A method for screening key elements of user charging and changing behavior based on principal component analysis

Abstract

This invention discloses a method for screening key elements of user charging and swapping behavior based on principal component analysis. The method includes: collecting multi-source user charging and swapping behavior data; preprocessing the multi-source user charging and swapping behavior data, including outlier removal and data standardization; calculating key influencing factors based on the preprocessed data, including covariance matrices, eigenvalues, and eigenvectors; determining key principal components based on the principal component key influencing factors, and screening out the principal component key influencing factors; and applying the principal component key influencing factors to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization. This application can achieve the extraction, interpretation, and structured output of the core driving factors of user charging and swapping behavior, significantly improving the intelligent operation level and resource allocation efficiency of charging and swapping networks.

Description

Technical Field

[0001] This invention relates to the field of user charging and swapping technology, and in particular to a method for screening key elements of user charging and swapping behavior based on principal component analysis. Background Technology

[0002] With the continuous increase in the penetration rate of electric vehicles (EVs), their dual attributes of transportation and charging behavior have significantly impacted the power system, driving a deep coupling between user choices and grid operation. This coupling has played a positive role in reducing transportation's dependence on fossil fuels and promoting energy structure optimization, but it has also brought new challenges to the safe and stable operation of the power grid. As the main energy replenishment channel for EVs, the power grid can provide users with efficient energy replenishment services at different times. However, with the rapid expansion of EV connectivity, the load fluctuations and operational pressures on the power grid are becoming increasingly prominent. How to scientifically guide the charging and swapping behavior of EVs, and improve the economy and reliability of power grid operation while ensuring users' travel and energy replenishment needs, has become a core issue that urgently needs to be addressed for the large-scale application of EVs.

[0003] To address this challenge, vehicle-to-grid (V2G) technology has emerged, offering a novel solution to alleviate the load pressure caused by the concentrated access of electric vehicles. As the technology matures, the charging and battery swapping service system has gradually evolved into a multi-dimensional data fusion network integrating "vehicle-charging station-road-grid-weather." User charging and battery swapping behavior data is a key foundation of this system, characterized by high dimensionality, large scale, and multi-source heterogeneity. It encompasses various information such as vehicle battery status (e.g., capacity, health, SOC preference), charging spatiotemporal characteristics (e.g., start time, location, frequency), travel demand (e.g., mileage and time), electricity prices and market mechanisms (e.g., time-of-use pricing, incentive pricing), and weather conditions (e.g., temperature, weather conditions). Systematic analysis and key element extraction of this data not only support the power grid in peak shaving and valley filling and load optimization scheduling but also provide operators with crucial information for targeted marketing and improving user service and satisfaction.

[0004] With the continuous growth in the number of electric vehicles, users' charging and swapping behaviors exhibit high complexity across time, space, and energy dimensions, becoming a significant factor influencing grid load changes. To accurately characterize user behavior and identify its potential patterns, existing research primarily employs methods based on descriptive statistics, empirical rules, and traditional machine learning models to analyze and model user behavior. However, these methods still have significant limitations in feature extraction capabilities, model objectivity, and the processing of high-dimensional data.

[0005] First, descriptive statistics and empirical rules are currently the most widely used analytical methods. These methods typically perform statistical calculations such as mean, variance, and frequency distribution on historical charging and swapping data, and combine this with expert experience to summarize behavioral patterns. For example, Guo Yu, in his research "Operational Optimization Research of Electric Taxi Charging Stations," uses electric taxis as the subject, and systematically analyzes the impact of personal attributes, operational characteristics, and charging station service levels on drivers' charging and swapping choices based on behavioral surveys (RP / SP) and disaggregate models (BL and MNL). He Xing, in his research "A Charging and Swapping User Behavior Pattern Analysis System and Method," utilizes multi-source data analysis, behavioral pattern recognition, and abnormal behavior probability calculation methods to construct a dynamic scheduling and resource allocation model, achieving accurate prediction of user charging and swapping behavior. While such research can reveal behavioral influence mechanisms well, the models generally rely on expert experience and struggle to capture complex nonlinear relationships and deep correlations between multidimensional features.

