Charging demand prediction method based on regional distribution of user portraits

Abstract

The invention relates to a charging demand prediction method based on regional distribution of user portraits, and belongs to the technical field of electric power. The method comprises the following steps: collecting geographical location information and charging behavior data of a user, and preprocessing the collected data; analyzing the preprocessed data, and creating a portrait of the user by using the charging place, the charging frequency, the charging time and the charging electric quantity of the user; carrying out regional distribution analysis on the user portraits, and researching user charging demands and behavior modes of different regions; and predicting the charging demand of the specific user group based on the created user portrait and the regional distribution analysis result. According to the invention, through the analysis of the user portrait and the regional distribution, the future charging demand can be predicted more accurately, and the situation of resource waste or short supply caused by inaccurate estimation is effectively avoided.

Description

Technical Field

[0001] The present invention belongs to the field of electric power technology, and specifically relates to a charging demand prediction method based on the regional distribution of user portraits. Background Art

[0002] In the development of the new energy vehicle market, the construction and distribution of charging facilities is an important component. Due to factors such as vehicle type, driving habits, and travel distance, the charging needs of different users vary greatly, resulting in significant differences in the spatial distribution characteristics of charging facility demand.

[0003] In the past few years, the prediction of charging demand has mainly relied on experience and observation. For example, some city or regional power grid companies and property companies will estimate future charging demand based on local population density, number of vehicles, electric vehicle sales data and other information. However, this method is difficult to consider the specific behavior habits and needs of each individual user.

[0004] In recent years, with the development of big data and artificial intelligence technology, it has become possible to integrate and analyze a large amount of user data. Among them, user profiling is an effective means to deeply analyze and understand user needs based on information such as user behavior habits and consumption habits. At the same time, regional analysis is often used to understand the geographical distribution characteristics of users, thereby guiding decisions related to geographical location.

[0005] However, in the current user portrait analysis, regional analysis and charging demand forecasting research, there is no systematic application of these technologies to the research on electric vehicle charging demand forecasting. Therefore, how to overcome the shortcomings of existing technologies is an urgent problem to be solved in the current power technology field. Summary of the invention

[0006] The purpose of the present invention is to address the deficiencies of the prior art and to provide a charging demand prediction method based on the regional distribution of user portraits. The method is expected to improve the accuracy of charging demand prediction and guide the planning and layout of charging facilities.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0008] A charging demand prediction method based on user profile regional distribution includes the following steps:

[0009] Step 1: Collect user data: collect user's geographic location information and charging behavior data, and pre-process the collected data;

[0010] Step 2: Create a user profile: Analyze the data pre-processed in step 1 and use the user's charging location, charging frequency, charging time, and charging power to create a user profile;

[0011] Step 3: Regional distribution analysis: Conduct regional distribution analysis on user portraits to study user charging needs and behavior patterns in different regions; build a prediction model for future charging demand in specific regions using time series analysis or multivariate linear regression analysis;

[0012] Step 4: Charging demand prediction: Obtain the user portrait created in step 2 for the region to be predicted and the results of the regional distribution analysis in step 3, and use the prediction model constructed in step 3 to predict the charging demand in each region.

[0013] Furthermore, preferably, in step one, preprocessing includes missing value processing, outlier processing and data standardization.

[0014] Furthermore, preferably, the interpolation method is used for missing value processing; the box plot method is used for outlier processing; and the z-score normalization method is used for data normalization.

[0015] Furthermore, preferably, in step one, the collected data also needs to be desensitized.

[0016] Further, preferably, the specific method of step 2 is:

[0017] (2.1) Feature selection: Select features to describe user charging behavior and needs;

[0018] (2.2) Data analysis: Analyze the features selected in step (2.1), including descriptive statistics analysis and correlation analysis;

[0019] (2.3) Profile creation: Create a user profile based on the data analysis results of step (2.2);

[0020] The user portrait includes the following contents: user charging frequency, user charging time, user charging power and charging location.

