Electric vehicle user classification method and system
By obtaining dynamic behavior data of electric vehicle users and using gradient enhancement decision tree algorithm for classification, the problem that static characteristics in the existing technology cannot reflect dynamic changes is solved, and more accurate and efficient user classification is achieved, and the electric vehicle users are improved. Scheduling potential and system operation economy.
Patent Information
- Application Number
- CN202411168052.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-12-03
AI Technical Summary
The existing electric vehicle user classification method is based on static characteristics and cannot accurately reflect the dynamic behavior changes of users, resulting in insufficient processing of large-scale, high-dimensional and dynamically changing data, affecting the utilization rate of charging facilities and the efficiency of energy management.
By obtaining the travel data and charging data of electric vehicle users, extracting charging behavior characteristics, such as charging time, power and preference characteristics, using the gradient improvement decision tree algorithm to classify users, combining real-time data of the Internet of Vehicles platform, carefully considering the dynamic changes of users characteristic.
It significantly improves the accuracy and efficiency of user classification, improves the utilization rate of charging facilities and the economicality of energy management, adapts to the scheduling potential of different electric vehicle users, and improves the economicality of system operation.
Smart Images

Figure CN119066536A_ABST
Abstract
Description
Technical Field
[0001] The present invention mainly relates to the technical field of electric vehicles, and in particular to a method and system for classifying users of electric vehicles. Background Art
[0002] As the global demand for sustainable energy increases, the penetration rate of electric vehicles (EVs) is rising rapidly, which not only promotes the green transformation of the transportation sector, but also brings new challenges to smart grids and charging infrastructure. The charging behavior of electric vehicle users has a direct impact on grid load management, especially during peak hours, when a large number of electric vehicle charging demands may cause grid overload, affecting the stability and economy of the power system.
[0003] The behavioral characteristics of electric vehicle users are an important basis for charging facility layout planning, charging scheduling optimization, and grid load management. Traditional electric vehicle user classification methods are mostly based on static characteristics and cannot accurately reflect the dynamic behavioral changes of users. There are significant individual differences in the travel patterns and charging habits of electric vehicle users, and these behavioral characteristics change with time, location, and other external factors. Therefore, studying the analysis and classification methods of electric vehicle user behavioral characteristics based on dynamic data is of great significance for improving the utilization rate of charging facilities, optimizing energy management, and improving user satisfaction. Most of the existing electric vehicle user classification methods use traditional classification methods, which have many shortcomings when dealing with large-scale, high-dimensional, and dynamically changing data. Summary of the invention
[0004] In view of the technical problems existing in the prior art, the present invention provides an electric vehicle user classification method and system which can improve classification accuracy and efficiency.
[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is:
[0006] A method for classifying electric vehicle users comprises the steps of:
[0007] 1) Obtain travel and charging data of electric vehicle users;
[0008] 2) Obtaining charging behavior characteristics of electric vehicles based on travel data and charging data of electric vehicle users; the charging behavior characteristics include charging time characteristics, charging power characteristics, and charging preference characteristics of electric vehicles;
[0009] 3) Based on the charging behavior characteristics of electric vehicles, the gradient boosting decision tree algorithm is used to classify electric vehicle users.
[0010] Preferably, in step 2), battery parameters and charging duration are used to describe the charging behavior characteristics of the electric vehicle;
[0011] Different types of electric vehicles have different battery capacities. After statistically analyzing the battery capacity data of various types of electric vehicles, the battery capacities of different types of electric vehicles are as follows:
[0012]
[0013] in is the battery capacity of the electric vehicle, p is the type of electric vehicle, α and β are the parameters of the gamma function, μ and σ are the parameters of the normal function;
[0014] The expected power consumption of electric vehicle users is set Following the normal distribution of N(0.85, 0.3), the charging time of electric vehicles is as follows:
[0015]
[0016] Where: is the state of charge at the beginning of charging, η is the charging efficiency, p c is the charging power.
