Electric vehicle charging behavior identification method based on user portrait

Abstract

The invention discloses an electric vehicle charging behavior identification method based on a user portrait, and the method comprises the steps: obtaining user portrait features and charging behavior data from a user database, processing missing values and abnormal values through a data cleaning technology, and obtaining a standardized user feature data set and a charging behavior data set; extracting space-time dynamic characteristics from the charging behavior data set, and decomposing charging frequency and charging duration distribution by adopting a time sequence analysis method to obtain charging behavior modes of the user in different time periods and places; according to the charging pile layout schemes of the commercial district and the residential district, optimizing power grid dispatching by adopting a linear programming algorithm, and determining power distribution proportions of different districts; and extracting personalized service requirements from the occupational background and the charging behavior mode of the user, generating exclusive charging time and place recommendation of the user by adopting a decision tree algorithm, and obtaining a personalized charging service scheme.

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle charging technology, and in particular relates to a method for recognizing electric vehicle charging behavior based on user profiles. Background Technology

[0002] The widespread adoption of electric vehicles has placed new demands on energy management and urban planning. Charging behavior recognition, as a key element in optimizing charging infrastructure and improving user experience, is receiving increasing attention. Analyzing users' charging habits can support charging pile deployment, grid dispatching, and personalized services. However, current research and applications have significant shortcomings in integrating user characteristics with charging behavior. Existing methods often rely on single charging data points, such as time or location, neglecting the dynamic correlation between user background information and behavior. This results in a lack of depth in the analysis and an inability to accurately reflect the actual needs of different user groups. For example, existing systems struggle to accurately distinguish the charging preferences of users in residential and commercial areas, limiting the optimal allocation of charging facilities.

[0003] The core challenge in this field lies in effectively integrating user profiles and charging behavior data to uncover behavioral patterns. First, the diverse characteristics in user profiles, such as age, occupation, income level, and driving habits, have complex interactions with charging behavior, making effective integration difficult through simple statistical methods. For example, professional drivers may prefer fast charging at highway service areas, while ordinary office workers are more likely to use slow charging overnight in residential areas; this difference requires systematic data fusion to capture. Second, the spatiotemporal characteristics of charging behavior are highly dynamic. Existing technologies often fail to accurately correlate users' geographical location preferences with charging frequency and duration when processing multidimensional data, resulting in analysis results that cannot provide precise guidance for real-world scenarios. For example, a user might choose to charge in a commercial area on weekdays but prefer to charge in a residential area on weekends; if this dynamic change cannot be effectively identified, it will directly affect the rationality of charging station deployment.

[0004] Therefore, the key issue of this study is how to combine the diverse features of user profiles with the spatiotemporal dynamics of charging behavior to build a data system that can accurately identify user charging preferences in different scenarios. Summary of the Invention

[0005] This invention proposes a method for recognizing electric vehicle charging behavior based on user profiles to solve the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides a method for recognizing electric vehicle charging behavior based on user profiles, comprising the following steps:

[0007] Acquire user profile features and charging behavior data and preprocess them to obtain standardized user feature datasets and charging behavior datasets;

[0008] Based on the standardized user feature dataset and charging behavior dataset, user profile features are classified to obtain classification results of user groups' occupational background and geographical location preferences;

[0009] Spatiotemporal dynamic characteristics are extracted from the charging behavior dataset, and time series analysis is used to decompose the charging frequency and charging duration distribution to obtain the charging behavior patterns of users in different time periods and locations.

[0010] Based on charging behavior patterns and combined with geographical location preference classification results, we can obtain charging pile layout schemes for commercial and residential areas.

[0011] Based on the charging pile layout plan for commercial and residential areas, a linear programming algorithm is used to optimize the power grid dispatch and determine the power allocation ratio for different areas.

[0012] Personalized service needs are extracted from users' professional backgrounds and charging behavior patterns. A decision tree algorithm is used to generate personalized charging time and location recommendations for each user, resulting in a personalized charging service plan.

[0013] Optionally, the obtained standardized user feature dataset and charging behavior dataset include:

[0014] User profile features and charging behavior data are obtained from the user database. SQL query language is used to extract data including user ID, age, gender, occupation, charging time, charging frequency and charging location to obtain the raw dataset.

[0015] For the original dataset, the mean imputation method is used to handle missing values. If the missing rate of a field exceeds a preset threshold, the field is removed to obtain a preliminary cleaned dataset.

[0016] Outlier detection was performed on the initially cleaned dataset using a standard deviation-based method. If a data point exceeded three times the standard deviation of the mean, it was marked as an outlier and removed, resulting in an outlier-free dataset.

[0017] Based on the dataset without anomalies, the numerical features are normalized using the Z-score normalization method to generate a standardized user feature dataset and a charging behavior dataset.

[0018] Optionally, the obtained user group occupational background and geographic location preference classification results include:

[0019] The K-means clustering algorithm is used to process the standardized user feature dataset to determine the user classification results;

[0020] Based on the user classification results, the occupational background features of each classification group are extracted to obtain the occupational distribution feature set;

[0021] By analyzing the charging behavior preferences of each category group in terms of geographical location through the occupational distribution feature set, the location preference distribution can be obtained.

[0022] Optionally, obtaining the user's charging behavior patterns at different time periods and locations includes:

[0023] Spatiotemporal dynamic features are obtained from the charging behavior dataset, and the charging frequency and charging duration distributions are separated by the time series decomposition method to determine the user's behavior pattern set in different time periods and locations.

[0024] By analyzing the relationship between time period division and location distribution characteristics through behavioral pattern set, the time-location correlation coefficient is calculated by correlation analysis method to obtain spatiotemporal correlation rule set;

[0025] If the charging frequency in a certain time period in the spatiotemporal association rule set is higher than a preset threshold, it is marked as a high-frequency time period, and a set of high-frequency time periods is obtained.

[0026] Based on the high-frequency time period set, the corresponding location distribution characteristics are extracted, and the location distribution characteristics are grouped using cluster analysis to determine the location cluster set;

[0027] By analyzing the characteristics of each cluster in terms of charging time distribution through location clustering sets, a set of time distribution characteristics can be obtained.

