomings of operational strategy mismatch caused by sudden changes in user demand structure under special conditions, this invention provides a comprehensive operation and management method and system for charging stations based on multi-dimensional user segmentation.

[0006] The technical implementation scheme of the present invention is as follows: a comprehensive operation and management method for charging stations based on multi-dimensional user segmentation, comprising the following steps: S1: Obtain user charging demand data and user value-added demand data, and construct a complete set of mutual information based on the user charging demand data and user value-added demand data; S2: Determine the first mutual information set and the second mutual information set based on the complete set of mutual information, perform multi-dimensional user segmentation based on the first mutual information set and the second mutual information set, and construct a charging station operation dimension combination library; S3: Obtain special weather conditions and determine the user group integration degree. Based on the user group integration degree, calculate the user charging demand conversion degree and the user value-added demand conversion degree. S4: For each standard operation dimension combination in the charging station operation dimension combination library, perform conversion matching degree calculation and obtain comprehensive matching degree, and determine the comprehensive operation management method of the charging station based on the comprehensive matching degree.

[0007] Preferably, the acquisition of user charging demand data and user value-added demand data includes: The user charging demand data refers to a structured set of information describing all behavioral attributes, equipment status, energy exchange parameters, spatiotemporal constraints, and direct cost preferences involved in the process of a user replenishing their energy at a charging station. The user value-added demand data refers to a multimodal set of information that records users' behavioral choices, consumption characteristics, experience feedback, and payment preferences when using additional services in charging stations or related scenarios to meet non-charging-related life, consumption, or experience goals.

[0008] Preferably, the step of constructing a complete set of mutual information based on the user charging demand data and the user value-added demand data includes: User charging demand data is divided into N1 dimensions, and user value-added demand data is divided into N2 dimensions, where N1 is not equal to N2. Normalize each of the N1 dimensions of the user charging demand data to obtain normalized user charging demand data with N1 dimensions. Normalize each of the N2 dimensions of the user value-added needs data to obtain normalized user value-added needs data with N2 dimensions; Calculate the information entropy of each dimension in the N1 dimensions of the normalized user charging demand data, and determine the dimension with the largest information entropy as the main dimension. Calculate the mutual information between the main dimension and each of the N2 dimensions of the normalized user value-added needs data to obtain N2 mutual information values; Calculate the mutual information between each of the remaining dimensions in the N1 dimensions of the normalized user charging demand data and each of the N2 dimensions of the normalized user value-added demand data, and obtain (N1-1) multiplied by N2 mutual information values. The N2 mutual information values ​​obtained and the (N1-1) multiplied by N2 mutual information values ​​obtained are combined to form the complete set of mutual information for the charging demand dimension and the value-added demand dimension. The complete set of mutual information contains N1 multiplied by N2 mutual information values.

[0009] Preferably, determining the first mutual information set and the second mutual information set based on the complete mutual information set includes: For each of the N1 dimensions of the normalized user charging demand data, extract the mutual information value between the current dimension and all N2 dimensions of the normalized user value-added demand data from the mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current charging demand dimension; the average mutual information values ​​of the N1 dimensions of the normalized user charging demand data form the first mutual information set. For each of the N2 dimensions of normalized user value-added demand data, extract the mutual information value between the current dimension and all N1 dimensions of normalized user charging demand data from the mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current value-added demand dimension; the average mutual information values ​​of the N2 dimensions of normalized user value-added demand data form the second mutual information set.

[0010] Preferably, the step of performing multi-dimensional user segmentation based on the first mutual information set and the second mutual information set and constructing a charging station operation dimension combination library includes: Based on the first mutual information set and the second mutual information set, a clustering algorithm is used to perform multi-dimensional user grouping to obtain multi-dimensional user grouping results; the multi-dimensional user grouping results include the group label of each user, the charging demand dimension feature vector and the value-added demand dimension feature vector of each group; For each charging station, count the number of users in each user group within the current charging station, calculate the proportion distribution of each user group, and identify the user group with the largest proportion as the main user group of the current charging station. Extract the dimension with the largest value from the feature vector of the charging demand dimension of the main user group as the optimal dimension for charging service of the current charging station; extract the dimension with the largest value from the feature vector of the value-added demand dimension of the main user group as the optimal dimension for value-added service of the current charging station. The optimal dimension of charging service and the optimal dimension of value-added service together constitute a single optimal combination of operating dimensions for the current charging station; The charging station operation dimension combination library contains K standard operation dimension combinations; each standard operation dimension combination contains a standard charging service dimension and a standard value-added service dimension; the standard operation dimension combination is generated in the following way: collect the single optimal operation dimension combination of K charging stations at different operation stages, remove duplicate combinations to form a standard operation dimension combination set.

[0011] Preferably, the step of acquiring special weather conditions and determining user group integration includes: Obtain special weather conditions, including heavy rain, heavy snow, high temperature, strong wind and freezing; Acquire user charging demand data and user value-added demand data at current charging stations under special weather conditions; Based on user charging demand data and user value-added demand data under special weather conditions, the same method as under normal conditions is used to perform multi-dimensional user grouping to obtain user grouping results under special weather conditions. Calculate the user group fusion degree, which is used to measure the degree of convergence of user behavioral characteristics under special weather conditions; The formula for calculating the user group integration degree is: ,in, To improve the integration of user groups, This represents the sum of inter-class distances for user groups under special weather conditions. This represents the sum of inter-class distances for user groups under normal conditions; the sum of inter-class distances is calculated by summing the pairwise Euclidean distances between all group center points. The value range of F is [0,1]. The closer the F value is to 1, the more similar the user behaviors are. The closer the F value is to 0, the more obvious the differences in user behaviors are.