[0017] Preferably, in step 2), for the i-th electric vehicle in a charging station, the charging time characteristics include the arrival time Departure time Network access time Off-grid moments Waiting time Charging time Duration of stay after charging is completed until leaving the charging station and the total length of stay at the charging station Among them, the waiting time Charging time Stop time and total length of stay Obtained by the following formula:
[0018]
[0019] The eight time features are divided into a set of electric vehicle charging time state quantity features and a set of electric vehicle charging time process quantity features, as shown in the following formula:
[0020]
[0021] Where n is the number of all electric vehicles in the study area, τ is the feature set of all electric vehicle charging time state quantities, and T is the feature set of all electric vehicle charging time process quantities.
[0022] Preferably, in step 2), the user power characteristics of the i-th electric vehicle include the power when connected to the power grid Power consumption when off the grid Actual charging capacity ΔE i , Expected power Charging demand And battery capacity
[0023] Among them, the expected power It follows the normal distribution shown in the following formula:
[0024]
[0025] Actual charging capacity ΔE i for:
[0026]
[0027] Charging demand for:
[0028]
[0029] The six user power characteristics are divided into an electric vehicle user power state quantity feature set and an electric vehicle user power process quantity feature set, as shown in the following formula:
[0030]
[0031] Where S is the power state quantity feature set of all electric vehicle users in the study area, and E is the power process quantity feature set of all electric vehicle users in the study area.
[0032] Preferably, in step 2), the vehicle parameter characteristic ω is used i , response willingness σ i and charging cost C i To describe the user's charging preference characteristics.
[0033] Preferably, the vehicle parameter characteristic ω i As shown below:
[0034]
[0035] Where: i is the vehicle parameter characteristic of the ith electric vehicle. When it is 0, it means that the ith electric vehicle has not signed a relevant response agreement with the electric vehicle aggregator and does not respond to the optimal dispatch of the power grid. When ω is 1, it means that the ith electric vehicle has signed a relevant response agreement with the electric vehicle aggregator, and whether it participates in the optimal dispatch of the power grid depends on other characteristic quantities.
[0036] Preferably, the response willingness σ is obtained i The process is:
[0037] Judging the current response willingness σ by historical response behavior i , defining the EV i response rate It is the ratio of the number of response dispatches to the number of responses initiated by the power grid within a period of time:
[0038]
[0039] Where: T i is the response rate of the i-th electric vehicle, N is the number of days, is the number of responses of the i-th electric vehicle on the j-th day, is the number of times the power grid responds to the i-th electric vehicle on the j-th day;
[0040] right and Make the following provisions: N c =[0,1,2,3], That is, the power grid initiates a response to the electric vehicle at least once and at most three times a day, and the number of responses of the electric vehicle is randomly selected between 0 and 3;
[0041]
[0042] Where: T is the average response rate of all EVs in the study area;
[0043] Therefore, the response willingness of electric vehicles in this i As shown below:
[0044]
[0045] When the response rate of the i-th electric vehicle is greater than the average response rate of all electric vehicles, the electric vehicle is willing to participate in this response, otherwise it is unwilling to participate in this response.
[0046] Preferably, the charging cost C is obtained i The specific process is:
[0047] According to the geographical location and charging price information of all charging stations within the research scope, the station matching is performed based on the GPS positioning of the data set of the i-th electric vehicle entering the charging station, and the electricity price information C in each charging data is obtained. a (t), and get the charging cost C of the i-th electric car i as follows:
[0048]
[0049] Where P i (t) is the charging power of the i-th electric vehicle at time t; C a (t) is the real-time electricity price of the charging station.