[0028] If the concentration of charging time in a certain cluster of time distribution characteristics is higher than a preset threshold, it is marked as a high concentration group, and the high concentration group set is determined.

[0029] Optionally, the acquisition of charging pile layout schemes for commercial and residential areas includes:

[0030] When charging behavior patterns show that users tend to charge in commercial areas on weekdays, a weighted average algorithm combined with geographical location preference classification results is used to calculate the priority distribution of charging piles in commercial areas and obtain a charging pile layout scheme for commercial areas.

[0031] When charging behavior patterns show that users tend to charge in residential areas on weekends, a density clustering algorithm is used to analyze residential charging preferences, generate the density of charging piles in residential areas, and obtain a residential charging pile layout scheme.

[0032] Optionally, the obtained commercial area charging pile layout scheme includes:

[0033] Using data collection methods, weekday tendency data is extracted from charging behavior patterns, and combined with time period distribution, the distribution of users' charging demand in commercial areas is determined.

[0034] If the charging demand distribution shows that the charging frequency is higher than a preset threshold during a certain period, it is marked as a high-demand period, and the set of high-demand periods is obtained.

[0035] Based on the high-demand time period set and combined with geographical location preferences, a weighted average algorithm is used to calculate the priority of each charging pile in the commercial area, and the priority distribution of charging piles is obtained.

[0036] By analyzing the charging pile priority distribution, the utilization rate of charging piles is determined, charging piles with utilization rates below a preset threshold are identified, and a set of charging piles with low utilization rates is obtained.

[0037] If the number of charging piles in the low-usage charging pile cluster exceeds a preset threshold, then a cluster analysis method is used to group the charging pile distribution to obtain a charging pile distribution cluster set.

[0038] Based on the cluster set of charging pile distribution, combined with users' charging needs and location preferences, an optimization plan for the layout of charging piles in commercial areas is generated, and the optimized distribution of charging piles in commercial areas is determined.

[0039] Optionally, the obtained residential area charging pile layout scheme includes:

[0040] Weekend residential charging data is obtained from charging behavior patterns, and charging demand distribution is generated by combining time period and geographical location information.

[0041] If the charging frequency in any time period of the charging demand distribution is higher than a preset threshold, it is marked as a peak period, and a peak period set is obtained.

[0042] Density clustering algorithm is used to analyze peak time period sets and geographical location distribution to generate the density distribution of charging pile configuration in residential areas;

[0043] Based on the configuration density distribution, the expected utilization rate of each charging pile in the residential area is calculated to obtain the utilization rate distribution set;

[0044] If there are charging piles in the usage distribution that have a usage rate lower than a preset threshold, they are marked as inefficient charging piles, and a set of inefficient charging piles is obtained.

[0045] By adjusting the location of inefficient charging pile clusters and their density distribution, an optimized charging pile layout scheme for residential areas can be generated.

[0046] Optionally, determining the power allocation ratio for different regions includes:

[0047] The power demand distribution is obtained from the charging pile layout schemes in commercial and residential areas, and power demand sets for each area are generated by combining time periods and regional characteristics.

[0048] If the peak charging demand in any region of concentrated electricity demand exceeds a preset threshold, it is marked as a high-load region, thus obtaining a set of high-load regions.

[0049] A linear programming algorithm is used to analyze the high-load area set and power grid resource constraints, generate a regional load balancing scheme, and determine the power allocation ratio set for each region.

[0050] By combining the regional power distribution ratio set with the layout of charging piles, the expected power supply of each charging pile is calculated, and the power supply distribution set is obtained.

[0051] If the expected power supply of charging piles in a concentrated power supply distribution is lower than a preset threshold, they are marked as low-supply charging piles, and a set of low-supply charging piles is obtained.

[0052] Based on the low-supply charging pile cluster and regional load balancing scheme, the power allocation ratio is adjusted to generate an optimized dynamic allocation scheme.

[0053] By using dynamic allocation schemes and grid resource constraints, the stability of power allocation ratios in different regions is verified, and a stable power allocation set is obtained.

[0054] Using a stable power distribution set, the power dispatch configuration of charging piles in commercial and residential areas is updated to generate the final power grid dispatch scheme.

[0055] Optionally, the personalized charging service solution includes:

[0056] User behavior data is obtained from users' professional background and charging behavior patterns. Feature extraction methods are used to analyze the occupation type and daily activity patterns to obtain a user behavior feature set.

[0057] Based on the user behavior feature set, the time preference analysis method is used to calculate the probability of users' charging needs in different time periods, and obtain the charging time preference set.

[0058] By combining the charging time preference set with the user behavior feature set, and using the location preference analysis method, the user's permanent location and activity area are matched to obtain the charging location preference set;

[0059] Using a decision tree algorithm, the system takes a set of charging time preferences and a set of charging location preferences as input, generates a user-specific charging time and location recommendation scheme, and obtains a charging service scheme.

[0060] Compared with the prior art, the present invention has the following advantages and technical effects:

[0061] This invention discloses a method for recognizing electric vehicle charging behavior based on user profiles. Addressing the business scenario problem of accurately meeting users' charging needs in different scenarios and optimizing power grid resource allocation, the method uses data cleaning techniques to process missing and outlier values ​​in the user database, constructing a standardized dataset of user characteristics and charging behavior. K-means clustering is used to analyze users' occupational background and geographical location preferences, and time series analysis is combined to extract charging frequency and duration distributions, identifying charging patterns in commercial areas on weekdays and residential areas on weekends. For charging preferences in commercial areas, a weighted average algorithm is used to calculate the priority distribution of charging piles; for residential areas, a density clustering algorithm is used to generate the charging pile configuration density. Finally, a linear programming algorithm is used to optimize power grid scheduling and determine the power allocation ratio, while a decision tree algorithm is used to generate personalized charging time and location recommendations. This invention achieves charging pile layout optimization and personalized services through the fusion of multiple algorithms, significantly improving charging efficiency and power grid resource utilization. Attached Figure Description

[0062] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0063] Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0064] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0066] Example 1

[0067] like Figure 1 As shown, this embodiment provides a method for recognizing electric vehicle charging behavior based on user profiles, including the following steps:

[0068] Acquire user profile features and charging behavior data and preprocess them to obtain standardized user feature datasets and charging behavior datasets;

[0069] Based on the standardized user feature dataset and charging behavior dataset, user profile features are classified to obtain classification results of user groups' occupational background and geographical location preferences;

[0070] Spatiotemporal dynamic characteristics are extracted from the charging behavior dataset, and time series analysis is used to decompose the charging frequency and charging duration distribution to obtain the charging behavior patterns of users in different time periods and locations.