[0012] Preferably, the step of calculating the conversion degree of user charging demand and the conversion degree of user value-added demand based on user segmentation integration degree includes: The steps for calculating the user charging demand conversion degree are as follows: Calculate the average charging demand feature vector of all users at the current charging station under special weather conditions; calculate the average charging demand feature vector of all users at the current charging station under normal conditions; the user charging demand conversion degree vector equals the average charging demand feature vector under special weather conditions minus the average charging demand feature vector under normal conditions; in the user charging demand conversion degree vector, a positive value for any component indicates that charging demand is stronger under special weather conditions compared to normal conditions in the corresponding dimension, while a negative value indicates a weakening; obtain the directional conversion degree component of user charging demand by multiplying the user group integration degree by the components of the user charging demand conversion degree vector on the preset key dimensions of charging services. ; The steps for calculating the conversion degree of user value-added needs are as follows: Calculate the average value-added needs feature vector of all users at the current charging station under special weather conditions; calculate the average value-added needs feature vector of all users at the current charging station under normal conditions; the user value-added needs conversion degree vector equals the average value-added needs feature vector under special weather conditions minus the average value-added needs feature vector under normal conditions; in the user value-added needs conversion degree vector, a positive value for any component indicates that the value-added needs under special weather conditions are stronger than under normal conditions in the corresponding dimension, while a negative value indicates a weakening; the directional conversion degree component of user value-added needs is obtained by multiplying the user group integration degree and the components of the user value-added needs conversion degree vector on the preset key dimensions of value-added services. .

[0013] Preferably, for each standard operational dimension combination in the charging station operational dimension combination library, performing a conversion matching degree calculation and obtaining a comprehensive matching degree includes: Charging demand conversion matching degree calculation: Obtain the value of the optimal dimension of normal charging service of the current charging station in the feature vector of charging demand dimension. ; Obtain the values ​​of the standard charging service dimension in the feature vector of the charging demand dimension, which is a combination of standard operation dimensions. Charging demand conversion matching degree : ,in, For cosine similarity, This is the conversion degree influence coefficient, with a value range of [0,1]. Value-added demand conversion matching degree calculation: Obtain the value of the optimal dimension of the current charging station's normal value-added services in the feature vector of the value-added demand dimension. ; Obtain the values ​​of the standard value-added service dimensions in the feature vector of the value-added demand dimensions, which represent the combination of standard operational dimensions. Value-added demand conversion matching degree : ,in This is the conversion degree influence coefficient, with a value range of [0,1]. The overall matching degree is the average of the matching degree of charging demand conversion and the matching degree of value-added demand conversion.

[0014] Preferably, the method for determining the comprehensive operation and management of charging stations based on the comprehensive matching degree includes: All standard operational dimension combinations are sorted by comprehensive matching degree, and the combination with the highest comprehensive matching degree is selected as the recommended operational dimension combination under special weather conditions. Charging station resource allocation and service strategies are adjusted based on the recommended operational dimension combination. During the duration of special weather conditions, user segmentation integration degree, user charging demand conversion degree, and user value-added demand conversion degree are periodically recalculated and the recommended operational dimension combination is updated. Charging station operation strategies are adjusted in real time based on the dynamically updated recommended operational dimension combination. After the special weather conditions end, the normal operational dimension combination of the charging station is restored. Matching degree data and operational effect data under this special weather condition are recorded, and the charging station operational dimension combination library and conversion degree impact coefficient are updated.

[0015] Preferably, the charging station integrated operation and management system based on multi-dimensional user segmentation includes: The user charging and value-added data collection module is used to acquire user charging demand data and user value-added demand data, and to construct a complete set of mutual information between the charging demand dimension and the value-added demand dimension. The multi-dimensional user segmentation and combination library construction module is used to determine the first mutual information set and the second mutual information set based on the complete mutual information set, perform multi-dimensional user segmentation, and construct a charging station operation dimension combination library. The Special Weather Demand Conversion Analysis Module is used to obtain special weather conditions and determine the user group integration degree, and calculate the user charging demand conversion degree and the user value-added demand conversion degree. The comprehensive matching degree decision and operation module is used to calculate the conversion matching degree and comprehensive matching degree of each standard operation dimension combination, and determine the comprehensive operation and management method of the charging station based on the comprehensive matching degree.

[0016] Beneficial Effects: This invention integrates user charging and value-added demand data to construct a complete set of mutual information that quantifies the relationship between the two. Based on average mutual information, it builds a multi-dimensional user segmentation and operational dimension combination library, enabling the extraction of standardized strategy knowledge from historical data. For external disturbances such as severe weather, this invention innovatively introduces user segmentation integration degree and directional demand conversion degree components, which can not only accurately perceive structural changes in user behavior but also quantify the direction of change in key operational dimensions. Furthermore, by calculating the conversion matching degree between preset strategies and the current demand state, which combines directional consistency and historical fit, strategies that adapt to change and suppress counter-change are intelligently rewarded, thereby dynamically selecting and activating the optimal operational dimension combination, achieving real-time adaptive adjustment of resource allocation and service strategies. Ultimately, this invention not only significantly enhances the operational resilience and precise response capability of charging stations in uncertain environments but also achieves continuous self-optimization of the strategy library and decision parameters through a closed-loop learning mechanism. Attached Figure Description

[0017] Figure 1 This is a flowchart of the integrated operation and management method for charging stations based on multi-dimensional user segmentation, as described in this invention. Figure 2 This is a structural diagram of the charging station integrated operation and management system based on multi-dimensional user segmentation, as described in this invention. Detailed Implementation

[0018] The present invention will be further described below with reference to specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0019] Example 1: A comprehensive operation and management method for charging stations based on multi-dimensional user segmentation, such as... Figure 1 As shown, it includes the following steps: S1-1: Obtain user charging demand data and user value-added demand data, including: The user charging demand data refers to a structured set of information describing all behavioral attributes, equipment status, energy exchange parameters, spatiotemporal constraints, and direct cost preferences involved in the process of a user replenishing their energy at a charging station. The user value-added demand data refers to a multimodal set of information that records users' behavioral choices, consumption characteristics, experience feedback, and payment preferences when using additional services in charging stations or related scenarios to meet non-charging-related life, consumption, or experience goals.