[0006] Secondly, some studies have begun to utilize traditional machine learning models to identify behavioral patterns or predict charging and battery swapping demand. For example, Liu Jiaming, in "A Method and System for Optimizing Driving and Charging Behavior of New Energy Vehicles," identified key factors influencing user charging behavior based on multidimensional data models and machine learning algorithms; Zhao Xingyu, in "A Deep Reinforcement Learning Optimization Method for Charging Behavior of Clustered Electric Vehicles," used deep reinforcement learning combined with real-time monitoring data and time-of-use electricity price signals to analyze user charging behavior characteristics. While such research has made progress in intelligence and prediction accuracy, it is still limited by insufficient high-dimensional feature extraction capabilities, multicollinearity interference of variables, and weak model interpretability.

[0007] Besides the two mainstream methods mentioned above, existing technologies, such as correlation coefficient method and recursive feature elimination, have been introduced for feature selection or dimensionality reduction. However, these methods still have significant limitations. The correlation coefficient method can only characterize the linear relationship between any two variables and cannot reflect the comprehensive effect under the interaction of multiple variables. Although recursive feature elimination has a certain variable screening capability, its calculation process is complex, and the interpretability of the results is weak, making it difficult to support in-depth analysis of user behavior mechanisms.

[0008] In summary, existing technologies still face a number of problems in analyzing user charging and swapping behavior: variable selection relies on expert experience and lacks a unified data-driven standard; it is difficult to make full use of information when faced with high-dimensional and multicollinear data structures; the analysis often remains at the level of superficial pattern description and fails to uncover the deep driving factors that dominate behavior, thus making it difficult to provide reliable support for user behavior prediction and charging and swapping strategy optimization. Summary of the Invention

[0009] The main objective of this invention is to provide a method for screening key elements of user charging and swapping behavior based on principal component analysis. Through objective mathematical calculations, a few key comprehensive elements that can represent the vast majority of behavioral information can be automatically and efficiently extracted from massive and complex user behavior data.

[0010] Another objective of this invention is to propose a device for screening key elements of user charging and swapping behavior based on principal component analysis.

[0011] The third objective of this invention is to provide an electronic device.

[0012] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0013] To achieve the above objectives, a first aspect of the present invention proposes a method for screening key elements of user charging and swapping behavior based on principal component analysis, comprising: Collect charging and battery swapping behavior data from multiple user sources; Preprocessing of charging and swapping behavior data from multiple users includes outlier removal and data standardization; Calculate key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors; Based on the key influencing factors of the principal components, the key principal components are determined, and the key influencing factors of the principal components are screened out. The key influencing factors of the principal components are applied to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

[0014] Optionally, the collection of multi-source user charging and swapping behavior data includes: Obtain the original dataset of charging and swapping behavior of the target user group. The original dataset of charging and swapping behavior contains multiple user samples and their corresponding multidimensional feature variables. The feature variables include starting SOC, ending SOC, charging duration, average daily travel mileage, charging ratio during off-peak / peak / flat periods, charging power preference, price sensitivity, and correlation coefficients with temperature and rainfall.

[0015] Optionally, preprocessing of multi-source user charging and swapping behavior data includes: Remove obviously illogical abnormal data and fill or delete missing data; All feature variables after cleaning are standardized to eliminate the influence of different variable dimensions and orders of magnitude. The Z-score standardization method is used to transform each variable into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardization formula is:

[0016] In the formula: It is the first The first sample Standardized values ​​of each variable. It is its original value. It is the first The mean of each variable, It is the first The standard deviation of each variable.

[0017] Optionally, key influencing factors can be calculated based on the preprocessed data, including: Calculate the covariance matrix among the feature variables to characterize the degree of linear correlation between each pair of variables. If a dataset consists of... There are samples, each sample has One variable, , The formula for calculating the covariance matrix is:

[0018] In the formula, : indicates the first One variable; : indicates the first One variable; : is the first The first sample One variable value; : is the first The mean of the variables is calculated using the following formula: ; : is the first The first sample The values ​​of the variables; : is the first The mean of each variable; : indicates the first The first variable and the second The covariance of each variable, when At that time, it is the first The variance of each variable; Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues. (satisfy ) and their corresponding eigenvectors Eigenvalues ​​represent the variance of the corresponding principal component, reflecting the amount of original data information it carries; eigenvectors determine the orientation of the principal component in the variable space and its linear combination weights with respect to each variable.