[0021] Furthermore, preferably, in (2.1), the selected features include the user's charging location, charging frequency, charging time and charging power; in (2.2), the descriptive statistical analysis includes the mean and standard deviation; when creating the portrait, the mean and standard deviation of the charging location, charging frequency, charging time and charging power are used to create the portrait.

[0022] Further, preferably, the specific method of step three is:

[0023] (3.1) Based on the user portrait data, the regional distribution analysis of the user portrait is conducted to study the charging needs and behavior patterns of users in different regions to obtain regional characteristic data; at the same time, the region to be analyzed is determined; the regional characteristic data includes region, population density, traffic conditions and economic conditions;

[0024] (3.2) Compare and analyze the distribution of charging demand in various regions, and construct a prediction model for charging demand in a specific region in the future using time series analysis or multivariate linear regression analysis.

[0025] Furthermore, preferably, the time series analysis adopts the ARIMA model; and the multiple linear regression analysis adopts the multiple linear regression model.

[0026] Furthermore, preferably, in the ARIMA model, user profile data is used as input, and charging demand in the region in the future period is used as output for training;

[0027] In the multiple linear regression model, charging demand is used as the dependent variable Y, and region, population density, traffic conditions, economic conditions and user portrait data are used as independent variables X for modeling.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. Improve accuracy: Through the analysis of user portraits and geographical distribution, future charging demand can be predicted more accurately, effectively avoiding resource waste or supply shortages caused by inaccurate estimates.

[0030] 2. Enhance personalization: By deeply understanding the user's charging behavior patterns and charging needs, we can provide more personalized services and improve the user's charging experience and service satisfaction.

[0031] 3. Optimize resource allocation: By accurately predicting charging demand, the construction of charging facilities can be scientifically planned and deployed, which helps to achieve the rational allocation of power resources and optimize efficiency.

[0032] 4. Assist the development of new energy vehicles: Correctly predicting charging demand and rationally planning charging facilities will reduce users’ concerns about the difficulty of charging electric vehicles, stimulate the development of the new energy vehicle market, and also help reduce oil consumption and reduce environmental pollution.

[0033] 5. Realize data-driven: Through the analysis and application of big data, valuable information can be extracted from massive data to drive precise decision-making and provide effective decision-making support for charging operators.

[0034] When using the traditional prediction model, the prediction accuracy is 70%, while after using the charging demand prediction method based on user portrait regional distribution of the present invention, the prediction accuracy is increased to 85%. Accurate prediction can help allocate resources more effectively and reduce waste caused by inaccurate predictions. For example, after using the prediction method of the present invention, each charging station can save 1,000 yuan in electricity costs per month on average because of the reduction of idle time and insufficient power supply caused by inaccurate predictions. If a city has 100 charging stations, it can save 100,000 yuan in electricity costs per month.

[0035] On the other hand, personalized services for users can improve user satisfaction. For example, after using the prediction method of the present invention, through a user satisfaction survey, the user satisfaction with the charging service increased from 70% to 90%.

[0036] Furthermore, based on the prediction results, resource optimization configuration can achieve faster charging services. After using the prediction method of the present invention, the time users wait for charging is shortened by 20% on average, which can greatly improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flow chart of a charging demand prediction method based on user profile regional distribution. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below in conjunction with embodiments.

[0039] Those skilled in the art will appreciate that the following examples are only used to illustrate the present invention and should not be considered to limit the scope of the present invention. If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the art or the product specifications are used. If the manufacturer of the materials or equipment used is not specified, they are all conventional products that can be purchased.