[0020] It's important to note that improving the operational efficiency of charging stations relies on refined insights and modeling of user behavior. Therefore, the primary task of this step is to build a complete data collection system, the core of which is to acquire and clearly define "user charging demand data" and "user value-added demand data." Charging demand data is mainly collected from charging pile operation logs, vehicle communication units, and user application software, specifically including the initial charging power level, the selected charging power level, the time window of the charging service, the complete charging process duration, and the settlement fee. Value-added demand data is collected through transaction records in the station's service system, user-submitted evaluation information, and auxiliary IoT sensing devices. This data details the user's occupancy time in rest facilities, consumption details at retail outlets, the types of food purchased, and feedback on the services received.

[0021] Traditional operational models often focus on the single transaction of power replenishment, failing to incorporate the diverse consumption behaviors of users within the site into a unified analytical framework. This lack of data dimension renders pre-set operational plans ineffective when faced with sudden events like severe weather, where user behavior deviates from historical norms. This step, by simultaneously acquiring and standardizing two types of structured and multimodal datasets—one characterizing core functions and the other representing derivative service needs—provides comprehensive information input for subsequent in-depth analysis of the coupling relationship between the two. This approach enables operational decisions to be based on an understanding of the complete user consumption journey, thereby supporting the system in developing more resilient resource allocation and service management strategies in complex and ever-changing environments.

[0022] S1-2: Based on the user charging demand data and user value-added demand data, a complete set of mutual information is constructed, including: User charging demand data is divided into N1 dimensions, and user value-added demand data is divided into N2 dimensions, where N1 is not equal to N2. Normalize each of the N1 dimensions of the user charging demand data to obtain normalized user charging demand data with N1 dimensions. Normalize each of the N2 dimensions of the user value-added needs data to obtain normalized user value-added needs data with N2 dimensions; Calculate the information entropy of each dimension in the N1 dimensions of the normalized user charging demand data, and determine the dimension with the largest information entropy as the main dimension. Calculate the mutual information between the main dimension and each of the N2 dimensions of the normalized user value-added needs data to obtain N2 mutual information values; Calculate the mutual information between each of the remaining dimensions in the N1 dimensions of the normalized user charging demand data and each of the N2 dimensions of the normalized user value-added demand data, and obtain (N1-1) multiplied by N2 mutual information values. The N2 mutual information values ​​obtained and the (N1-1) multiplied by N2 mutual information values ​​obtained are combined to form the complete set of mutual information for the charging demand dimension and the value-added demand dimension. The complete set of mutual information contains N1 multiplied by N2 mutual information values.

[0023] It's important to note that, to address the limitation of traditional operating models where the dynamic relationship between user charging behavior and derivative service demands is difficult to quantify, the core of this stage lies in establishing a quantitative foundation capable of systematically measuring the statistical dependency between the two—namely, the complete set of mutual information. In practice, the raw data is first decomposed into a structured framework. Charging demand data is broken down into multiple analytical dimensions, denoted as N1, including charging time preference, charging power selection, and initial power level. Similarly, value-added demand data is decomposed into multiple analytical dimensions, denoted as N2, including lounge usage tendency, food and beverage consumption level, and product browsing time. By applying the min-max normalization method to these dimensions, differences in the units and ranges of measurement among different indicators are eliminated, paving the way for subsequent unified calculations.

[0024] Next, we need to identify the most valuable features for analysis in charging demand. Here, we introduce information entropy as a metric. The higher the information entropy of a dimension, the more dispersed and uncertain the user's behavior is in that dimension, and thus the richer the differential information it contains. After calculating the information entropy of all charging demand dimensions, the dimension with the highest entropy value is determined as the primary dimension, because the primary dimension best represents the overall heterogeneity of user charging behavior. Subsequently, we use mutual information as a tool to quantify the correlation. The magnitude of mutual information directly reflects the amount of shared information between two random variables; the higher the value, the stronger the trend of coordinated change between the two. Calculating the mutual information between the primary dimension and each value-added demand dimension aims to explore the initial correlation between core charging features and various value-added services. To obtain a global correlation view, we also need to calculate the mutual information between the remaining charging dimensions and all value-added dimensions.

[0025] Finally, all calculated mutual information values ​​are aggregated to form a complete mutual information set. This set is essentially a detailed quantitative correlation network, where each specific value indicates the coupling strength between a particular pair of charging and value-added dimensions. For example, a high value clearly shows a significant correlation between "initial charging capacity" and "the likelihood of a user purchasing emergency goods." This quantitative network provides indispensable analytical basis for subsequently identifying overall shifts in user demand structure during sudden changes in the external environment and dynamically adjusting operational strategies accordingly.

[0026] In practice, for continuous variables or discretized variables, the probability distribution is estimated using histogram-based statistical methods, and then mutual information is calculated; or the K-nearest neighbor nonparametric estimation method is used.

[0027] S2-1: Determine the first mutual information set and the second mutual information set based on the complete set of mutual information, including: For each of the N1 dimensions of the normalized user charging demand data, extract the mutual information value between the current dimension and all N2 dimensions of the normalized user value-added demand data from the mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current charging demand dimension; the average mutual information values ​​of the N1 dimensions of the normalized user charging demand data form the first mutual information set. For each of the N2 dimensions of normalized user value-added demand data, extract the mutual information value between the current dimension and all N1 dimensions of normalized user charging demand data from the mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current value-added demand dimension; the average mutual information values ​​of the N2 dimensions of normalized user value-added demand data form the second mutual information set.