[0019] Optionally, key principal components are determined based on the key influencing factors of the principal components, and key influencing factors of the principal components are screened, including: Based on the key influencing factors of the principal components, the number of key principal components to be retained is determined using one or a combination of the Kaiser criterion and the cumulative variance contribution rate criterion. , will the original condensing dimensional variable information into A series of unrelated principal components .

[0020] Optionally, the process of identifying key influencing factors of principal components includes: For each retained key principal component, its loading vector is analyzed. The loading vector is composed of the product of each eigenvector and the square root of the corresponding eigenvalue. Each loading value represents the degree of correlation between the original variable and the principal component. The calculation formula is as follows.

[0021]

[0022] In the formula: : The loading vector of the m-th principal component; : The eigenvalue of the m-th principal component; : The eigenvector corresponding to the m-th principal component; By filtering by loading amplitude, the original variables of each principal component are sorted by the absolute value of the loading, and the variables with the preset ranking are selected as the key elements of the principal component. At the same time, the behavioral meaning of high loading variables is interpreted by combining loading sign and business domain knowledge, so as to clarify the potential behavioral dimensions represented by each principal component and their corresponding behavioral mechanisms.

[0023] To achieve the above objectives, a second aspect of the present invention provides a device for screening key elements of user charging and swapping behavior based on principal component analysis, comprising: The first module is used to collect charging and battery swapping behavior data from multiple users. The second module is used to preprocess the charging and swapping behavior data of multi-source users, including outlier removal and data standardization. The third module is used to calculate key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors. The fourth module is used to determine the key principal components based on the key influencing factors of the principal components, and to screen out the key influencing factors of the principal components. The fifth module is used to apply the key influencing factors of the principal components to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

[0024] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code stored in the memory to implement the method described in the first aspect.

[0025] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0026] The embodiments of this invention have the following beneficial effects: they not only overcome the problems of traditional analysis, such as variable selection relying on experience, high data dimensionality, and collinearity, but also provide a unified and quantifiable technical path for identifying the dominant factors influencing users' charging and swapping behavior. This method can provide scientific, reliable, and interpretable decision support for the precise planning, intelligent operation, demand forecasting, user segmentation, behavioral modeling, and multi-scenario collaborative scheduling of charging and swapping facilities, thereby supporting the high-quality development of vehicle-to-grid interaction, intelligent charging, and urban energy management. Attached Figure Description

[0027] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A simplified text flowchart illustrating a method for screening key elements of user charging and swapping behavior based on principal component analysis, provided in this embodiment of the invention. Figure 2 This is a flowchart illustrating the overall details of a method for screening key elements of user charging and swapping behavior based on principal component analysis, provided in an embodiment of the present invention. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] To overcome the problems of existing technologies, such as reliance on manual experience in variable selection, significant interference from multicollinearity among features in analysis results, and difficulty in effectively revealing deep-seated patterns in user behavior, this invention proposes a method for screening key elements of user charging and swapping behavior based on principal component analysis (PCA). The core idea of ​​this invention is to use PCA, a typical data-driven mathematical tool, to reduce the dimensionality of a large number of highly correlated original behavioral variables into several independent comprehensive principal components. Through quantitative analysis of the loading structure of each principal component, the key elements that dominate user charging and swapping behavior are automatically identified and extracted. This method achieves objective, efficient, and reproducible screening of core influencing factors from massive amounts of complex behavioral data, providing more scientific and robust technical support for charging and swapping facility planning, intelligent operation optimization, and user-side strategy formulation. The overall process of the PCA-based method for screening key elements of user charging and swapping behavior described in this invention is as follows: Figure 1 and Figure 2 As shown, it includes several interrelated steps.

[0031] Reference Figure 1 and Figure 2 The method includes the following steps: S1 collects charging and battery swapping behavior data from multiple user sources.

[0032] In this embodiment of the invention, the original dataset of charging and swapping behavior of the target user group is first obtained. This dataset contains multiple user samples and their corresponding multidimensional feature variables.

[0033] It should be noted that these multidimensional characteristic variables include, but are not limited to, initial SOC, final SOC, charging duration, average daily mileage, charging ratio during off-peak / peak / flat periods, charging power preference, price sensitivity, and correlation coefficients with external factors such as temperature and rainfall.

[0034] S2 preprocesses charging and swapping behavior data from multiple users, including outlier removal and data standardization.