[0071] Based on charging behavior patterns and combined with geographical location preference classification results, we can obtain charging pile layout schemes for commercial and residential areas.

[0072] The acquisition of charging pile deployment plans for commercial and residential areas includes:

[0073] When charging behavior patterns show that users tend to charge in commercial areas on weekdays, a weighted average algorithm combined with geographical location preference classification results is used to calculate the priority distribution of charging piles in commercial areas and obtain a charging pile layout scheme for commercial areas.

[0074] When charging behavior patterns show that users tend to charge in residential areas on weekends, a density clustering algorithm is used to analyze residential charging preferences, generate the density of charging piles in residential areas, and obtain a residential charging pile layout scheme.

[0075] Based on the charging pile layout plan for commercial and residential areas, a linear programming algorithm is used to optimize the power grid dispatch and determine the power allocation ratio for different areas.

[0076] Personalized service needs are extracted from users' professional backgrounds and charging behavior patterns. A decision tree algorithm is used to generate personalized charging time and location recommendations for each user, resulting in a personalized charging service plan.

[0077] Specifically, the following steps are included:

[0078] S101. Obtain user profile features and charging behavior data from the user database, and use data cleaning techniques to process missing and outlier values ​​to obtain a standardized user feature dataset and charging behavior dataset.

[0079] Specifically, user profile features and charging behavior data are obtained from the user database. SQL query language is used to extract data including user ID, age, gender, occupation, charging time, charging frequency, and charging location to obtain the raw dataset. For the raw dataset, mean imputation is used to handle missing values. If the missing value rate of a field exceeds a preset threshold, the record is removed, resulting in a pre-cleaned dataset. Outlier detection is performed on the pre-cleaned dataset using a standard deviation-based method. If a data point exceeds three times the standard deviation of the mean, it is marked as an outlier and removed, resulting in an outlier-free dataset. Based on the outlier-free dataset, Z-score normalization is used to normalize the numerical features, generating a standardized user feature dataset and a charging behavior dataset.

[0080] For example, in processing user profile features and charging behavior data, SQL query language is used to extract required fields from the database. Suppose an electric vehicle charging service platform stores information such as user ID, age, gender, occupation, charging time, charging frequency, and charging location. The raw dataset can be extracted using an SQL statement such as `SELECT user_id, age, gender, occupation, charge_time, charge_frequency, charge_location FROM user_data`. The raw dataset may contain 10,000 records, each corresponding to a user's features and behavioral data.

[0081] In this embodiment, the mean imputation method is used to handle missing values ​​in the original dataset.

[0082] For example, the age field might have 5% missing records. The average of the non-missing ages, such as 35, can be calculated and used to fill in the missing ages. If the missing rate of the charging frequency field exceeds a preset threshold of 20%, the record is removed. Assuming 500 records are deleted due to excessive missing charging frequency records, the initially cleaned dataset contains 9500 records. This method ensures data integrity, reduces the interference of missing values ​​on subsequent analysis, and helps improve the accuracy of model training.

[0083] Specifically, outlier detection uses a standard deviation-based method.

[0084] For example, the average charging frequency is 3 times per week, with a standard deviation of 1. If a user charges 10 times per week, exceeding three times the standard deviation of the mean, it is marked as an anomaly and removed. Assuming 100 records are removed due to anomalies, a dataset of 9400 records without anomalies is obtained. This step effectively removes extreme values, ensuring the reasonableness of the data distribution and providing a reliable foundation for subsequent standardization.

[0085] In this embodiment, Z-score normalization is used to handle numerical features such as age and charging frequency.

[0086] For example, with a mean age of 35 and a standard deviation of 10, a user aged 45 would have a Z-score of (45-35) / 10 = 1. Similarly, charging frequency is standardized, with a mean of 3 times and a standard deviation of 1. A user charging 4 times would have a Z-score of (4-3) / 1 = 1. After standardization, the mean of numerical features is 0 and the standard deviation is 1, generating a standardized user feature dataset and a charging behavior dataset. This method makes features of different dimensions comparable, facilitating the training of subsequent clustering or prediction models.

[0087] Understandably, the above processing significantly improves data quality through layers of cleaning and standardization.

[0088] For example, standardized datasets can be used for user behavior clustering, identifying high-frequency charging user groups, and optimizing charging pile layout. Missing value imputation and outlier removal reduce data noise, while Z-score standardization ensures feature fairness. Together, they support the construction of accurate user profiles and the analysis of charging behavior, providing the platform with data-driven operational strategies.

[0089] S102. Based on the standardized user feature dataset and charging behavior dataset, the K-means clustering algorithm is used to classify the user profile features to obtain the classification results of user group occupational background and geographical location preference.

[0090] Specifically, the K-means clustering algorithm is used to process the standardized user feature dataset to determine the user classification results. Based on the user classification results, the occupational background features of each category are extracted to obtain an occupational distribution feature set. Through the occupational distribution feature set, the charging behavior preferences of each category in terms of geographical location are analyzed to obtain the location preference distribution.

[0091] For example, when processing a standardized user feature dataset using the K-means clustering algorithm, user groups can be divided by determining the number of clusters. K-means clustering is an unsupervised learning method that iteratively calculates the distance between data points and cluster centers, assigning data points to the nearest centers to form user classifications.

[0092] Specifically, assuming the dataset contains features such as user ID, age, gender, and occupation, we can set the K value to 4, which means that users are divided into 4 categories.

[0093] In this embodiment, after initializing four cluster centers, the algorithm assigns each user to the nearest center based on Euclidean distance, recalculates the center position, and repeats this process until the centers are stable.

[0094] For example, a clustering result might divide users into four categories: young white-collar workers, middle-aged technical personnel, students, and retirees. Each category contains users with different ages and occupational characteristics.