[0028] It should be noted that, addressing the limitations of traditional operational strategies that become inaccurate under specific conditions due to the break in demand correlation, this phase introduces a correlation strength condensation method based on average mutual information. In practice, each charging demand dimension undergoes mutual information calculation with all value-added demand dimensions, and these values ​​are then averaged. This average essentially measures the comprehensive impact of that charging dimension on various value-added services. For example, if the average mutual information value of the charging power dimension is prominent, it suggests a general correlation between the user's chosen charging speed and their subsequent consumption intentions. The first mutual information set, composed of the averages of all charging dimensions, identifies key indicators of the service-oriented potential inherent in charging behavior. In parallel, each value-added demand dimension also calculates its average mutual information with all charging dimensions; this value is used to assess the overall sensitivity of the service to the user's charging behavior patterns. The second mutual information set, formed by these averages, focuses on filtering service items closely coupled with the charging process. These two sets, based on the charging origin and service endpoint respectively, construct complementary analytical perspectives: the former quantifies the driving effect of charging characteristics on service consumption, while the latter quantifies the responsiveness of service items to charging characteristics. The establishment of this two-way quantitative framework provides a key measurement basis for the system to keenly identify the overall changes in user demand interaction patterns when facing external disturbances.

[0029] S2-2: Based on the first mutual information set and the second mutual information set, perform multi-dimensional user segmentation and construct a charging station operation dimension combination library, including: Based on the first mutual information set and the second mutual information set, a clustering algorithm is used to perform multi-dimensional user grouping to obtain multi-dimensional user grouping results; the multi-dimensional user grouping results include the group label of each user, the charging demand dimension feature vector and the value-added demand dimension feature vector of each group; For each charging station, count the number of users in each user group within the current charging station, calculate the proportion distribution of each user group, and identify the user group with the largest proportion as the main user group of the current charging station. Extract the dimension with the largest value from the feature vector of the charging demand dimension of the main user group as the optimal dimension for charging service of the current charging station; extract the dimension with the largest value from the feature vector of the value-added demand dimension of the main user group as the optimal dimension for value-added service of the current charging station. The optimal dimension of charging service and the optimal dimension of value-added service together constitute a single optimal combination of operating dimensions for the current charging station; The charging station operation dimension combination library contains K standard operation dimension combinations; each standard operation dimension combination contains a standard charging service dimension and a standard value-added service dimension; the standard operation dimension combination is generated in the following way: collect the single optimal operation dimension combination of K charging stations at different operation stages, remove duplicate combinations to form a standard operation dimension combination set.

[0030] It should be noted that, to address the issue of traditional static clustering models losing their guiding role under external environmental disturbances, this stage focuses on deriving practically valuable user segmentation solutions and strategy knowledge bases from quantitatively correlated data. The entire process begins with the automated classification of users. Using clustering analysis, users are divided based on their individual characteristic profiles defined by the first and second mutual information sets, forming user groups with distinct behavioral differences, such as "frequent fast charging and preference for simple meals" and "regular slow charging and emphasis on rest." The group labels assigned to each user serve as classification identifiers, while the charging and value-added service demand feature vectors for each group statistically depict the group's concentration tendency in these two types of demands. The clustering analysis is performed on an individual user basis; each user's characteristics are composed of normalized data in the N1 dimensions of charging demand and normalized data in the N2 dimensions of value-added service demand, forming a (N1+N2)-dimensional feature vector. The first and second mutual information sets are used to assist in understanding group characteristics but are not directly used as clustering input.

[0031] Focusing on the operation of a specific charging station, analyzing the proportion of users in each group becomes crucial. The calculated proportions directly reflect the actual structure of user demand at that station. The user group with the highest proportion is defined as the primary user segment, clearly identifying the target group for service resource allocation. To translate the common needs of the primary user segment into specific operational guidance, it's necessary to identify the strongest features from its feature vector. The dimension with the highest value in each dimension of the charging demand feature vector is established as the optimal dimension for charging services, and the dimension with the highest value in each dimension of the value-added demand feature vector is established as the optimal dimension for value-added services. The coupling of these two dimensions defines a highly targeted "single optimal operational dimension combination," which clearly indicates the service focus and resource allocation direction that the station should strengthen most during its normal operating cycle.

[0032] Furthermore, to generate transferable and reusable strategic knowledge, a standardized operational dimension combination library needs to be built. High-performing new dimension combinations, after validation, are added to the operational dimension combination library to expand strategy selection; simultaneously, historical combinations that have not been matched for a long time or whose matching effect is consistently below average are moved to a secondary library or marked as pending observation. The generation of this knowledge base relies on collecting and deduplicating single optimal operational dimension combinations from a wide range of sources, ultimately culminating in a series of standard paradigms such as "peak-hour fast charging paired with fast food supply" and "off-peak charging combined with retail experience." The establishment of this combination library provides the operations management system with a wealth of strategic contingency plans, enabling it to quickly match and activate the most suitable response from this pre-built knowledge base based on real-time diagnostics when facing special weather emergencies, thereby effectively enhancing the agility and scientific nature of operational decision-making.

[0033] The number of clusters K is determined automatically using the silhouette coefficient method combined with business interpretability constraints. The silhouette coefficient is used to quantify the clustering effect. The K value that maximizes the coefficient is selected as the technical candidate, and its value is agreed to be between 3 and 8 to ensure that the number of generated user groups is both distinguishable and convenient for the formulation and execution of operational strategies.

[0034] S3-1: Obtain special weather conditions and determine user group integration, including: Obtain special weather conditions, including heavy rain, heavy snow, high temperature, strong wind and freezing; Acquire user charging demand data and user value-added demand data at current charging stations under special weather conditions; Based on user charging demand data and user value-added demand data under special weather conditions, the same method as under normal conditions is used to perform multi-dimensional user grouping to obtain user grouping results under special weather conditions. Calculate the user group fusion degree, which is used to measure the degree of convergence of user behavioral characteristics under special weather conditions; The formula for calculating the user group integration degree is: ,in, To improve the integration of user groups, This represents the sum of inter-class distances for user groups under special weather conditions. This represents the sum of inter-class distances for user groups under normal conditions; the sum of inter-class distances is calculated by summing the pairwise Euclidean distances between all group center points. The value range of F is [0,1]. The closer the F value is to 1, the more similar the user behaviors are. The closer the F value is to 0, the more obvious the differences in user behaviors are.