[0040] Example 1

[0041] like Figure 1 As shown, a charging demand prediction method based on user portrait regional distribution is characterized by comprising the following steps:

[0042] Step 1: Collect user data: collect user's geographic location information and charging behavior data, and pre-process the collected data;

[0043] Step 2: Create a user profile: Analyze the data pre-processed in step 1 and use the user's charging location, charging frequency, charging time, and charging power to create a user profile;

[0044] Step 3: Regional distribution analysis: Conduct regional distribution analysis on user portraits to study user charging needs and behavior patterns in different regions; build a prediction model for future charging demand in specific regions using time series analysis or multivariate linear regression analysis;

[0045] Step 4: Charging demand prediction: Obtain the user portrait created in step 2 for the region to be predicted and the results of the regional distribution analysis in step 3, and use the prediction model constructed in step 3 to predict the charging demand in each region.

[0046] Example 2

[0047] like Figure 1 As shown, a charging demand prediction method based on user profile regional distribution includes the following steps:

[0048] Step 1: Collect user data: collect user's geographic location information and charging behavior data, and pre-process the collected data;

[0049] Step 2: Create a user profile: Analyze the data pre-processed in step 1 and use the user's charging location, charging frequency, charging time, and charging power to create a user profile;

[0050] Step 3: Regional distribution analysis: Conduct regional distribution analysis on user portraits to study user charging needs and behavior patterns in different regions; build a prediction model for future charging demand in specific regions using time series analysis or multivariate linear regression analysis;

[0051] Step 4: Charging demand prediction: Obtain the user portrait created in step 2 for the region to be predicted and the results of the regional distribution analysis in step 3, and use the prediction model constructed in step 3 to predict the charging demand in each region.

[0052] In step one, preprocessing includes missing value processing, outlier processing and data standardization.

[0053] Interpolation method was used to deal with missing values; box plot method was used to deal with outliers; and z-score standardization method was used for data standardization.

[0054] In step one, the collected data also needs to be desensitized.

[0055] The specific method of step 2 is:

[0056] (2.1) Feature selection: Select features to describe user charging behavior and needs;

[0057] (2.2) Data analysis: Analyze the features selected in step (2.1), including descriptive statistics analysis and correlation analysis;

[0058] (2.3) Profile creation: Create a user profile based on the data analysis results of step (2.2);

[0059] The user portrait includes the following contents: user charging frequency, user charging time, user charging power and charging location.

[0060] In (2.1), the selected features include the user's charging location, charging frequency, charging time and charging power; in (2.2), the descriptive statistical analysis includes the mean and standard deviation; when creating the portrait, the mean and standard deviation of the charging location, charging frequency, charging time and charging power are used to create the portrait.

[0061] The specific method of step three is:

[0062] (3.1) Based on the user portrait data, the regional distribution analysis of the user portrait is conducted to study the charging needs and behavior patterns of users in different regions to obtain regional characteristic data; at the same time, the region to be analyzed is determined; the regional characteristic data includes region, population density, traffic conditions and economic conditions;

[0063] (3.2) Compare and analyze the distribution of charging demand in various regions, and construct a prediction model for charging demand in a specific region in the future using time series analysis or multivariate linear regression analysis.

[0064] The ARIMA model was used for time series analysis, and the multiple linear regression model was used for multiple linear regression analysis.

[0065] In the ARIMA model, user profile data is used as input, and the charging demand in the region in the future period is used as output for training;

[0066] In the multiple linear regression model, charging demand is used as the dependent variable Y, and region, population density, traffic conditions, economic conditions and user portrait data are used as independent variables X for modeling.

[0067] Example 3

[0068] A charging demand prediction method based on the regional distribution of user portraits mainly determines the charging needs and preferences of users in different regions through observation and analysis of the regional distribution of user portraits, and then predicts future charging demand based on this information.

[0069] This method mainly includes the following steps:

[0070] Step 1: Collect user data

[0071] Collect user's geographic location information and charging behavior data. This data can be obtained from various sources, such as mobile phone applications, website access records, and vehicle GPS systems.

[0072] The detailed implementation process of collecting user data is as follows:

[0073] 1.1 Data source selection and access: User data can come from different sources, including charging records at charging stations and mobile apps used by users. Data access usually requires an agreement with the data provider to meet the privacy protection requirements of the data.