[0095] In this embodiment, when extracting the occupational background features of each classification group, an occupational distribution feature set can be generated through statistical analysis.

[0096] For example, young white-collar workers may mainly work in the finance and internet industries, accounting for 40% and 30% respectively; middle-aged technical personnel may mainly be engineers and programmers, accounting for 50% and 20% respectively.

[0097] Specifically, the frequency of the occupation field in each category can be counted using SQL queries to generate a distribution table.

[0098] For example, the occupational distribution characteristics of young white-collar workers show that financial professionals tend to charge their devices more frequently, averaging five times a week, while internet professionals prefer to charge at fixed locations, with 80% concentrated at charging stations in city centers.

[0099] For example, when analyzing the charging behavior preferences of each group in terms of geographical location, the charging location field in the charging behavior dataset can be combined to count the charging frequency of each group in different regions.

[0100] In this embodiment, young white-collar workers may have the highest usage rate of charging stations in downtown business districts, accounting for 60%, while in suburbs it is only 10%; students tend to prefer charging stations near their schools, accounting for 70%.

[0101] Specifically, charging locations can be mapped using a Geographic Information System (GIS) to generate heat maps that visually display location preferences.

[0102] For example, young white-collar workers tend to prefer charging in commercial areas, possibly due to their concentrated workplaces and high commuting needs, while students tend to use charging stations around their schools due to their campus life. This kind of analysis can help optimize the layout of charging stations and improve resource utilization efficiency.

[0103] In this embodiment, the location preference distribution can be further refined through correlation analysis.

[0104] For example, the difference in charging behavior between weekdays and weekends can be analyzed by combining the time dimension. Young white-collar workers might choose city center charging stations 90% of the time on weekdays, while on weekends they tend to disperse to suburban shopping malls, accounting for 30%. In this way, the generated distribution data not only reflects spatial preferences but also reveals behavioral patterns over time, providing precise data for charging station operation.

[0105] Understandably, the above methods, through clustering, occupational distribution extraction, and location preference analysis, progressively form a complete user behavior profile. Each step supports the others, jointly constructing a comprehensive analytical framework from user characteristics to charging behavior.

[0106] For example, clustering results provide a classification basis for occupational distribution, which in turn provides user background support for location preference analysis. This logically rigorous analytical chain ensures in-depth data mining and targeted applications.

[0107] S103. Extract spatiotemporal dynamic characteristics from the charging behavior dataset, and use time series analysis to decompose the charging frequency and charging duration distribution to obtain the user's charging behavior patterns in different time periods and locations.

[0108] Specifically, spatiotemporal dynamic features are obtained from the charging behavior dataset. Time series decomposition is used to separate the charging frequency and charging duration distributions, identifying user behavior patterns across different time periods and locations. These behavior patterns are then analyzed to determine the correlation between time period divisions and location distribution characteristics. Correlation analysis is used to calculate the time-location correlation coefficient, resulting in a spatiotemporal correlation rule set. If the charging frequency in a given time period exceeds a preset threshold, it is marked as a high-frequency time period, thus obtaining a high-frequency time period set. Based on this set, corresponding location distribution features are extracted, and cluster analysis is used to group these features, determining location clusters. The characteristics of each cluster in the charging duration distribution are analyzed, obtaining a duration distribution characteristic set. If the charging duration concentration of a particular cluster in the duration distribution characteristic set exceeds a preset threshold, it is marked as a high-concentration group, thus determining a high-concentration group set. Based on the high-concentration group set, a charging behavior prediction model is generated for each group. Regression analysis is used to predict charging demand trends, resulting in a demand trend set.

[0109] It should be noted that when obtaining spatiotemporal dynamic features from charging behavior datasets, the periodic and trend features of charging frequency and duration can be extracted using time series decomposition methods.

[0110] For example, a dataset of electric vehicle users in a certain city contains daily charging records. The day can be divided into three time periods: morning, noon, and evening, to analyze the charging frequency of users during each period. Assuming the charging frequency during the morning peak (7:00-9:00) is 20 times per day, and the evening peak (17:00-19:00) is 30 times per day, time series decomposition extracts the periodic fluctuations in frequency, revealing that charging demand is more concentrated during the evening peak. This decomposition method clearly separates the regularities in user behavior, facilitating subsequent analysis.

[0111] Specifically, when analyzing the correlation between time periods and location distribution characteristics, correlation analysis can be used to calculate the time-location correlation coefficient.

[0112] For example, in the charging station data for commercial and residential areas, the charging frequency during the evening peak hours showed a strong correlation with the geographical location of charging stations in commercial areas, with a correlation coefficient of 0.85, while it was only 0.3 in residential areas. This indicates that users are more inclined to charge in commercial areas during the evening peak hours, possibly because shopping or social activities are concentrated there after get off work. Such analysis helps to accurately locate high-demand areas.

[0113] In this embodiment, when marking high-frequency time periods, the charging frequency threshold is set to 25 times per day. The evening peak period from 17:00 to 19:00 is marked as a high-frequency time period because the frequency is 30 times. Based on this, the location distribution characteristics of this period are extracted, revealing that 80% of charging behavior is concentrated in charging stations in commercial areas and transportation hubs. Using cluster analysis, these locations are grouped according to geographical density, resulting in three clusters: downtown commercial areas, suburban transportation hubs, and the outskirts of residential areas. The downtown commercial area has the highest density of charging stations, indicating that users' charging demand is more concentrated in this area.

[0114] For example, analyzing the charging duration distribution characteristics of clustered groups reveals that the average charging duration of the cluster in the city center business district is 2.5 hours, with a standard deviation of 0.4 hours, indicating a high concentration of duration distribution. Setting a concentration threshold of a standard deviation less than 0.5 hours, this group is labeled as a high-concentration group. This high concentration reflects relatively regular user charging behavior in business districts, possibly due to fixed work or consumption scenarios. Based on this, when generating a charging behavior prediction model, regression analysis can be used to combine historical data to predict charging demand for the coming week.

[0115] For example, it is predicted that charging demand in downtown business districts will increase by 10% during the evening rush hour due to increased foot traffic from weekend promotions. Such forecasts can provide a basis for the allocation of charging station resources.