[0035] It's important to note that this step focuses on addressing the core issue of static user segmentation models becoming inaccurate under sudden environmental changes. This is achieved by quantitatively characterizing the convergence effect of user behavior triggered by special weather conditions. This is done by integrating external meteorological data sources to obtain real-time information on clearly defined special weather events such as heavy rain and blizzards. Under this premise, "user segmentation convergence degree" is introduced as a core diagnostic variable. This variable is designed to measure the degree of convergence in user behavior patterns under stress: a convergence degree value close to 1 indicates a high degree of homogeneity in user needs, while a value close to 0 means that significant differences still exist.

[0036] The calculation of cluster integration degree relies on a comparative formula. The denominator is the cumulative Euclidean distance between the centroids of various user groups obtained through cluster analysis under normal conditions, essentially characterizing the dispersion or diversity baseline of user demand structure during a stable operation phase. The numerator corresponds to the sum of inter-class distances generated after re-clustering under special weather conditions. If a weather shock causes the originally dispersed user behavior patterns to converge at a single point, the numerator will drop sharply, leading to... The ratio decreases, ultimately driving the degree of fusion. Climbing towards 1. This calculation process transforms the qualitative observation of "behavioral convergence" into precise numerical signals, enabling the operating system to automatically and promptly detect moments when fundamental changes occur in the structure of user needs, providing quantifiable trigger conditions for subsequent dynamic switching of operating strategies.

[0037] User group integration This needs to be used in conjunction with a dynamically calibrated threshold to determine the "high convergence" of behavior and trigger a strategy switch. The dynamically calibrated threshold is obtained and updated through the following specific steps: Initial threshold calculation: Collect user grouping results for all normal days (without special weather) in the history of this charging station, calculate the "simulated convergence" value for each day according to the convergence calculation formula, and form a dataset. Take the 95th percentile of the values ​​in this dataset as the initial threshold. Dynamic optimization mechanism: After each special weather event, evaluate the timeliness and final operational effect of the strategy switch under the threshold adopted this time. If the effect is not good (such as switching too early or too late), fine-tune it in steps of 0.05 within the range of [initial threshold - 0.15, initial threshold + 0.10], and record the threshold with the best performance. After several events, the threshold will converge to the optimal value that adapts to the user behavior characteristics of this station.

[0038] S3-2: Based on user segmentation and integration degree, calculate the conversion degree of user charging demand and the conversion degree of user value-added demand, including: The steps for calculating the user charging demand conversion degree are as follows: Calculate the average charging demand feature vector of all users at the current charging station under special weather conditions; calculate the average charging demand feature vector of all users at the current charging station under normal conditions; the user charging demand conversion degree vector equals the average charging demand feature vector under special weather conditions minus the average charging demand feature vector under normal conditions; in the user charging demand conversion degree vector, a positive value for any component indicates that charging demand is stronger under special weather conditions compared to normal conditions in the corresponding dimension, while a negative value indicates a weakening; obtain the directional conversion degree component of user charging demand by multiplying the user group integration degree by the components of the user charging demand conversion degree vector on the preset key dimensions of charging services. ; The steps for calculating the conversion degree of user value-added needs are as follows: Calculate the average value-added needs feature vector of all users at the current charging station under special weather conditions; calculate the average value-added needs feature vector of all users at the current charging station under normal conditions; the user value-added needs conversion degree vector equals the average value-added needs feature vector under special weather conditions minus the average value-added needs feature vector under normal conditions; in the user value-added needs conversion degree vector, a positive value for any component indicates that the value-added needs under special weather conditions are stronger than under normal conditions in the corresponding dimension, while a negative value indicates a weakening; the directional conversion degree component of user value-added needs is obtained by multiplying the user group integration degree and the components of the user value-added needs conversion degree vector on the preset key dimensions of value-added services. .

[0039] It's important to note that the core objective of this step is to quantify and characterize the structural changes in user demand caused by special weather conditions, particularly the directional changes in key operational dimensions, in order to overcome the failure of traditional operational models in such scenarios. Specifically, user data collected during special weather periods and under historical normal conditions is aggregated and analyzed to generate corresponding average charging demand feature vectors. The average vector under special weather conditions encapsulates the common charging behavior characteristics of users during that period, while the average vector under normal conditions constitutes a stable demand baseline. By calculating the difference between the two average vectors, a user charging demand transformation vector is obtained. This vector clearly indicates the specific direction and dimensional components of the user's charging behavior migrating from the baseline state to the stress state in a multi-dimensional feature space.

[0040] To focus multidimensional changes on the key dimensions most instructive for operational decisions and to integrate the convergence of user behavior, this step introduces the indicator of "user charging demand directional conversion degree component". The "preset key dimension of charging service" is the "optimal dimension of charging service" determined for the current charging station in step S2-2. First, based on business logic or historical data analysis, a key dimension of charging service is preset (e.g., "charging power preference" dimension). Then, component values ​​on this key dimension are extracted from the user charging demand conversion degree vector. The sign of the component value directly indicates whether user demand is stronger (positive value) or weaker (negative value) compared to the normal state on that dimension. Finally, the user charging demand directional conversion degree component is calculated, and its value is the product of the user group integration degree and the component.