[0035] After obtaining the raw data through step S1, this embodiment of the invention further preprocesses the raw data, including data cleaning, missing value imputation, outlier sample identification and processing, unit unification and standardization, etc., to improve the stability and reliability of subsequent analysis. The specific steps are as follows: (1) Data cleaning: Remove obviously illogical abnormal data (such as charging time exceeding 24 hours, SOC value greater than 100%), and fill or delete missing data.

[0036] (2) Data Standardization: All feature variables after cleaning are standardized to eliminate the influence of different variable dimensions and orders of magnitude. The Z-score standardization method is used to transform each variable into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardization formula is:

[0037] In the formula: It is the first The first sample Standardized values ​​of each variable. It is its original value. It is the first The mean of each variable, It is the first The standard deviation of each variable.

[0038] S3 calculates key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors.

[0039] In this embodiment of the invention, based on the standardized data matrix, the covariance matrix between each feature variable is first calculated to characterize the degree of linear correlation between each pair of variables. This application specifies that a dataset consists of... There are samples, each sample has One variable, , The specific calculation formula is shown below:

[0040] In the formula, : indicates the first One variable; : indicates the first One variable; : is the first The first sample One variable value; : is the first The mean of each variable is calculated using the following formula:

[0041] : is the first The first sample The values ​​of the variables; : is the first The mean of each variable; : indicates the first The first variable and the second The covariance of each variable, when At that time, it is the first The variance of each variable.

[0042] Subsequently, eigenvalues ​​were obtained by eigenvalue decomposition of the covariance matrix. (satisfy ) and their corresponding eigenvectors Eigenvalues ​​represent the variance of the corresponding principal component, reflecting the amount of original data information it carries; eigenvectors determine the orientation of the principal component in the variable space and its linear combination weights with respect to each variable.

[0043] S4. Determine the key principal components based on the key influencing factors of the principal components, and screen out the key influencing factors of the principal components.

[0044] In one embodiment of the present invention, the number of key principal components that need to be retained is determined based on the feature decomposition results. (in , (This represents the total number of original feature variables).

[0045] Preferably, embodiments of the present invention may employ one or a combination of the following criteria for judgment: Kaiser's criterion is a commonly used rule for retaining components or factors in principal component analysis and factor analysis. Its core idea is that only when the eigenvalue is greater than 1 does the principal component (or factor) have a stronger explanatory power than a single original variable, and therefore should be retained. This criterion is based on the characteristics of the covariance matrix or correlation matrix, arguing that components with eigenvalues ​​less than 1 contribute insufficiently to the population variance and are not worth retaining, thus providing a simple and intuitive standard for determining the number of principal components to retain.

[0046] Cumulative variance contribution rate criterion: Usually based on the cumulative variance contribution rate (i.e., the previous...) The number of principal components to be retained is determined by the proportion of the sum of the eigenvalues ​​of each principal component to the total sum of all eigenvalues. When this proportion reaches a preset threshold (such as 80% or 85%), the remaining principal components are selected. Principal components. This criterion can effectively achieve dimensionality reduction of data, transforming the original... Information from dimensional variables is concentrated in A series of unrelated principal components ( )superior.

[0047] It is understandable that, based on the Kaiser criterion and the cumulative variance contribution rate criterion, the embodiments of the present invention can efficiently achieve data dimensionality reduction, transforming the original... condensing dimensional variable information into A series of unrelated principal components This approach retains the most relevant information while significantly reducing redundant variables and improving the simplicity and accuracy of subsequent analyses.

[0048] Next, this embodiment of the invention analyzes the loading vectors of each key principal component retained in the previous step. The loading vector is composed of the product of each eigenvector and the square root of its corresponding eigenvalue, where each loading value represents the degree of correlation between the original variable and the principal component, calculated using the following formula.

[0049]

[0050] In the formula: The loading vector of the m-th principal component; : The eigenvalue of the m-th principal component; : The eigenvector corresponding to the m-th principal component; It should be noted that by quantitatively analyzing the loading matrix, the original variables with large absolute values ​​of loading on each key principal component can be identified. These variables can be regarded as the main contributing factors of the corresponding principal components, which can be used to reveal the core characteristics of users' charging and swapping behavior in the corresponding potential dimensions, thereby achieving the objective extraction and accurate determination of key influencing factors.