[0116] In this embodiment, the extended solution can be further refined by combining the user's professional background.

[0117] For example, frequent charging users in commercial areas are mostly white-collar workers, with charging times concentrated at 2-3 hours, reflecting their fixed off-get off work hours. In contrast, users at transportation hubs are mostly drivers, with charging times more dispersed, with a standard deviation of 0.8 hours. This difference can be used to optimize charging station layout, such as increasing fast charging piles in commercial areas and setting up diverse charging equipment at hubs. Such analysis supports accurate user behavior insights from multiple dimensions of time, location, and duration.

[0118] It should be noted that the combination of spatiotemporal correlation rules and high-concentration group analysis can effectively identify core charging demand scenarios.

[0119] For example, the high frequency and concentrated duration of charging during evening rush hour in commercial areas suggest that operators could prioritize increasing the supply of charging stations in these areas and time periods. This multi-faceted analysis ensures logical consistency from data extraction to predictive models, providing comprehensive support for optimizing charging infrastructure.

[0120] S104. If the charging behavior pattern shows that users tend to charge in commercial areas on weekdays, a weighted average algorithm combined with geographical location preference is used to calculate the priority distribution of charging piles in commercial areas, and a charging pile layout scheme for commercial areas is obtained.

[0121] Specifically, data collection methods are employed to extract weekday preference data from charging behavior patterns. Combined with time-period distribution, this determines the distribution of user charging demand in commercial areas. If the charging demand distribution shows that the charging frequency during a certain time period exceeds a preset threshold, it is marked as a high-demand time period, and a set of high-demand time periods is obtained. Based on this set of high-demand time periods and considering geographical location preferences, a weighted average algorithm is used to calculate the priority of each charging pile within the commercial area, resulting in a charging pile priority distribution. The charging pile utilization rate is analyzed using this priority distribution to identify charging piles with utilization rates below a preset threshold, thus obtaining a set of low-utilization charging piles. If the number of charging piles in the low-utilization charging pile set exceeds a preset threshold, cluster analysis is used to group the charging pile distribution, resulting in a charging pile distribution cluster set. Based on the charging pile distribution cluster set, combined with user charging demand and location selection preferences, an optimized charging pile layout plan for the commercial area is generated, determining the optimized charging pile distribution for the commercial area.

[0122] For example, when extracting weekday preference data from charging behavior patterns, one can analyze charging time and location preferences by collecting users' charging records on weekdays. Assume a charging dataset for a city's commercial district includes user ID, charging start time, charging duration, and charging station location. Data collection methods can be based on smart meter or charging station log systems, filtering charging data from Monday to Friday, and statistically analyzing the daily charging frequency and duration distribution. Results show that charging demand between 8:00 and 10:00 on weekdays accounts for 30% of the total daily demand, indicating that users tend to charge before going to work in the morning.

[0123] Specifically, when determining the distribution of charging demand in commercial areas, a day can be divided into six four-hour periods, and the charging frequency for each period can be statistically analyzed. If the charging frequency from 9:00 to 13:00 accounts for 25% of the daily total, exceeding a preset threshold of 20%, it is marked as a high-demand period. Analysis of one week's data confirmed that 9:00-13:00 and 17:00-21:00 are the high-demand period sets. This reflects peak charging times for users during lunch breaks and after get off work.

[0124] In this embodiment, a weighted average algorithm is used to calculate the priority of charging piles, combining the high-demand time period set and geographical location preference. Assuming there are 10 charging piles in a commercial area, location preference is based on the number of times each charging pile is used in the user's charging records, with a weight of 0.6; the time period demand weight is 0.4. Charging pile A accounts for 40% of the total usage during high-demand time periods, has a high geographical preference score, and its priority score reaches 0.85, ranking first. The priority distribution shows that 3 charging piles have scores below 0.5, and are marked as a low-usage charging pile set.

[0125] For example, when analyzing a set of charging stations with low utilization rates, if 4 out of 10 charging stations have a utilization rate below 30%, cluster analysis is triggered. Using the K-means algorithm, the charging stations are grouped into 3 clusters based on their geographical coordinates and utilization rates. Cluster 1 contains charging stations near the edge of commercial areas, with low utilization rates and infrequent user visits. By combining user needs and location preferences, the optimized layout can relocate these charging stations to high-demand areas, such as near subway stations.

[0126] Specifically, when generating the optimization plan, the system considers both high-demand time periods and clusters, prioritizing increasing charging pile density in peak demand areas from 9:00 AM to 1:00 PM. Assuming the original layout had 10 charging piles, the optimization adds 2 more charging piles in the high-traffic area of ​​cluster 2, expected to increase overall utilization by 15%. Simultaneously, the locations of less-used charging piles are adjusted to shorten walking distances for users and improve convenience.

[0127] It should be noted that the optimized distribution of charging stations in the commercial area needs to be verified periodically. Charging data should be collected for one week after the optimization, and usage rate changes analyzed. If the usage rate of all charging stations under the new layout exceeds 40%, the solution is considered effective. This method, through data-driven layout adjustments, significantly improves the operational efficiency of charging stations and meets users' charging needs on weekdays.

[0128] S105. If the charging behavior pattern shows that users tend to charge in residential areas on weekends, then the density clustering algorithm is used to analyze the charging preferences in residential areas, generate the configuration density of charging piles in residential areas, and obtain the layout scheme of charging piles in residential areas.

[0129] Specifically, charging behavior patterns in residential areas on weekends are obtained, and charging demand distribution is generated by combining time period and geographical location information. If the charging frequency in any time period of the charging demand distribution is higher than a preset threshold, it is marked as a peak period, resulting in a peak period set. Density clustering algorithm is used to analyze the peak period set and geographical location distribution to generate a residential area charging pile configuration density distribution. Based on the configuration density distribution, the expected utilization rate of each charging pile in the residential area is calculated, resulting in a utilization rate distribution set. If there are charging piles in the utilization rate distribution set with utilization rates lower than a preset threshold, they are marked as inefficient charging piles, resulting in an inefficient charging pile set. Based on the inefficient charging pile set and configuration density distribution, the locations of charging piles are adjusted to generate an optimized residential area charging pile layout scheme.