[0050] Preferably, in step 3), the steps of the gradient boosting decision tree algorithm are:
[0051] S31. Build initial model
[0052] For the training sample set {x i ,y i )}n, i=1, 2, .., n, where x i is the behavior characteristic of the i-th sample, y i is the category label of the i-th sample, the loss function is L[y,F(x)], γ is the decision category in this iteration, and then the weak learner is initialized:
[0053]
[0054] Lt represents the load at time t, F0(x) represents the number of charging vehicles classified at time i, and Pc represents the charging parameters of each vehicle charging pile;
[0055] S32, loop iteration process
[0056] The number of iterations M is the number of weak learners; the model of iteration m is denoted as F m (x), when m = 1, 2, ... M, M is the number of iterations, perform the following operations; for i = 1, 2, ..., n, calculate the negative gradient of the loss function L as:
[0057]
[0058] γ m is the weighting coefficient of the fitting residual of the training decision tree, specifically:
[0059]
[0060] By h m (x) The model obtained after iteration is as follows:
[0061] F m (x) = F m-1 (x)+γ m h m (x)
[0062] After M iterations, the final strong classification model F is obtained M (x).
[0063] The present invention also discloses an electric vehicle user classification system, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, and when the computer program is run by the processor, the steps of the method described above are executed.
[0064] Compared with the prior art, the advantages of the present invention are:
[0065] The present invention is based on the actual travel and charging data of users, which reflects the real behavior of users, making the classification closer to the actual use of users; the charging behavior characteristics include charging time, power characteristics and preference characteristics, which can comprehensively characterize the user's behavior pattern and provide rich information for classification; the present invention applies the gradient boosting decision tree algorithm to the classification of electric vehicle users, combined with the real-time data collected by the Internet of Vehicles platform, which can significantly improve the accuracy and efficiency of classification. The present invention carefully considers the dynamic change characteristics of electric vehicle users, and can adapt to different electric vehicles by extracting different electric vehicle user characteristics, which is of great significance for effectively improving the scheduling potential of electric vehicle users and improving the economy of system operation. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 The figure is a flow chart of an embodiment of the electric vehicle user classification method of the present invention. DETAILED DESCRIPTION
[0067] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.
[0068] like Figure 1 As shown, the electric vehicle user classification method provided by the embodiment of the present invention specifically includes the following steps:
[0069] S1. Collect travel data of electric vehicle users, and clean and analyze the collected data.
[0070] The electric vehicle travel data and charging data of a certain area in November 2023 obtained from a charging station are selected as the electric vehicle travel database and charging database. The number of daily trips of electric vehicles is counted. The standard for judging the number of daily trips is that when the GPS coordinates of the electric vehicle do not change for more than thirty minutes, it is recorded as the end of a trip; the set of trajectory points of the complete itinerary chain is counted. The daily trajectory data points are divided into several sets according to the number of daily trips. There are several complete itinerary chains for the number of daily trips of electric vehicles, corresponding to several sets of trajectory data points. In this way, a set of trajectory points of electric vehicles can be generated, which is sorted by time point and based on the number of trips.
[0071] The process of processing and cleaning the collected electric vehicle travel data is as follows:
[0072] A. Convert the longitude and latitude coordinates of the track points in the travel data. The GPS data format of the electric vehicle data directly obtained uses the Martian coordinate system (GCJ-02), while the geocentric coordinate system (WGS-28) is used when analyzing the electric vehicles in the selected area based on the data-driven method. In order to match the electric vehicle track point data with the WGS-28 coordinates and improve the accuracy of longitude and latitude, the coordinates of the track points should be converted. Take the longitude and latitude coordinates of the kth track point in the GCJ-02 coordinate system (a k ,b k ) as an example, the conversion method of the two coordinate systems is as follows:
[0073]
[0074] Where: σ is the parameter of the normal function, e is 6378245, e′ is 0.06693422, α k ′ and b k ′ are the values of longitude and latitude in the GCJ-02 coordinate system after special algorithm processing, dLon and dLat are the deviation values of longitude and latitude respectively;
[0075] Finally, the longitude and latitude coordinates of the kth trajectory point in the WGS-28 coordinate system are:
[0076]
[0077] Where: x k and k are the longitude and latitude coordinates of the kth trajectory point in the WGS-28 coordinate system.
[0078] B. Delete the travel data trajectory point sets with travel time less than one minute, travel distance less than 500 meters, and less than 10 trajectory data points during the travel process.