[0051] In some embodiments of the present invention, the load amplitude can be used for screening, that is, the original variables of each principal component are sorted according to the absolute value of the load, and the top N (e.g., top 3 or top 5) variables are selected as the key elements of the principal component; at the same time, the behavioral meaning of the above high load variables is interpreted by combining the load sign (positive / negative) and business domain knowledge, so as to clarify the potential behavioral dimensions represented by each principal component and their corresponding behavioral mechanisms.

[0052] S5, apply the key influencing factors of the principal components to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

[0053] It is understandable that the aforementioned key elements can systematically and comprehensively characterize the core driving factors influencing users' charging and swapping behavior, providing a reliable basis for further analysis and business applications. Based on this, results analysis and business applications can be conducted on the key principal components and their related elements to achieve a structured expression and interpretable output of user behavior characteristics, thereby providing support for subsequent operational strategy formulation and scheduling optimization. The specific implementation path is as follows: (1) Visualization and structural feature display: By drawing principal component bipolar plots, the distribution pattern of user samples in principal component space and the directional relationship of key elements are displayed; and by constructing principal component load heatmaps, the importance level and contribution structure of key elements on different principal components are clearly presented, so as to realize the intuitive presentation of behavioral features.

[0054] (2) Refined user group segmentation and identification: Based on the scores of each user on key principal components, cluster analysis or hierarchical identification of user samples is performed to form typical user profiles such as "high travel intensity users", "price-sensitive users" and "battery anxiety users" to support the formulation of differentiated operation strategies and the construction of a precise service system.

[0055] (3) Business strategy design and operation decision optimization: Key principal components and their related behavioral elements are used as input variables and applied to business scenarios such as optimization of charging and swapping facility layout, formulation of time-of-use pricing and incentive strategies, design of demand-side response strategies and personalized service recommendations, so as to provide scientific and reliable decision-making basis for power grid dispatch, operator management and user services.

[0056] Through the above results analysis and application steps, this method can extract, explain and structure the core driving factors of users' charging and swapping behavior, significantly improving the intelligent operation level and resource allocation efficiency of the charging and swapping network.

[0057] Furthermore, this application provides a specific example to illustrate the method proposed in this application, as follows: This embodiment uses the charging and battery swapping behavior data of 5,000 taxis in a certain city from June 2023 to January 2024 as an example, and applies the method described in this invention to screen the key factors affecting their charging and battery swapping behavior.

[0058] S101: Data preparation and preprocessing.

[0059] Raw data was obtained from the charging operation platform and vehicle monitoring platform. The raw data contained 5,000 user samples, each with 42 initial feature variables, including: starting charging SOC (SOC0), ending charging SOC, charging duration, charging capacity, average daily travel mileage, charging location concentration, charging ratio during off-peak / peak / flat periods, charging power preference, price sensitivity, and correlation coefficients between user charging / travel volume and summer temperature, winter temperature, and rainfall.

[0060] First, perform data cleaning: remove abnormal records such as SOC not within the range of 0-100, average daily mileage not within the range of 0-600 kilometers, and charging location not within the land range.

[0061] Then, the variables are standardized: the Z-score standardization method is used to standardize the above 42 feature variables so that their mean is 0 and their standard deviation is 1, thus obtaining the standardized dataset Z.

[0062] S102: Principal component analysis modeling and key principal component identification.

[0063] Based on the standardized dataset Z, its 42×42 covariance matrix was calculated, and eigenvalues ​​and their corresponding eigenvectors were obtained by eigenvalue decomposition. The eigenvalues ​​and cumulative variance contribution rates are shown in Table 1.

[0064] Table 1. Principal Component Eigenvalues ​​and Cumulative Variance Contribution Rate

[0065] Based on the Kaiser criterion (i.e., principal components with eigenvalues ​​greater than or equal to 1 are retained), the top 10 principal components (PC1 to PC10) were ultimately determined as key principal components. Furthermore, the cumulative variance contribution rate of these 10 principal components reached 71.9%, effectively carrying most of the core information from the original 42 variables, demonstrating high representativeness and explanatory power.

[0066] S103: Screening key elements based on load matrix.

[0067] The aforementioned principal component analysis results systematically analyze the business meaning of the first ten key principal components (PC1–PC10), calculate their load matrices, and screen key elements based on the absolute value of the loads. At the same time, the business meaning of each principal component is explained, thereby achieving a structured identification of the driving mechanism of user charging and swapping behavior.