[0130] In this embodiment, obtaining weekend residential area charging data from charging behavior patterns requires extracting charging behavior during specific time periods on weekends from user charging records.

[0131] For example, assuming a residential area has 1,000 households, log data from smart charging stations can be analyzed to determine the number, time, and location of charging on Saturdays and Sundays. The data might show higher charging demand between 6:00 PM and 10:00 PM on weekend evenings, as residents are more likely to return home during this time. Combined with geographic information, such as the distribution of residential buildings, a charging demand distribution map can be generated, reflecting the charging frequency in each area.

[0132] For example, in the distribution of charging demand, if the charging frequency during a certain time period, such as Sunday at 8:00 PM, exceeds a preset threshold (e.g., 50 charging requests per hour), it is marked as a peak period. Analysis revealed that the peak period is from 6:00 PM to 10:00 PM on weekends. This step, through statistical analysis of time periods, ensures that the pattern of concentrated user charging is captured, which helps in subsequent optimization of resource allocation.

[0133] In this embodiment, a density clustering algorithm is used to analyze the peak time period set and geographical location distribution to generate a charging pile configuration density distribution.

[0134] Specifically, density clustering can identify high-density areas based on the geographical coordinates of charging stations and their usage frequency during peak hours.

[0135] For example, charging stations in the central area of ​​a residential area have a peak usage rate of 80%, while those in the peripheral areas only reach 20%. An algorithm generates a density distribution map showing that the central area requires a higher density of charging stations. This method can accurately pinpoint demand hotspots, facilitating the rational planning of charging station layout.

[0136] Understandably, based on the configuration density distribution, the expected utilization rate of each charging pile is calculated to obtain the utilization rate distribution set.

[0137] For example, the expected utilization rate of charging piles in the central area might be 75%, while in the peripheral areas it might only be 15%. If a preset threshold of 30% is set, some charging piles in the peripheral areas are marked as inefficient charging piles, forming a set of inefficient charging piles. This analysis, by quantifying utilization rates, clearly distinguishes between efficient and inefficient equipment, providing data support for optimization.

[0138] For example, by analyzing the distribution of inefficient charging station clusters and their density, the location of charging stations can be adjusted to optimize the layout. Suppose a residential area has 50 charging stations, 10 of which are marked as inefficient and concentrated in the peripheral areas. An optimization plan might relocate 5 of these charging stations to high-density demand points in the central area, while retaining a small number of charging stations in the peripheral areas to cover occasional demand. This adjustment ensures that resource allocation better aligns with user habits and improves overall charging efficiency.

[0139] In this embodiment, the optimized residential area charging pile layout scheme also takes into account user convenience.

[0140] For example, newly added charging stations in the central area can be located near parking lot entrances for residents' quick access. This design combines users' location preferences and actual needs, improving the accessibility and satisfaction of charging services.

[0141] S106. Based on the layout plan of charging piles in commercial and residential areas, use linear programming algorithm to optimize power grid dispatch and determine the power distribution ratio of different areas.

[0142] Specifically, the power demand distribution is obtained from the charging pile layout schemes in commercial and residential areas. Combined with time periods and regional characteristics, a power demand set for each region is generated. If the peak charging demand in any region of the power demand set exceeds a preset threshold, it is marked as a high-load region, resulting in a high-load region set. A linear programming algorithm is used to analyze the high-load region set and grid resource constraints to generate a regional load balancing scheme and determine the power allocation ratio set for each region. Using the regional power allocation ratio set and the charging pile layout, the expected power supply for each charging pile is calculated, resulting in a power supply distribution set. If the expected power supply for a charging pile in the power supply distribution set is lower than a preset threshold, it is marked as a low-supply charging pile, resulting in a low-supply charging pile set. Based on the low-supply charging pile set and the regional load balancing scheme, the power allocation ratio is adjusted to generate an optimized dynamic allocation scheme. The stability of the power allocation ratio in each region is verified using the dynamic allocation scheme and grid resource constraints, resulting in a stable power allocation set. Using the stable power allocation set, the power dispatch configuration of charging piles in commercial and residential areas is updated to generate the final grid dispatch scheme.

[0143] For example, obtaining the electricity demand distribution for charging pile deployment schemes in commercial and residential areas can be achieved by analyzing user charging behavior data combined with regional characteristics. Commercial areas experience high charging demand during weekday daytime hours, while residential areas see demand concentrated at night and on weekends. Suppose a commercial area experiences a peak charging demand of 500kW between 9:00 AM and 11:00 AM, exceeding a preset threshold of 400kW, and can be marked as a high-load area. Similarly, a residential area experiences a peak demand of 450kW between 8:00 PM and 10:00 PM, also marked as a high-load area. Data sources can include charging pile usage records and smart meter data, ensuring accurate reflection of time periods and regional characteristics.

[0144] In this embodiment, when using a linear programming algorithm to analyze the high-load area set and grid resource constraints, the objective can be set as minimizing the power allocation deviation. Assuming the total grid capacity is 2000kW, 600kW needs to be allocated to the commercial area during high-load periods, and 500kW to the residential area. Through algorithm optimization, the grid is ensured not to be overloaded. The power allocation ratio set can be obtained as 55% for the commercial area and 45% for the residential area.

[0145] Specifically, if the expected power supply of a charging pile in a commercial area is 50kW, and it is lower than the threshold of 60kW, it will be marked as a low-supply charging pile and included in the low-supply charging pile set.

[0146] For example, when adjusting the power allocation ratio, data from low-supply charging piles can be used in conjunction with a regional load balancing scheme to increase the power input to these piles. Suppose a residential area's original charging pile supply was 40kW, and after adjustment, it is increased to 65kW, optimizing the dynamic allocation scheme. To verify stability, grid load fluctuations during peak hours can be simulated to check if the allocation ratio causes voltage instability. A stable power allocation set can ensure that volatility in both commercial and residential areas is below 5%.