[0079] C. Delete the track point data outside the research area. The travel of electric vehicles is extremely random. When the driving range of electric vehicles exceeds the research area of the present invention, the track data of this trip cannot provide a complete data chain for the research of the present invention. Therefore, all track point data of the vehicle during this trip are deleted.
[0080] D. Delete multiple duplicate track point data within a short distance during the trip. When traffic jams and other conditions cause the electric vehicle to be unable to drive normally, the GPS device of the electric vehicle still uploads location information to the terminal, thus forming a large amount of duplicate track point data. In order to reduce the difficulty of data processing, the duplicate data should be deleted.
[0081] S2. Model the charging behavior characteristics of electric vehicles, and infer the time characteristics, power characteristics, and preference characteristics of electric vehicle users based on the SOC data, GPS coordinates, and other information after the electric vehicles arrive at the charging station.
[0082] Specifically, battery parameters and charging time are used to describe the charging behavior characteristics of electric vehicles; the battery parameters and charging time are as follows:
[0083] 1. Battery parameters
[0084] Different types of electric vehicles have different battery capacities. After statistically analyzing the battery capacity data of various types of electric vehicles, the battery capacities of different types of electric vehicles are obtained as shown in formulas (5) and (6):
[0085]
[0086] in is the battery capacity of the electric vehicle, p is the type of electric vehicle, α and β are the parameters of the gamma function, and μ and σ are the parameters of the normal function.
[0087] 2. Charging time
[0088] Due to the influence of the next travel demand, mileage, and charging power of electric vehicles, the charging time of electric vehicles varies. Set the expected power consumption of electric vehicle users. It obeys the normal distribution of N(0.85, 0.3), so the charging time of electric vehicles is shown in formula (7).
[0089]
[0090] Where: refers to the battery capacity of the ith electric vehicle, is the state of charge at the beginning of charging, η is the charging efficiency, p c is the charging power.
[0091] For the i-th electric vehicle in a charging station, the charging time characteristics mainly include the arrival time Departure time Network access time Off-grid moments Waiting time Charging time Duration of stay after charging is completed until leaving the charging station and the total length of stay at the charging station Among them, the waiting time Charging time Stop time and total length of stay It can be calculated by equation (8) to equation (11).
[0092]
[0093] The eight charging time features are divided into a set of electric vehicle charging time state quantity features and a set of electric vehicle charging time process quantity features, as shown in equations (12)-(13).
[0094]
[0095] Where n is the number of all electric vehicles in the study area, τ is the feature set of all electric vehicle charging time state quantities, and T is the feature set of all electric vehicle charging time process quantities.
[0096] The charging characteristics of the i-th electric vehicle include: Power consumption when off the grid Actual charging capacity ΔE i , Expected power Charging demand And battery capacity Among them, the expected power The actual charging capacity ΔE follows the normal distribution shown in formula (14). i and charging demand As shown in formula (15)-formula (16).
[0097]
[0098] The six power characteristics are divided into an electric vehicle user power state quantity feature set and an electric vehicle user power process quantity feature set, as shown in equations (17) and (18).
[0099]
[0100] Where: S is the charging status power feature set of all electric vehicle users in the study area, and E is the charging process power feature set of all electric vehicle users in the study area.
[0101] Specifically, the vehicle parameter ω is used i , response willingness σ i and charging cost C i To describe the user's charging preference characteristics;
[0102] 1. Vehicle parameters
[0103] Vehicle parameter characteristicsω i As shown in formula (19):
[0104]
[0105] Where: i is the vehicle parameter characteristic of the ith electric vehicle. When it is 0, it means that the ith electric vehicle has not signed a relevant response agreement with the electric vehicle aggregator and does not respond to the optimal dispatch of the power grid. When ω is 1, it means that the ith electric vehicle has signed a relevant response agreement with the electric vehicle aggregator, and whether it participates in the optimal dispatch of the power grid depends on other characteristic quantities.
[0106] The parameter feature set of all electric vehicles is ω=(ω 1 ,ω 2 , ..., ω n ).