[0074] 1.2 Data acquisition and processing: The acquired user data needs to undergo certain processing, such as missing value processing, outlier processing, and data standardization. These data preprocessing steps help improve the quality and accuracy of subsequent data analysis.

[0075] Missing value processing: Interpolation method is used. For data with time series characteristics, linear interpolation can be performed by referring to nearby data (such as previous and next time points). The formula for linear interpolation is: Y = Y1 + (X-X1) * (Y2-Y1) / (X2-X1), where (X, Y) is the point to be interpolated, and (X1, Y1) and (X2, Y2) are two known points.

[0076] Outlier handling: box plot method.

[0077] The box plot method is based on the five-number summary (minimum value, lower quartile Q1, median, upper quartile Q3, maximum value), and regards "outliers" (data less than Q1-1.5IQR, or greater than Q3+1.5IQR) outside the upper and lower boxes as abnormal values. IQR = Q3-Q1.

[0078] Data standardization: z-score standardization (zero-mean normalization): The formula is z = (x-μ) / σ, where x is the original value, μ is the mean, and σ is the standard deviation.

[0079] 1.3 Data storage and protection: User data needs to be stored in a database that can be used for mathematical analysis and model training, and must comply with data protection regulations, which means that sensitive data needs to be desensitized.

[0080] Step 2: Create a user profile

[0081] By analyzing the collected user data, such as location, charging frequency, charging time and power, a user profile is created. It can be used to represent the characteristics and patterns of user charging behavior.

[0082] 2.1 Feature selection: Based on business needs and data analysis results, determine the features used to describe user charging behavior and needs, such as location, charging frequency, charging time, and charging power.

[0083] 2.2 Data analysis: Apply mathematical and statistical methods to analyze the above characteristics, including descriptive statistical analysis (such as mean, standard deviation) and correlation analysis.

[0084] Correlation analysis: Use the Pearson correlation coefficient to measure the linear relationship between two features. The formula for the Pearson correlation coefficient is as follows:

[0085]

[0086] Among them, xi and yi are the observed values ​​of the two features, μx and μy are the means of the two features. The value range of the Pearson correlation coefficient is -1 to 1, 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no linear relationship.

[0087] Calculate the correlation coefficient between charging frequency and charging capacity. If the correlation coefficient is close to 1, it means that the more frequently users charge, the more they charge. This information is very useful for predicting users' charging needs.

[0088] 2.3 Portrait creation: Create a user portrait based on the data analysis results. User portrait is an abstraction and summary of the user, describing the user's typical behavior and demand patterns.

[0089] The application of mathematical formulas is mainly in the data analysis stage. For example, the average formula can be used to calculate the average charging frequency of users:

[0090] Average charging frequency = total charging times / total time period

[0091] The standard deviation formula can be used to calculate the degree of fluctuation in the user's charging frequency:

[0092] The average value of .

[0093] The final user portrait includes the following: the frequency of user charging and its fluctuation (for example, users charge 3 times a week on average, and the fluctuation is not large), the user's charging time (for example, users prefer to charge between 5pm and 7pm every day), the user's charging power and location (for example, users need about 50 kWh of power each time they charge, and prefer to charge at a charging station near their home), etc. Through user portrait analysis, we can have a deeper understanding of users' charging needs and behavior patterns, thus providing a basis for predicting charging needs.

[0094] In step 2.1, four features representing the user's charging behavior are first selected: charging location, charging frequency, charging time, and charging power. These four features should be sufficient to describe a user's charging behavior habits in most cases and provide a basis for creating a profile.

[0095] In step 2.2, descriptive statistical analysis is usually performed on the four selected features, that is, the mean and standard deviation of these features are calculated. The mean can be used to obtain the average level of each feature, which can help understand the general charging behavior of users; the standard deviation can be used to understand the degree of change of each feature, which reflects the stability or heterogeneity of the user's charging behavior.