[0147] In this embodiment, when updating the power dispatch configuration of charging piles, the daytime dispatch priority for commercial areas can be adjusted based on a stable power allocation set, while the nighttime priority for residential areas can be prioritized. Assuming that the daytime dispatch power for a charging pile in a commercial area is increased to 70kW, and the nighttime dispatch power in a residential area is increased to 60kW, the final grid dispatch scheme can balance supply and demand, improve the utilization efficiency of charging piles, and reduce grid pressure.

[0148] Understandably, the above scheme ensures accurate and efficient power distribution, reduces inefficient charging piles, and improves the overall stability of the power grid through regional characteristics, time period analysis, and algorithm optimization.

[0149] S107. Extract personalized service needs from users' professional background and charging behavior patterns, and use decision tree algorithm to generate personalized charging time and location recommendations for users, thus obtaining personalized charging service solutions.

[0150] Specifically, user behavior data is obtained from users' professional backgrounds and charging behavior patterns. Feature extraction methods are used to analyze occupational types and daily activity patterns to obtain a user behavior feature set. Based on the user behavior feature set, time preference analysis is used to calculate the probability of users' charging needs at different time periods, resulting in a charging time preference set. Combining the charging time preference set with the user behavior feature set, location preference analysis is used to match users' frequent locations and activity areas, resulting in a charging location preference set. A decision tree algorithm is then used, inputting the charging time preference set and the charging location preference set, to generate personalized charging time and location recommendations for each user, resulting in a preliminary charging service set. If there are mismatches between the recommendations in the preliminary charging service set and the user behavior data, the time and location parameters are adjusted through comparative analysis to obtain an optimized charging service set. Finally, by combining the optimized charging service set with pre-set charging pile availability data, the feasibility of the recommended solutions is verified, resulting in the final charging service plan.

[0151] For example, when acquiring user behavior data, smart charging station systems can collect users' occupational background and charging behavior patterns. Occupational background includes occupation type, such as white-collar worker, driver, or freelancer; charging behavior patterns cover charging time, location, and frequency. Suppose a user is a white-collar worker who leaves their residential area at 8 am every morning and returns at 6 pm, with a charging habit of charging at home overnight. The system extracts this information through questionnaires or historical charging records to form a user behavior feature set, including occupation type "white-collar worker," activity pattern "fixed commuting," and charging preference "at home overnight." This feature extraction method relies on data cleaning and cluster analysis to ensure that the data accurately reflects user habits.

[0152] In this embodiment, the time preference analysis method calculates the probability of charging demand based on a user behavior feature set. For example, analyzing a user's charging records for a week reveals that white-collar users have a 70% probability of charging between 8 PM and 10 PM on weekdays and a 20% probability of charging on weekend afternoons. Through statistical analysis, a charging time preference set is generated, clearly indicating that users are more likely to charge at night. This method utilizes time series analysis, combined with historical data, to predict future demand, improving the accuracy of charging services.

[0153] For example, location preference analysis combines charging time preference sets and user behavior feature sets to match users' frequent locations. Assuming a user resides in residential area A and works in commercial area B, the system analyzes GPS positioning and charging records to determine that 90% of the user's charging activity occurs at charging stations in residential area A. This generates a charging location preference set, recommending nighttime charging stations in residential area A. This method uses spatial analysis techniques to ensure that recommended locations closely match the user's activity trajectory.

[0154] In this embodiment, the decision tree algorithm can be used to generate user-specific charging time and location recommendations. Given a set of charging time preferences and a set of charging location preferences, the algorithm generates a preliminary charging service set based on conditional branches (e.g., "If the user is a white-collar worker, then prioritize recommending nighttime residential charging").

[0155] For example, it might recommend that white-collar users charge at charging stations in residential area A at 8 PM. If the recommendation doesn't match the user's actual behavior, such as charging late at night due to overtime work, the time parameters are adjusted through comparative analysis to generate an optimized charging service set, recommending charging at 10 PM.

[0156] For example, when verifying the feasibility of a recommended solution, the system combines charging pile availability data to check whether there are any available charging ports in residential area A between 8 PM and 10 PM. Assuming residential area A has 10 charging piles and an 80% availability rate at 8 PM, the system confirms the feasibility of the recommended solution and ultimately generates a charging service plan. This verification method, through real-time data analysis, ensures that the recommended solution matches actual resources, improving service reliability.

[0157] Understandably, the above methods generate personalized charging solutions through detailed analysis and multi-dimensional matching of user behavior data. Each step is closely integrated with the actual needs of users and the status of charging pile resources to ensure the efficiency and feasibility of the solution. Especially in high-load areas, optimizing charging time and location recommendations can effectively divert demand and improve grid dispatch efficiency.

[0158] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for recognizing electric vehicle charging behavior based on user profiles, characterized in that, Includes the following steps: Acquire user profile features and charging behavior data and preprocess them to obtain standardized user feature datasets and charging behavior datasets; Based on the standardized user feature dataset and charging behavior dataset, user profile features are classified to obtain classification results of user groups' occupational background and geographical location preferences; Spatiotemporal dynamic characteristics are extracted from the charging behavior dataset, and time series analysis is used to decompose the charging frequency and charging duration distribution to obtain the charging behavior patterns of users in different time periods and locations. Based on charging behavior patterns and combined with geographical location preference classification results, we can obtain charging pile layout schemes for commercial and residential areas. Based on the charging pile layout plan for commercial and residential areas, a linear programming algorithm is used to optimize the power grid dispatch and determine the power allocation ratio for different areas. Personalized service needs are extracted from users' professional backgrounds and charging behavior patterns. A decision tree algorithm is used to generate personalized charging time and location recommendations for each user, resulting in a personalized charging service plan.

2. The method according to claim 1, characterized in that, The standardized user feature dataset and charging behavior dataset obtained include: User profile features and charging behavior data are obtained from the user database. SQL query language is used to extract data including user ID, age, gender, occupation, charging time, charging frequency and charging location to obtain the raw dataset. For the original dataset, the mean imputation method is used to handle missing values. If the missing rate of a field exceeds a preset threshold, the field is removed to obtain a preliminary cleaned dataset. Outlier detection was performed on the initially cleaned dataset using a standard deviation-based method. If a data point exceeded three times the standard deviation of the mean, it was marked as an outlier and removed, resulting in an outlier-free dataset. Based on the dataset without anomalies, the numerical features are normalized using the Z-score normalization method to generate a standardized user feature dataset and a charging behavior dataset.