[0107] 2. Willingness to respond
[0108] Different travel habits of electric vehicle users make the response willingness each time uncertain. The present invention uses a large number of historical response behaviors to determine the response willingness σ i , defining the EV i response rate It is the ratio of the number of response dispatches to the number of responses initiated by the power grid within a period of time.
[0109]
[0110] Where: T i is the response rate of the i-th electric vehicle, N is the number of days, is the number of responses of the i-th electric vehicle on the j-th day, is the number of times the power grid responds to the i-th electric vehicle on the j-th day. and Make the following provisions: N c =[0,1,2,3], That is, the power grid initiates a response to the electric vehicle at least once and at most three times a day, and the number of responses of the electric vehicle is randomly selected between 0 and 3.
[0111]
[0112] Where: T is the average response rate of all electric vehicles in the study area.
[0113] Therefore, the response willingness of electric vehicles in this i As shown in formula (22):
[0114]
[0115] When the response rate of the i-th electric vehicle is greater than the average response rate of all electric vehicles, the electric vehicle is willing to participate in this response, otherwise it is unwilling to participate in this response.
[0116] The response willingness characteristic set of all electric vehicles is σ=(σ 1 , σ 2 , ..., σ n ).
[0117] 3. Charging costs
[0118] When the power grid is at its peak, the power price is higher than that in normal times. When the power grid is at its low point, the power price is lower than that in normal times. The power grid adjusts the demand side response by setting time-of-use electricity prices to avoid the occurrence of power grid peaks. Electric vehicle aggregators also set time-of-use electricity prices to encourage electric vehicles to participate in the optimal scheduling of the power grid. Therefore, the time-of-use electricity price of the charging station set by the aggregator is shown in formula (23).
[0119] C a (t) = C0(1+C k )k=1,2,3 (23)
[0120] Where: C a (t) is the real-time electricity price of the charging station, C0 is the basic electricity price, the unit is RMB / KW·h, k=1,2,3 represents the peak period, normal period and valley period respectively, C k It is the floating ratio compared with the basic electricity price for each period.
[0121] Furthermore, according to the geographical location and charging electricity price information of all charging stations within the research scope, the station matching is performed based on the GPS positioning of the data set of the i-th electric vehicle entering the charging station, and the electricity price information C in each charging data is obtained. a (t), we can get the charging cost C of the i-th electric car i as follows:
[0122]
[0123] Where P i (t) is the charging power of the i-th electric vehicle at time t.
[0124] The characteristic set of all electric vehicle charging costs is C = (C 1 ,C 2 ,…,C n ).
[0125] S3, Gradient Boosting Decision Tree (GBDT) uses classification and regression trees as weak learners to iteratively minimize the loss function. The gradient boosting classification model is composed of a series of weak classification models, and there is a strong connection between them. The steps of the gradient boosting decision tree algorithm are as follows:
[0126] S31. Build initial model
[0127] For the training sample set {x i ,y i )}n, i=1, 2, .., n, where x i is the behavior characteristic of the i-th sample, y i is the category label of the i-th sample, the loss function is L[y,F(x)], γ is the decision category in this iteration, and then the weak learner is initialized:
[0128]
[0129] Lt represents the load at time t, F0(x) represents the number of charging vehicles classified at time i, and Pc represents the charging parameters of the charging pile for each vehicle.
[0130] S32, loop iteration process
[0131] The number of iterations M is the number of weak learners. The model of iteration m is denoted as F m (x). When m = 1, 2, ...M, perform the following operations. For i = 1, 2, ..., n, calculate the negative gradient of the loss function L as:
[0132]
[0133] γ m is the weighting coefficient of the fitting residual of the training decision tree, specifically:
[0134]
[0135] By h m (x) The model obtained after iteration is as follows:
[0136]
[0137] After M iterations, the final strong classification model F is obtained M(x). Python programming language is used for modeling training and prediction. It should be noted that if the given n estimators are too small, it is easy to underfit. Therefore, in this patent, n estimators are set to 10, 30 or 50. The learning rate is called the step size, and its value range is (0, 1]. A smaller step size means that more iterations are required, so the step size is adjusted during the iteration process. The maximum feature is set to "None", which means that all features are considered when dividing. Due to the limited model features and training data, the maximum depth is not limited, and the default value is used to prevent the model from overfitting.