[0041] This calculation relationship contains important business logic: only when user demand shows a clear shift in direction (large absolute value of the component) in key dimensions, and the user group exhibits a high degree of convergence in this change (large user segmentation integration value), will the absolute value of the directional shift in user charging demand significantly increase, thus indicating a clear and widespread shift in demand paradigm. The sign (positive or negative) of the directional shift in user charging demand indicates the direction of change in key demand, and its absolute value quantifies the intensity and consensus of this directional change. Following the exact same principles and steps, the directional shift in user value-added demand is calculated in parallel. The resulting two directional shift components constitute a precise measure of the direction and intensity of demand shifts in charging and value-added services in key dimensions, providing a direct and quantitative basis for subsequent intelligent selection of the most suitable operational strategy from the contingency plan library.

[0042] S4-1: For each standard operational dimension combination in the charging station operational dimension combination library, perform a conversion matching degree calculation and obtain a comprehensive matching degree, including: Charging demand conversion matching degree calculation: Obtain the value of the optimal dimension of normal charging service of the current charging station in the feature vector of charging demand dimension. ; Obtain the values ​​of the standard charging service dimension in the feature vector of the charging demand dimension, which is a combination of standard operation dimensions. Charging demand conversion matching degree : ,in, For cosine similarity, This is the conversion degree influence coefficient, with a value range of [0,1]. Value-added demand conversion matching degree calculation: Obtain the value of the optimal dimension of the current charging station's normal value-added services in the feature vector of the value-added demand dimension. ; Obtain the values ​​of the standard value-added service dimensions in the feature vector of the value-added demand dimensions, which represent the combination of standard operational dimensions. Value-added demand conversion matching degree : ,in This is the conversion degree influence coefficient, with a value range of [0,1]. The overall matching degree is the average of the matching degree of charging demand conversion and the matching degree of value-added demand conversion.

[0043] It should be noted that this step aims to address the issue of inaccurate adaptation of static strategy libraries when demand changes abruptly. It uses quantitative calculations to evaluate the degree of matching between each pre-defined solution and the current specific conditions. The charging demand transformation matching degree measures the fit between the charging strategy in the standard combination and the current situation; a higher value indicates a better fit.

[0044] When calculating the matching degree of charging demand conversion, two key factors are considered simultaneously: first, the degree of alignment between the strategy and the site's historical advantages; and second, the consistency between the strategy's direction and the current direction of key demand changes. The specific calculation formula is as follows: .in, It is the cosine similarity between the preset charging strategy and the optimal charging dimension characteristics of the station under normal conditions, used to measure the historical consistency. The aforementioned calculation shows that the directional conversion degree component of user charging demand is a scalar with positive and negative values. User group integration Components of the charging demand conversion vector on preset key dimensions Multiply them to get the result. The sign (positive / negative) indicates whether user demand is stronger or weaker than normal in key dimensions, and its absolute value indicates the strength of the directional change and the degree of user consensus.

[0045] coefficient (0≤ ≤1) is the core adjustment parameter, determining the weighting of historical consistency and current directional change. This design embodies a clear operational logic: under special weather conditions, the coefficient... The increasing trend is causing decision-making to shift from relying on historical experience to dynamically adapting to current changes. The item constitutes an intelligent reward / punishment mechanism: when This item is positive when >0 (critical demand enhancement). If the strategy also tends to enhance this service (…), then… If the similarity calculation is too large (leading to an implicit tendency towards this), then this item will improve the matching degree. This serves to reward those who adapt to change. When the value is less than 0 (indicating weakened critical demand), this item is negative. If the strategy still favors strengthening the service, this item will reduce the matching degree. This has the effect of suppressing adverse changes.

[0046] This automatically filters out operational strategies that are not only historically effective but also proactively adapt to or at least do not contradict key changes in user needs under current special conditions.

[0047] Value-added demand conversion matching degree This is used to symmetrically evaluate the value-added service strategy in the standard combination, and its calculation formula and design logic are the same. Exactly the same: .in, To measure the similarity between the preset value-added strategies and the site's historical advantages, It is the component of the directional conversion degree of user value-added needs, and the coefficient. Play a role with The same regulatory effect. This formula ensures that the system can effectively identify and prioritize value-added service options that align with the changing direction of value-added demand under special weather conditions.

[0048] The conversion degree influence coefficient and The determination and update methods are as follows: Coefficient initialization (reverse calibration): retrieve historical data. Data on the first completed special weather event. For the first... For this event, the optimal operational strategy is known. We initialize it by solving an optimization problem. and Minimize the loss function ,in, For loss function, The total number of historical events. For indexing historical events, The actual optimal strategy in the j-th event is based on the current... , The calculated overall matching degree. A grid search method is used to find the value that minimizes L within the interval [0,1] with a step size of 0.1. , The initial values ​​are a combination of these. Coefficients are updated online (fine-tuned): after each special weather event, a new set of data is obtained. , (The actual optimal strategy). Using this data as a training sample, online stochastic gradient descent is employed to... and Perform an iterative update with a learning rate of 0.01, aiming to minimize the difference between the "predicted optimal policy" and the "actual optimal policy" in this event.

[0049] S4-2: Determine the comprehensive operation and management method for charging stations based on the aforementioned comprehensive matching degree, including: All standard operational dimension combinations are sorted by comprehensive matching degree, and the combination with the highest comprehensive matching degree is selected as the recommended operational dimension combination under special weather conditions. Charging station resource allocation and service strategies are adjusted based on the recommended operational dimension combination. During the duration of special weather conditions, user segmentation integration degree, user charging demand conversion degree, and user value-added demand conversion degree are periodically recalculated and the recommended operational dimension combination is updated. Charging station operation strategies are adjusted in real time based on the dynamically updated recommended operational dimension combination. After the special weather conditions end, the normal operational dimension combination of the charging station is restored. Matching degree data and operational effect data under this special weather condition are recorded, and the charging station operational dimension combination library and conversion degree impact coefficient are updated.