[0096] For the four features of charging location, charging frequency, charging time and charging power, their means and standard deviations are calculated. The mean of charging frequency can help understand how often users usually charge, and its standard deviation reflects the stability of charging frequency. The mean of charging time can help understand in which time period users prefer to charge, and its standard deviation reflects the stability of charging time. The mean of charging power can help understand how much power users usually need for each charge, and its standard deviation reflects the stability of charging power. Similarly, the mean and standard deviation are calculated for charging locations. Specifically, the charging location can be recorded as longitude and latitude coordinates, the mean reflects the user's main charging location, and the standard deviation reflects the user's charging situation at different locations.

[0097] For example, for the charging frequency feature, the above calculation formula can be used to obtain the mean and standard deviation of the charging frequency. Then, the same operation can be performed on the charging time, charging power and other features to obtain the mean and standard deviation of all features. These are all part of calculating the user profile. Then, through data analysis and prediction models, the user's charging needs can be predicted.

[0098] For example, if the number of times a user charges in a week is collected as [3, 4, 2, 3, 3, 4, 2], then the average number of times he charges in a week and the standard deviation can be calculated as follows:

[0099] The mean value μ of charging frequency:

[0100] μ=(3+4+2+3+3+4+2) / 7=21 / 7=3

[0101] The standard deviation of charging frequency σ:

[0102] σ=√[(3-3)^2+(4-3)^2+(2-3)^2+(3-3)^2+(3-3)^2+(4-3)^2+(2-3)^2] / 7=√[0+1+1+0+0+1+1] / 7=√4 / 7≈0.75

[0103] Therefore, the mean of this user’s charging frequency is 3 and the standard deviation is 0.75.

[0104] For example, in the calculation of charging power, suppose the power of a user's charging every time in a week is recorded, and the data obtained is [50, 45, 55, 52, 46, 50, 48] kWh. The average charging power and standard deviation of this user can be calculated to represent his charging behavior. The specific calculation process is as follows:

[0105] The mean value μ of the charging capacity:

[0106] μ=(50+45+55+52+46+50+48) / 7=346 / 7=49.42 (round to two decimal places)

[0107] The standard deviation of the charge capacity σ:

[0108] σ=√[(50-49.42)^2+(45-49.42)^2+(55-49.42)^2+(52-49.42)^2+(46-49.42)^2+(50-49 .42)^2+(48-49.42)^2] / 7=√[0.34+19.63+31.11+6.59+11.92+0.34+2] / 7=√71.93 / 7≈3.21

[0109] That is, we can get that the average amount of electricity consumed by this user per charging arrow is 49.42 kWh, and the standard deviation is 3.21, which describes his charging behavior.

[0110] Next, in step 2.3, a user profile is created based on the results of step 2.2. Specifically, each user's profile can be composed of the mean of its charging location, charging frequency, charging time, and charging power, as well as the standard deviation of these features. Both the mean and the standard deviation provide information about user behavior and can help better understand and predict users' charging needs.

[0111] In summary, through feature selection, analysis, and descriptive statistical analysis of user data, we can obtain the user's charging behavior pattern and integrate this information into the user profile, thus providing a basis for subsequent predictions. This is how steps 2.1, 2.2, and 2.3 are connected and work together to support the creation of user profiles.

[0112] Step 3: Regional distribution analysis

[0113] Conduct regional distribution analysis on user portraits to study user charging needs and behavior patterns in different regions. For example, users in city centers may need faster charging services, while users in suburban areas may prefer to charge at home or at work.

[0114] 3.1 Based on the user portrait data, determine the area to be analyzed, which can be the city level, district level, or a smaller scope, such as a street or community.

[0115] 3.2 Compare and analyze the distribution of charging demand in different regions, and use time series analysis to predict the future charging demand in specific regions or use multivariate linear regression analysis to predict the future charging demand in specific regions. These analysis strategies are specifically dedicated to modeling the time dependence of regional charging demand and the relationship between different variables.