[0138] Based on the above algorithm, the feature data and classification criteria of electric vehicle classification are studied. Having "good" feature data can improve the prediction accuracy of the GBDT algorithm. Based on the electric vehicle behavior characteristics and historical charging data collected from user surveys, an electric vehicle user behavior feature dataset is constructed as training data for the classification model.
[0139] In order to evaluate whether electric vehicles respond to dispatch to shift peaks and fill valleys, the following classification criteria are considered and implemented:
[0140] (1) Electric vehicle users with dispatch willingness and demand response capabilities;
[0141] (2) Electric vehicle users who have the willingness and time margin to respond;
[0142] (3) Electric vehicle users who have the willingness and power margin to respond;
[0143] (4) Electric vehicle users who have no intention of scheduling.
[0144] The present invention firstly collects the travel data and charging data of electric vehicle users in combination with the Internet of Vehicles platform, and cleans and analyzes the collected data to obtain a data set of the number of electric vehicle trips; then, the charging behavior characteristics of electric vehicles are modeled based on data-driven, and the time characteristics, user power characteristics and user preference characteristics of electric vehicles are inferred based on the state of charge SOC data, GPS coordinates and other information after the electric vehicle arrives at the charging station; and then, based on these characteristics, the gradient boosting decision tree algorithm is used to classify electric vehicle users.
[0145] The present invention is based on the actual travel and charging data of users, which reflects the real behavior of users, making the classification closer to the actual use of users; the charging behavior characteristics include charging time, power characteristics and preference characteristics, which can comprehensively characterize the user's behavior pattern and provide rich information for classification; the gradient boosting decision tree algorithm is applied to the classification of electric vehicle users, combined with the real-time data collected by the Internet of Vehicles platform, which can significantly improve the accuracy and efficiency of classification. The present invention carefully considers the dynamic change characteristics of electric vehicle users, and can adapt to different electric vehicles by extracting different electric vehicle user characteristics, which is of great significance for effectively improving the scheduling potential of electric vehicle users and improving the economy of system operation.
[0146] The present invention also discloses an electric vehicle user classification system, comprising a memory and a processor connected to each other, wherein a computer program is stored on the memory, and the computer program executes the steps of the above method when being executed by the processor. The system of the present invention corresponds to the above method and also has the advantages described in the above method. The present invention implements all or part of the processes in the above embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above method embodiment when being executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0147] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.
Claims
1. A method for classifying electric vehicle users, characterized in that: Includes steps: 1) Obtain travel and charging data of electric vehicle users; 2) Obtaining the charging behavior characteristics of electric vehicles based on the travel data and charging data of electric vehicle users; The charging behavior characteristics include electric vehicle charging time characteristics, charging power characteristics, and charging preference characteristics; 3) Based on the charging behavior characteristics of electric vehicles, the gradient boosting decision tree algorithm is used to classify electric vehicle users.
2. The electric vehicle user classification method according to claim 1, characterized in that: In step 2), the battery parameters and charging time are used to describe the charging behavior characteristics of the electric vehicle; Different types of electric vehicles have different battery capacities. After statistically analyzing the battery capacity data of various types of electric vehicles, the battery capacities of different types of electric vehicles are as follows: in is the battery capacity of the electric vehicle, p is the type of electric vehicle, α and β are the parameters of the gamma function, μ and σ are the parameters of the normal function; The expected power consumption of electric vehicle users is set Following the normal distribution of N(0.85, 0.3), the charging time of electric vehicles is as follows: Where: is the state of charge at the beginning of charging, η is the charging efficiency, p c is the charging power.