[0068] 1) First principal component (PC1) – Basic travel demand intensity factor.

[0069] In PC1, "average daily mileage" showed the highest load (approximately 0.30), followed by "off-peak charging ratio" (approximately 0.17) and "charging power preference" (approximately 0.12).

[0070] This principal component mainly reflects the decisive influence of users' travel intensity on their charging behavior. In particular, in high-intensity vehicles such as taxis, the greater the travel demand, the more obvious the rigidity and regularity of their charging behavior.

[0071] (2) Second principal component (PC2) – Battery anxiety factor.

[0072] In PC2, the "Initial Charging SOC" load is the highest (approximately 0.13), indicating that the user actively replenishes the battery before the battery level has significantly decreased, reflecting a preventative charging behavior.

[0073] This principal component reflects the user's psychological sensitivity to low battery levels and can characterize the intensity of battery anxiety.

[0074] (3) Third principal component (PC3) – Temperature sensitivity and charging regularity factor.

[0075] The main contributing variables for PC3 include “Initial Charge SOC”, “End Charge SOC”, and “Summer Temperature – Charge Amount Correlation Coefficient”.

[0076] This combination structure demonstrates that users exhibit relatively stable energy replenishment habits under temperature changes, especially high-temperature conditions, reflecting their sensitivity to ambient temperature and regular energy replenishment patterns.

[0077] (4) Analysis of the business meaning of the fourth to tenth principal components (PC4–PC10).

[0078] PC4: Dominated by "Charging Power Preference" and "Simple Charging Ratio", it describes users' preferences for charging speed and charging time.

[0079] PC5: Determined by both "average daily mileage" and "off-peak charging ratio", it portrays the user's strategy of balancing driving needs and charging costs.

[0080] PC6–PC7: mainly composed of variables such as “rainfall-charging correlation” and “peak charging ratio”, reflecting the differences in energy replenishment among different users under severe weather conditions, and divided into “environmentally sensitive type” and “rigid demand type”.

[0081] PC8: Key variables include "peak charging ratio" and "winter temperature-charging volume correlation", reflecting users' ability to avoid peak electricity prices and changes in energy replenishment demand in low-temperature environments.

[0082] PC9–PC10: The core elements are closely related to temperature sensitivity and price sensitivity, revealing the differences in the intensity of users' responses to the external environment and market mechanisms.

[0083] By analyzing the principal components mentioned above, representative key latent factors were effectively extracted from charging and battery swapping behavior data, enabling an interpretable characterization of the user behavior-driven mechanism. The results can directly support the planning of charging and battery swapping facilities, the design of time-of-use pricing and incentive policies, and the formulation of demand-side response strategies, providing reliable data and theoretical basis.

[0084] S104: Verification conclusions and application analysis.

[0085] To verify the effectiveness of the key element screening method proposed in this invention, this embodiment conducted a "key element control effect verification." First, a statistical analysis was performed on the score distribution of users on each principal component. The results showed that the user group exhibited significant clustering characteristics in the principal component space. Among them, PC1 (basic travel intensity factor) users accounted for as high as 33.6%, far exceeding the uniform distribution ratio of 10%, indicating that this type of principal component can effectively distinguish differences in user behavior.

[0086] (1) Verification of PC1 key element regulation.

[0087] This embodiment focuses on the "basic travel demand intensity factor" represented by PC1, using the key element "average daily travel mileage" as the control target. It selects the top 20% of users with high travel intensity based on PC1 scores and implements a differentiated operational strategy, pushing off-peak charging discount packages that match long-distance travel scenarios to them. Verification results show that the off-peak charging volume of this type of user increased by approximately 15% compared to the baseline period, indicating that "average daily travel mileage" and "off-peak charging ratio" are indeed core factors influencing charging and battery swapping behavior. This demonstrates that operational control based on key elements can significantly guide users to migrate to lower-price periods, achieving peak-shaving and valley-filling load management benefits.

[0088] (2) Verification of PC2 key element regulation.