3. The method according to claim 1, characterized in that, The obtained user group occupational background and geographic location preference classification results include: The K-means clustering algorithm is used to process the standardized user feature dataset to determine the user classification results; Based on the user classification results, the occupational background features of each classification group are extracted to obtain the occupational distribution feature set; By analyzing the charging behavior preferences of each category group in terms of geographical location through the occupational distribution feature set, the location preference distribution can be obtained.

4. The method according to claim 1, characterized in that, The obtained user charging behavior patterns at different time periods and locations include: Spatiotemporal dynamic features are obtained from the charging behavior dataset, and the charging frequency and charging duration distributions are separated by the time series decomposition method to determine the user's behavior pattern set in different time periods and locations. By analyzing the relationship between time period division and location distribution characteristics through behavioral pattern set, the time-location correlation coefficient is calculated by correlation analysis method to obtain spatiotemporal correlation rule set; If the charging frequency in a certain time period in the spatiotemporal association rule set is higher than a preset threshold, it is marked as a high-frequency time period, and a set of high-frequency time periods is obtained. Based on the high-frequency time period set, the corresponding location distribution characteristics are extracted, and the location distribution characteristics are grouped using cluster analysis to determine the location cluster set; By analyzing the characteristics of each cluster in terms of charging time distribution through location clustering sets, a set of time distribution characteristics can be obtained. If the concentration of charging time in a certain cluster of time distribution characteristics is higher than a preset threshold, it is marked as a high concentration group, and the high concentration group set is determined.

5. The method according to claim 1, characterized in that, The proposed layout scheme for charging stations in commercial and residential areas includes: When charging behavior patterns show that users tend to charge in commercial areas on weekdays, a weighted average algorithm combined with geographical location preference classification results is used to calculate the priority distribution of charging piles in commercial areas and obtain a charging pile layout scheme for commercial areas. When charging behavior patterns show that users tend to charge in residential areas on weekends, a density clustering algorithm is used to analyze residential charging preferences, generate the density of charging piles in residential areas, and obtain a residential charging pile layout scheme.

6. The method according to claim 5, characterized in that, The proposed layout scheme for charging stations in the commercial area includes: Using data collection methods, weekday tendency data is extracted from charging behavior patterns, and combined with time period distribution, the distribution of users' charging demand in commercial areas is determined. If the charging demand distribution shows that the charging frequency is higher than a preset threshold during a certain period, it is marked as a high-demand period, and the set of high-demand periods is obtained. Based on the high-demand time period set and combined with geographical location preferences, a weighted average algorithm is used to calculate the priority of each charging pile in the commercial area, and the priority distribution of charging piles is obtained. By analyzing the charging pile priority distribution, the utilization rate of charging piles is determined, charging piles with utilization rates below a preset threshold are identified, and a set of charging piles with low utilization rates is obtained. If the number of charging piles in the low-usage charging pile cluster exceeds a preset threshold, then a cluster analysis method is used to group the charging pile distribution to obtain a charging pile distribution cluster set. Based on the cluster set of charging pile distribution, combined with users' charging needs and location preferences, an optimization plan for the layout of charging piles in commercial areas is generated, and the optimized distribution of charging piles in commercial areas is determined.

7. The method according to claim 5, characterized in that, The obtained residential area charging pile layout scheme includes: Weekend residential charging data is obtained from charging behavior patterns, and charging demand distribution is generated by combining time period and geographical location information. If the charging frequency in any time period of the charging demand distribution is higher than a preset threshold, it is marked as a peak period, and a peak period set is obtained. Density clustering algorithm is used to analyze peak time period sets and geographical location distribution to generate the density distribution of charging pile configuration in residential areas; Based on the configuration density distribution, the expected utilization rate of each charging pile in the residential area is calculated to obtain the utilization rate distribution set; If there are charging piles in the usage distribution that have a usage rate lower than a preset threshold, they are marked as inefficient charging piles, and a set of inefficient charging piles is obtained. By adjusting the location of inefficient charging pile clusters and their density distribution, an optimized charging pile layout scheme for residential areas can be generated.

8. The method according to claim 1, characterized in that, Determining the power allocation ratio for different regions includes: The power demand distribution is obtained from the charging pile layout schemes in commercial and residential areas, and power demand sets for each area are generated by combining time periods and regional characteristics. If the peak charging demand in any region of concentrated electricity demand exceeds a preset threshold, it is marked as a high-load region, thus obtaining a set of high-load regions. A linear programming algorithm is used to analyze the high-load area set and power grid resource constraints, generate a regional load balancing scheme, and determine the power allocation ratio set for each region. By combining the regional power distribution ratio set with the layout of charging piles, the expected power supply of each charging pile is calculated, and the power supply distribution set is obtained. If the expected power supply of charging piles in a concentrated power supply distribution is lower than a preset threshold, they are marked as low-supply charging piles, and a set of low-supply charging piles is obtained. Based on the low-supply charging pile cluster and regional load balancing scheme, the power allocation ratio is adjusted to generate an optimized dynamic allocation scheme. By using dynamic allocation schemes and grid resource constraints, the stability of power allocation ratios in different regions is verified, and a stable power allocation set is obtained. Using a stable power distribution set, the power dispatch configuration of charging piles in commercial and residential areas is updated to generate the final power grid dispatch scheme.

9. The method according to claim 1, characterized in that, The personalized charging service solution includes: User behavior data is obtained from users' professional background and charging behavior patterns. Feature extraction methods are used to analyze the occupation type and daily activity patterns to obtain a user behavior feature set. Based on the user behavior feature set, the time preference analysis method is used to calculate the probability of users' charging needs in different time periods, and obtain the charging time preference set. By combining the charging time preference set with the user behavior feature set, and using the location preference analysis method, the user's permanent location and activity area are matched to obtain the charging location preference set; Using a decision tree algorithm, the system takes a set of charging time preferences and a set of charging location preferences as input, generates a user-specific charging time and location recommendation scheme, and obtains a charging service scheme.

Images