[0116] Time series analysis implementation process:

[0117] The input of the ARIMA model is user profile data, including data on user charging frequency, charging time, charging power and charging location in different regions. The charging demand forecast for the future can be obtained through historical data. The output of the model is the charging demand forecast for each region under the given user profile conditions in the future.

[0118] 3.2.1 Data collection: Charging demand data at different time points over a period of time needs to be collected, including:

[0119] Charging start and end time: Users generally have fixed charging habits and times, such as charging before the morning rush hour or after the evening rush hour.

[0120] Charging capacity: the amount of electricity required for each charging, as well as the remaining power of the user's vehicle and other information.

[0121] Charging type: fast charging or slow charging, general electric vehicles may have different charging needs.

[0122] Charging location: The charging location or station selected by the user, which may be related to the user's residence, workplace, or frequented places.

[0123] 3.2.2 Based on the collected time series data, descriptive statistical analysis can be performed on the data, such as calculating the mean, variance, maximum and minimum values ​​of the demand.

[0124] 3.2.3 Introduce the ARIMA model (AutoRegressive Integrated Moving Average model) for time series forecasting. The ARIMA model consists of three parts: autoregressive model (AR), difference (I) and moving average (MA). The ARIMA model formula is as follows:

[0125]

[0126] Among them, X tis the observed value of the current period (period t); are the parameters of the AR model; θ j is the parameter of the MA model; Z t is the error term, which is assumed to be a white noise series; μ is a constant term. AR stands for autoregression, I stands for non-seasonal difference, and MA stands for moving average.

[0127] Use the ARIMA time series model to predict future total charging demand: The ARIMA model is a powerful time series forecasting model that can capture the time dependency in the data, so it is often used to predict future demand or sales, etc. In this example of predicting charging demand, the ARIMA model can first be applied to learn the historical total charging demand data and predict the future total charging demand.

[0128] The training process of the ARIMA model mainly includes two steps: determining the model parameters and fitting the model. The determination of parameters is mainly assisted by ACF (Autocorrelation Function) and PACF (Partial Autocorrelation Function) graphs. ACF is used to determine the order of the MA (moving average) part, and PACF is used to determine the order of the AR (autoregressive) part. After determining the parameters, the least squares method or maximum likelihood method is used to estimate the parameters to obtain the ARIMA model.

[0129] In addition, multiple linear regression analysis can also be used to predict the charging demand in a specific region in the future. The implementation process is as follows (it is parallel to the time series analysis above, and the two models are used as two prediction technology solutions):

[0130] The input variables of the multivariate linear regression model include region, population density, traffic conditions and economic conditions, as well as user profile data (including user charging frequency, charging time, charging power, charging location and other characteristic data). The output of the model is the predicted value of charging demand in various regions in the future after considering multiple influencing factors.

[0131] (1) First, we need to collect data on relevant variables including region, population density, traffic conditions, and economic conditions.

[0132] (2) Charging demand is taken as the dependent variable Y (model output), and region, population density, traffic conditions, economic conditions, and user profile data are taken as the independent variable X (model input).

[0133] Charging demand refers to the number of charging facilities or the amount of electricity required by users in a specific area. The specific data is comprehensively evaluated based on the recent facility utilization rate, facility load status, and feedback from facility users.

[0134] Regional data mainly refers to geographical location information, such as latitude, longitude, altitude, etc. Different regions and areas can be divided according to the geographical location information of charging facilities.

[0135] Population density can be obtained directly through statistical data, such as citizen population statistics, annual census, etc.

[0136] Traffic conditions can be the number of streets, the number of main roads, the daily traffic volume, the number of bus stops, and the number of subway stations.

[0137] Economic conditions can include regional per capita income, per capita consumption, regional GDP, and regional housing prices.

[0138] (3) Establish a multiple linear regression model: This model assumes that there is a linear relationship between the dependent variable Y and the independent variable X.