[0089] This embodiment focuses on the key element "initial charging SOC" of the "battery anxiety factor" represented by PC2, identifying "battery anxiety-prone" users whose initial charging SOC consistently exceeds 80%, and then providing them with targeted behavioral guidance information that "shallow charging and discharging are beneficial for improving battery life." Verification results show that the average initial charging SOC of these users decreased by approximately 5 percentage points compared to the baseline period, indicating that targeted regulation of the key element of "battery anxiety" can effectively reduce unnecessary preventative charging behavior. This also verifies that PC2 has good interpretability and controllability in characterizing user psychological tendencies and behavioral preferences.

[0090] (3) Verification conclusion.

[0091] Through experiments controlling key elements of PC1 and PC2, this embodiment verifies the effectiveness of the key element screening method based on principal component analysis of the present invention. This method objectively extracts high-impact key elements such as "average daily travel mileage," "initial charging SOC," "off-peak charging ratio," and "charging power preference" from 42 original variables, revealing the underlying driving factors such as "travel demand intensity," "power anxiety," and "price sensitivity." Actual operational control verifies the significant impact of these key elements on user behavior. The results show that this method not only has strong verifiability and high objectivity but also broad applicability, providing reliable support for various business scenarios such as time-of-use pricing optimization, refined user management, load control, and personalized service recommendations, demonstrating promising engineering application prospects.

[0092] The verification results of this embodiment fully demonstrate the effectiveness of the method of the present invention. Through a feature extraction mechanism based entirely on mathematical operations, the present invention objectively identifies key elements such as "average daily travel mileage," "initial charging SOC," "off-peak charging ratio," and "charging power preference" from 42 original behavioral variables, and further reveals the corresponding deeper driving factors such as "travel demand intensity," "power anxiety tendency," and "price sensitivity." With the help of the above-mentioned structured key elements, the present invention can provide charging operators with reliable data support and accurate decision-making basis for time-of-use pricing strategies, peak shaving and valley filling management, and personalized user service design, thereby achieving significant technical effects.

[0093] Compared with existing technologies, this invention has significant advantages in both methodological innovation and application value, specifically in the following five aspects: 1. High objectivity, enabling comprehensive data-driven analysis. This invention uses principal component analysis as its core, relying entirely on mathematical calculations to assess variable importance, eliminating the need for human experience or rule-based judgment. Compared to traditional screening methods that depend on expert experience, this invention effectively avoids subjective bias, making the selection of key elements more objective, reproducible, and scientifically rigorous.

[0094] 2. Completely resolves multicollinearity interference. This invention transforms the original high-dimensional variables into a set of pairwise orthogonal principal components through eigenvalue decomposition, fundamentally eliminating the impact of high correlation between variables on the analysis results, significantly improving the independence and interpretability of key elements, and providing a more stable and reliable data foundation for subsequent modeling.

[0095] 3. Possesses the ability to reveal deep behavioral mechanisms. Compared to traditional methods that can only analyze univariate features, this invention, through load structure analysis, can extract potential behavioral factors such as "travel demand intensity," "battery anxiety," and "price sensitivity" from high-dimensional data, achieving an improvement from surface statistical features to the essence of behavioral mechanisms, and significantly enhancing the explanatory power of user behavior analysis.

[0096] 4. Efficient dimensionality reduction while maximizing information retention. This invention retains the main information while significantly reducing dimensionality, thereby reducing redundant variables and noise, and lowering computational complexity. This provides a simpler and more reliable data input for subsequent model optimization, behavior prediction, and strategy design.

[0097] 5. Possesses outstanding engineering application value and commercial potential. The key elements and user principal component scores selected by this invention can be directly applied to various practical business scenarios such as charging and swapping facility planning, site power configuration, electricity pricing and incentive strategy formulation, demand-side response management, and user profile construction. It can significantly improve operational efficiency, reduce system costs, and enhance refined service capabilities, demonstrating good scalability and application prospects.

[0098] In summary, this invention not only theoretically overcomes the shortcomings of existing technologies in objective screening, multidimensional fusion, and mechanism identification, but also demonstrates strong operability and innovative value in engineering practice, and has important technical demonstration significance.