[0050] It's important to note that this step aims to transform the quantitative matching signals generated in the preliminary analysis into dynamic operational instructions for execution, thereby addressing the issue of rigid pre-set strategies. The decision-making basis directly stems from the ranking results of the comprehensive matching degree; the combination of standard operational dimensions with the highest score is selected as the execution plan, as it has been calculated and verified to be most compatible with the current demand structure changes. Subsequently, the operations system initiates specific resource allocation and service strategy adjustment processes based on the recommended combination. For example, for a recommended combination emphasizing "extremely high charging power and convenient hot beverage supply," on-site management will immediately improve the power availability of fast charging stations and increase the storage and display of portable food.

[0051] Given the fluctuating development stages of special weather and the varying user responses, decisions are not made on a one-off basis, but rather through a periodic reassessment mechanism. The reassessment cycle is pre-set based on the event type or dynamically triggered according to the rate of change in real-time data streams. By periodically recalculating core metrics and updating recommended combinations, rolling optimization and dynamic correction of operational strategies are achieved, ensuring that action plans remain synchronized with the real-time context. The "periodic recalculation" cycle is not fixed. Its base length is set based on the specific weather type and dynamically scales according to the rate of change in user segment integration: the cycle is shortened when user behavior rapidly converges for agile response, and extended when behavior stabilizes to conserve computing power.

[0052] Once the special weather conditions are lifted, pre-set instructions are executed to revert the charging station's operational focus back to the optimal combination of dimensions under normal conditions. Key data generated throughout the entire response process, including the matching degree sequence at each time point and the final operational performance indicators, are fully recorded by the system. This data is used for subsequent model iterations: new dimension combinations that perform exceptionally well, after validation, are added to the operational dimension combination library to expand strategy selection; simultaneously, by analyzing the difference between the predicted matching degree values ​​and the actual results, the conversion degree influence coefficient can be adaptively calibrated, thereby improving the accuracy and reliability of future decisions in similar scenarios.

[0053] Example 2: Based on Example 1, a comprehensive operation and management system for charging stations based on multi-dimensional user segmentation, such as... Figure 2 As shown, it includes: The user charging and value-added data collection module is used to acquire user charging demand data and user value-added demand data, and to construct a complete set of mutual information between the charging demand dimension and the value-added demand dimension. The multi-dimensional user segmentation and combination library construction module is used to determine the first mutual information set and the second mutual information set based on the complete mutual information set, perform multi-dimensional user segmentation, and construct a charging station operation dimension combination library. The Special Weather Demand Conversion Analysis Module is used to obtain special weather conditions and determine the user group integration degree, and calculate the user charging demand conversion degree and the user value-added demand conversion degree. The comprehensive matching degree decision and operation module is used to calculate the conversion matching degree and comprehensive matching degree of each standard operation dimension combination, and determine the comprehensive operation and management method of the charging station based on the comprehensive matching degree.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A comprehensive operation and management method for charging stations based on multi-dimensional user segmentation, characterized by: Includes the following steps: S1: Obtain user charging demand data and user value-added demand data, including: the user charging demand data refers to a structured information set describing all behavioral attributes, equipment status, energy exchange parameters, spatiotemporal constraints, and direct cost preferences involved in the process of a user replenishing energy at a charging station; the user value-added demand data refers to a multimodal information set recording the behavioral choices, consumption characteristics, experience feedback, and payment preferences of users when using additional services in charging stations or related scenarios to meet non-charging-related life, consumption, or experience goals. A complete mutual information set is constructed based on the user charging demand data and user value-added demand data, including: dividing the user charging demand data into N1 dimensions and the user value-added demand data into N2 dimensions, where N1 is not equal to N2; and normalizing each of the N1 dimensions of the user charging demand data to obtain normalized user data. The charging demand data has N1 dimensions; the N2 dimensions of the user value-added demand data are normalized to obtain N2 normalized user value-added demand data; the information entropy of each dimension in the normalized user charging demand data is calculated, and the dimension with the largest information entropy is determined as the main dimension; the mutual information between the main dimension and each dimension in the normalized user value-added demand data is calculated to obtain N2 mutual information values; the mutual information between each remaining dimension in the normalized user charging demand data and each dimension in the normalized user value-added demand data is calculated to obtain (N1-1) multiplied by N2 mutual information values; the obtained N2 mutual information values ​​and the obtained (N1-1) multiplied by N2 mutual information values ​​are combined to form the complete set of mutual information between the charging demand dimension and the value-added demand dimension, and the complete set of mutual information contains N1 multiplied by N2 mutual information values. S2: Determine the first mutual information set and the second mutual information set based on the complete mutual information set, including: for each of the N1 dimensions of normalized user charging demand data, extract the mutual information value between the current dimension and all N2 dimensions of normalized user value-added demand data from the complete mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current charging demand dimension; the average mutual information values ​​of the N1 dimensions of normalized user charging demand data form the first mutual information set; for each of the N2 dimensions of normalized user value-added demand data, extract the mutual information value between the current dimension and all N1 dimensions of normalized user charging demand data from the complete mutual information set, calculate the average value of all mutual information values, and obtain the average mutual information value of the current value-added demand dimension; the average mutual information values ​​of the N2 dimensions of normalized user value-added demand data form the second mutual information set; perform multi-dimensional user grouping based on the first mutual information set and the second mutual information set and construct a charging station operation dimension combination library, including: based on the first mutual information set and the second mutual information set, use a clustering algorithm to perform multi-dimensional user grouping. The system generates multi-dimensional user segmentation results. These results include group tags for each user, and feature vectors for each group's charging demand and value-added service needs. For each charging station, the system counts the number of users in each user segment and calculates the proportional distribution of each segment. The user segment with the largest proportion is identified as the primary user segment of the current charging station. The system extracts the dimension with the largest value from the charging demand feature vector of the primary user segment as the optimal dimension for charging service at the current charging station. It also extracts the dimension with the largest value from the value-added service feature vector of the primary user segment as the optimal dimension for value-added service at the current charging station. The optimal charging service dimension and the optimal value-added service dimension together constitute a single optimal operational dimension combination for the current charging station. The charging station operational dimension combination library contains K standard operational dimension combinations. Each standard operational dimension combination includes a standard charging service dimension and a standard value-added service dimension. The standard operational dimension combinations are generated by collecting the single optimal operational dimension combinations of K charging stations at different operational stages, removing duplicate combinations, and then forming a set of standard operational dimension combinations. S3: Obtain special weather conditions and determine user group integration degree, including: obtaining special weather conditions, such as heavy rain, heavy snow, high temperature, strong wind, and freezing; obtaining user charging demand data and user value-added demand data at the current charging station under special weather conditions; based on the user charging demand data and user value-added demand data under special weather conditions, performing multi-dimensional user grouping using the same method as under normal conditions to obtain user grouping results under special weather conditions; calculating the user group integration degree, which measures the degree of convergence of user behavioral characteristics under special weather conditions; the formula for calculating the user group integration degree is: ,in, To improve the integration of user groups, This represents the sum of inter-class distances for user groups under special weather conditions. This represents the sum of inter-class distances for user groups under normal conditions; the sum of inter-class distances is calculated by summing the pairwise Euclidean distances between all group center points. The value range of F is [0,1]. The closer the F value is to 1, the more similar the user behaviors are; the closer the F value is to 0, the more obvious the differences in user behaviors are. Based on the user group integration degree, the user charging demand conversion degree and the user value-added demand conversion degree are calculated, including: the user charging demand conversion degree calculation steps: calculate the average charging demand feature vector of all users at the current charging station under special weather conditions; calculate the average charging demand feature vector of all users at the current charging station under normal conditions; the user charging demand conversion degree vector is equal to the average charging demand feature vector under special weather conditions minus the average charging demand feature vector under normal conditions; in the user charging demand conversion degree vector, a positive value of any component indicates that the charging demand under special weather conditions is stronger than that under normal conditions in the corresponding dimension, and a negative value indicates that it is weaker; the directional conversion degree component of user charging demand is obtained by multiplying the user group integration degree and the user charging demand conversion degree vector in the preset charging service key dimensions. The steps for calculating the conversion degree of user value-added needs are as follows: Calculate the average value-added needs feature vector of all users at the current charging station under special weather conditions; calculate the average value-added needs feature vector of all users at the current charging station under normal conditions; the user value-added needs conversion degree vector equals the average value-added needs feature vector under special weather conditions minus the average value-added needs feature vector under normal conditions; in the user value-added needs conversion degree vector, a positive value for any component indicates that the value-added needs under special weather conditions are stronger than under normal conditions in the corresponding dimension, while a negative value indicates a weakening; the directional conversion degree component of user value-added needs is obtained by multiplying the user group integration degree and the components of the user value-added needs conversion degree vector on the preset key dimensions of value-added services. ; S4: For each standard operational dimension combination in the charging station operation dimension combination library, perform conversion matching degree calculation and obtain the comprehensive matching degree, including: Charging demand conversion matching degree calculation: obtain the value of the optimal dimension of the current charging station's normal charging service in the charging demand dimension feature vector. ; Obtain the values ​​of the standard charging service dimension in the charging demand dimension feature vector of the standard operation dimension combination. Charging demand conversion matching degree : ,in, For cosine similarity, The conversion degree influence coefficient ranges from [0,1]. Value-added demand conversion matching degree calculation: Obtain the value of the optimal dimension of the current charging station's normal value-added services in the feature vector of the value-added demand dimension. ; Obtain the values ​​of the standard value-added service dimensions in the feature vector of the value-added demand dimensions, which represent the combination of standard operational dimensions. Value-added demand conversion matching degree : ,in The conversion degree influence coefficient has a value range of [0,1]. The overall matching degree is the average of the matching degree of charging demand conversion and the matching degree of value-added demand conversion. The overall operation and management method of charging stations is determined based on the overall matching degree.