[0139]

[0140] Among them, Y is the dependent variable, X n is the independent variable, β0 is the intercept, β n is the independent variable X n The coefficients describing X n The amount by which Y changes when Y changes by one unit. is the error term.

[0141] (4) Solve β using the least squares method or other methods n The estimated value of The value of is the smallest. The least squares estimation formula is:

[0142] β′=(X T X) (-1) X T Y

[0143] Among them, X, Y are matrices, -1 represents the inverse matrix, X T X is reversible.

[0144] In the present invention, the ARIMA model and the multivariate linear regression model can be used simultaneously, and the means of the two are used as the final prediction result values.

[0145] Step 4: Charging demand prediction

[0146] Based on the results of the user profile and regional distribution analysis, the model can predict future charging demand. If the user profile of a region shows that their demand for fast charging is increasing, then this region may need more fast charging stations in the future.

[0147] 4.1: Determine the prediction model: The previous steps have created user profiles and regional distribution information, as well as prediction models based on multivariate linear regression and time series analysis. These models can be combined or integrated to form a comprehensive prediction model.

[0148] 4.2: Model prediction: Use the established prediction model to predict the charging demand in the future.

[0149] 4.4: Forecast result analysis: Analyze the forecast results. For example, you can compare the difference between the forecast value and the actual value, calculate the accuracy of the forecast, or analyze which regions have the fastest growing charging demand.

[0150] Through the above steps, the future charging demand can be predicted, thus providing decision support for the planning and operation of electric vehicle charging stations.

[0151] Application Examples

[0152] First, we collected charging data from 50,000 users in a certain region over the past year, including the users’ charging locations, charging frequency, charging time, and charging power.

[0153] After user data is collected, the data is preprocessed to handle missing values ​​and outliers in the data and standardized. After processing, the data is stored in the database, analyzed and features are extracted from the data to create a user profile. The selected features include the user's charging location, charging frequency, charging time, and charging power. For example, it is found that a user charges an average of 3 times a week, most of the time between 5pm and 7pm, and each time requires 50kW of power.

[0154] Next, the regional distribution analysis of all user data was conducted, and it was found that users in urban center areas have an increased demand for fast charging, while users in suburban areas are more inclined to charge at home or at work.

[0155] Based on the regional distribution of user portraits and the results of regional distribution analysis, a model for predicting charging demand was established. The model uses the ARIMA time series model.

[0156] The ARIMA model is used to learn historical data and predict the total charging demand for the next month.

[0157] Finally, the results of the prediction model are analyzed. For example, the difference between the prediction results and the actual charging demand is compared, and the accuracy of the prediction is calculated. For example, it is found that the predicted value of the charging demand in a certain city center is 20,000kW, while the actual charging demand is 21,000kW, and the prediction accuracy is 95.24%.

[0158] After analysis, it was found that the charging demand in suburbs grew slowly, while the charging demand in city centers grew rapidly. Through these analysis results, we will be able to better plan and operate charging stations to meet the charging needs of users in different regions.

[0159] In summary, the charging demand prediction method based on the regional distribution of user portraits can better understand user needs, provide data support for charging station design, and improve the operational efficiency of charging stations.

[0160] The present invention has the following advantages:

[0161] Personalized user portrait: The present invention constructs a personalized user portrait based on the user's geographic location information and charging behavior data. Compared with the traditional method of predicting based on grid load or vehicle registration volume, it can better reflect the user's actual charging needs and preferences.

[0162] In-depth analysis of regional distribution: Through regional distribution analysis, we can observe and study the charging needs and behavior patterns of users in different regions. This regionalized demand analysis model can provide a more accurate decision-making basis for the planning and operation of charging stations in specific areas.

[0163] Combining time series analysis and multiple linear regression analysis: This prediction method combines time series analysis and multiple linear regression analysis. This combination can comprehensively consider and predict charging demand from both time and space dimensions, thereby improving the accuracy of the prediction.