[0099] This invention also provides a device for screening key elements of user charging and swapping behavior based on principal component analysis. The device includes: The first module is used to collect charging and battery swapping behavior data from multiple users. The second module is used to preprocess the charging and swapping behavior data of multi-source users, including outlier removal and data standardization. The third module is used to calculate key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors. The fourth module is used to determine the key principal components based on the key influencing factors of the principal components, and to screen out the key influencing factors of the principal components. The fifth module is used to apply the key influencing factors of the principal components to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

[0100] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0101] To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0102] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0104] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0105] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for screening key elements of user charging and swapping behavior based on principal component analysis, characterized in that, include: Collect charging and battery swapping behavior data from multiple user sources; Preprocessing of charging and swapping behavior data from multiple users includes outlier removal and data standardization; Calculate key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors; Based on the key influencing factors of the principal components, the key principal components are determined, and the key influencing factors of the principal components are screened out. The key influencing factors of the principal components are applied to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

2. The method according to claim 1, characterized in that, The collected multi-source user charging and swapping behavior data includes: Obtain the original dataset of charging and swapping behavior of the target user group. The original dataset of charging and swapping behavior contains multiple user samples and their corresponding multidimensional feature variables. The feature variables include starting SOC, ending SOC, charging duration, average daily travel mileage, charging ratio during off-peak / peak / flat periods, charging power preference, price sensitivity, and correlation coefficients with temperature and rainfall.

3. The method according to claim 2, characterized in that, Preprocessing of charging and battery swapping behavior data from multiple users includes: Remove obviously illogical abnormal data and fill or delete missing data; All feature variables after cleaning are standardized to eliminate the influence of different variable dimensions and orders of magnitude. The Z-score standardization method is used to transform each variable into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardization formula is: In the formula: It is the first The first sample Standardized values ​​of each variable. It is its original value. It is the first The mean of each variable, It is the first The standard deviation of each variable.

4. The method according to claim 3, characterized in that, Key influencing factors were calculated based on the preprocessed data, including: Calculate the covariance matrix among the feature variables to characterize the degree of linear correlation between each pair of variables. If a dataset consists of... There are samples, each sample has One variable, , Then the formula for calculating the covariance matrix is: In the formula, : indicates the first One variable; : indicates the first One variable; : is the first The first sample One variable value; : is the first The mean of the variables is calculated using the following formula: ; : is the first The first sample The values ​​of the variables; : is the first The mean of each variable; : indicates the first The first variable and the second The covariance of each variable, when At that time, it is the first The variance of each variable; Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues. (satisfy ) and their corresponding eigenvectors Eigenvalues ​​represent the variance of the corresponding principal component, reflecting the amount of original data information it carries; eigenvectors determine the orientation of the principal component in the variable space and its linear combination weights with respect to each variable.

5. The method according to claim 4, characterized in that, Based on the key influencing factors of the principal components, key principal components are determined, and key influencing factors of the principal components are screened, including: Based on the key influencing factors of the principal components, the number of key principal components to be retained is determined using one or a combination of the Kaiser criterion and the cumulative variance contribution rate criterion. , will the original condensing dimensional variable information into A series of unrelated principal components .

6. The method according to claim 5, characterized in that, The process of identifying key influencing factors of principal components includes: For each retained key principal component, its loading vector is analyzed. The loading vector is composed of the product of each eigenvector and the square root of the corresponding eigenvalue. Each loading value represents the degree of correlation between the original variable and the principal component. The calculation formula is as follows. In the formula: : The loading vector of the m-th principal component; : The eigenvalue of the m-th principal component; : The eigenvector corresponding to the m-th principal component; By filtering by loading amplitude, the original variables of each principal component are sorted by the absolute value of the loading, and the variables with the preset ranking are selected as the key elements of the principal component. At the same time, the behavioral meaning of high loading variables is interpreted by combining loading sign and business domain knowledge, so as to clarify the potential behavioral dimensions represented by each principal component and their corresponding behavioral mechanisms.

7. A device for screening key elements of user charging and swapping behavior based on principal component analysis, characterized in that, include: The first module is used to collect charging and battery swapping behavior data from multiple users. The second module is used to preprocess the charging and swapping behavior data of multi-source users, including outlier removal and data standardization. The third module is used to calculate key influencing factors based on the preprocessed data, including the covariance matrix, eigenvalues, and eigenvectors. The fourth module is used to determine the key principal components based on the key influencing factors of the principal components, and to screen out the key influencing factors of the principal components. The fifth module is used to apply the key influencing factors of the principal components to perform visualization and structural feature display, refined user group segmentation and identification, and business strategy design and operational decision optimization.

8. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images