2. The integrated operation and management method for charging stations based on multi-dimensional user segmentation as described in claim 1, characterized in that, The method for determining the comprehensive operation and management of charging stations based on the comprehensive matching degree includes: All standard operational dimension combinations are sorted by comprehensive matching degree, and the combination with the highest comprehensive matching degree is selected as the recommended operational dimension combination under special weather conditions. Charging station resource allocation and service strategies are adjusted based on the recommended operational dimension combination. During the duration of special weather conditions, user segmentation integration degree, user charging demand conversion degree, and user value-added demand conversion degree are periodically recalculated and the recommended operational dimension combination is updated. Charging station operation strategies are adjusted in real time based on the dynamically updated recommended operational dimension combination. After the special weather conditions end, the normal operational dimension combination of the charging station is restored. Matching degree data and operational effect data under this special weather condition are recorded, and the charging station operational dimension combination library and conversion degree impact coefficient are updated.

3. A charging station integrated operation and management system based on multi-dimensional user segmentation, used to implement the charging station integrated operation and management method based on multi-dimensional user segmentation as described in any one of claims 1-2, characterized in that, include: The user charging and value-added data collection module is used to acquire user charging demand data and user value-added demand data, and to construct a complete set of mutual information between the charging demand dimension and the value-added demand dimension. The multi-dimensional user segmentation and combination library construction module is used to determine the first mutual information set and the second mutual information set based on the complete mutual information set, perform multi-dimensional user segmentation, and construct the charging station operation dimension combination library. The Special Weather Demand Conversion Analysis Module is used to obtain special weather conditions and determine the user group integration degree, and calculate the user charging demand conversion degree and the user value-added demand conversion degree. The comprehensive matching degree decision and operation module is used to calculate the conversion matching degree and comprehensive matching degree of each standard operation dimension combination, and determine the comprehensive operation and management method of the charging station based on the comprehensive matching degree.

Images