[0164] Flexible forecasting model selection: Users can choose forecasting models based on multiple linear regression, time series analysis, or a combination of the two according to their needs. This provides users with more flexible choices and also improves the reliability of forecast results.

[0165] Strong decision support function: The forecast results can provide rich information for electric vehicle charging stations to make decisions, such as comparing the difference between the forecast value and the actual value, calculating the accuracy of the forecast, or analyzing which regions have the fastest growing charging demand, etc. These analyses and information can provide more comprehensive decision support for the investment, planning and operation of charging stations.

[0166] In summary, the method for predicting user charging demand of the present invention is more personalized, more regionally specific, more accurate in prediction, and provides more powerful decision support.

[0167] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A charging demand prediction method based on user profile regional distribution, characterized in that: The steps include: Step 1: Collect user data: collect user's geographic location information and charging behavior data, and pre-process the collected data; Step 2: Create a user profile: Analyze the data pre-processed in step 1 and use the user's charging location, charging frequency, charging time, and charging power to create a user profile; Step 3: Regional distribution analysis: Conduct regional distribution analysis on user portraits to study user charging needs and behavior patterns in different regions; build a prediction model for future charging demand in specific regions using time series analysis or multivariate linear regression analysis; Step 4: Charging demand prediction: Obtain the user portrait created in step 2 for the region to be predicted and the results of the regional distribution analysis in step 3, and use the prediction model constructed in step 3 to predict the charging demand in each region.

2. The charging demand prediction method based on user profile regional distribution according to claim 1 is characterized in that: In step one, preprocessing includes missing value processing, outlier processing and data standardization.

3. The charging demand prediction method based on user profile regional distribution according to claim 1 is characterized in that: Interpolation method was used to deal with missing values; box plot method was used to deal with outliers; and z-score standardization method was used for data standardization.

4. The charging demand prediction method based on user profile regional distribution according to claim 1 is characterized in that: In step one, the collected data also needs to be desensitized.

5. The charging demand prediction method based on user profile regional distribution according to claim 1 is characterized in that: The specific method of step 2 is: (2.1) Feature selection: Select features to describe user charging behavior and needs; (2.2) Data analysis: Analyze the features selected in step (2.1), including descriptive statistics analysis and correlation analysis; (2.3) Profile creation: Create a user profile based on the data analysis results of step (2.2); The user portrait includes the following contents: user charging frequency, user charging time, user charging power and charging location.

6. The charging demand prediction method based on user profile regional distribution according to claim 5 is characterized in that: In (2.1), the selected features include the user's charging location, charging frequency, charging time and charging power; in (2.2), the descriptive statistical analysis includes the mean and standard deviation; when creating the portrait, the mean and standard deviation of the charging location, charging frequency, charging time and charging power are used to create the portrait.

7. The charging demand prediction method based on user profile regional distribution according to claim 1 is characterized in that: The specific method of step three is: (3.1) Based on the user portrait data, conduct a regional distribution analysis of the user portraits, study the charging needs and behavior patterns of users in different regions, and obtain regional characteristic data; at the same time, determine the region to be analyzed; the regional characteristic data includes region, population density, traffic conditions and economic conditions; (3.2) Compare and analyze the distribution of charging demand in different regions, and construct a prediction model for charging demand in a specific region in the future using time series analysis or multivariate linear regression analysis.

8. The charging demand prediction method based on user profile regional distribution according to claim 7 is characterized in that: The ARIMA model was used for time series analysis, and the multiple linear regression model was used for multiple linear regression analysis.

9. The charging demand prediction method based on user profile regional distribution according to claim 8, characterized in that: In the ARIMA model, user profile data is used as input, and the charging demand in the region in the future period is used as output for training; In the multiple linear regression model, charging demand is used as the dependent variable Y, and region, population density, traffic conditions, economic conditions and user portrait data are used as independent variables X for modeling.

Images