3. The electric vehicle user classification method according to claim 1, characterized in that: In step 2), for the i-th electric vehicle in a charging station, the charging time characteristics include the arrival time Departure time Network access time Off-grid moments Waiting time Charging time Duration of stay after charging is completed until leaving the charging station and the total length of stay at the charging station Among them, the waiting time Charging time Stop time and total length of stay Obtained by the following formula: The eight time features are divided into a set of electric vehicle charging time state quantity features and a set of electric vehicle charging time process quantity features, as shown in the following formula: Where n is the number of all electric vehicles in the study area, τ is the feature set of all electric vehicle charging time state quantities, and T is the feature set of all electric vehicle charging time process quantities.
4. The electric vehicle user classification method according to claim 1, 2 or 3, characterized in that: In step 2), the user power characteristics of the i-th electric vehicle include the power consumption when connected to the grid Power consumption when off the grid Actual charging capacity ΔE i , Expected power Charging demand And battery capacity Among them, the expected power It follows the normal distribution shown in the following formula: Actual charging capacity ΔE i for: Charging demand for: The six user power characteristics are divided into an electric vehicle user power state quantity feature set and an electric vehicle user power process quantity feature set, as shown in the following formula: Where S is the power state quantity feature set of all electric vehicle users in the study area, and E is the power process quantity feature set of all electric vehicle users in the study area.
5. The electric vehicle user classification method according to claim 1, 2 or 3, characterized in that: In step 2), the vehicle parameter feature ω is used i , response willingness σ i and charging cost C i To describe the user's charging preference characteristics.
6. The electric vehicle user classification method according to claim 5, characterized in that: Vehicle parameter characteristicsω i As shown below: Where: i is the vehicle parameter characteristic of the ith electric vehicle. When it is 0, it means that the ith electric vehicle has not signed a relevant response agreement with the electric vehicle aggregator and does not respond to the optimal dispatch of the power grid. When ω is 1, it means that the ith electric vehicle has signed a relevant response agreement with the electric vehicle aggregator, and whether it participates in the optimal dispatch of the power grid depends on other characteristic quantities.
7. The electric vehicle user classification method according to claim 5, characterized in that: Get response willingness σ i The process is: Judging the current response willingness σ by historical response behavior i , defining the EV i response rate It is the ratio of the number of response dispatches to the number of responses initiated by the power grid within a period of time: Where: is the response rate of the i-th electric vehicle, N is the number of days, is the number of responses of the i-th electric vehicle on the j-th day, is the number of times the power grid responds to the i-th electric vehicle on the j-th day; right and Make the following provisions: N c =[0,1,2,3], That is, the power grid initiates a response to the electric vehicle at least once and at most three times a day, and the number of responses of the electric vehicle is randomly selected between 0 and 3; Where: T is the average response rate of all EVs in the study area; Therefore, the response willingness of electric vehicles in this i As shown below: When the response rate of the i-th electric vehicle is greater than the average response rate of all electric vehicles, the electric vehicle is willing to participate in this response, otherwise it is unwilling to participate in this response.
8. The electric vehicle user classification method according to claim 5, characterized in that: Get charging fee C i The specific process is: According to the geographical location and charging price information of all charging stations within the research scope, the station matching is performed based on the GPS positioning of the data set of the i-th electric vehicle entering the charging station, and the electricity price information C in each charging data is obtained. a (t), and get the charging cost C of the i-th electric car i as follows: Where P i (t) is the charging power of the i-th electric vehicle at time t; C a (t) is the real-time electricity price of the charging station.
9. The electric vehicle user classification method according to claim 1, 2 or 3, characterized in that: In step 3), the steps of the gradient boosting decision tree algorithm are: S31. Build initial model For the training sample set {x i ,y i )}n, i=1, 2, .., n, where x i is the behavior characteristic of the i-th sample, y i is the category label of the i-th sample, the loss function is L[y,F(x)], γ is the decision category in this iteration, and then the weak learner is initialized: Lt represents the load at time t, F0(x) represents the number of charging vehicles classified at time i, and Pc represents the charging parameters of each vehicle charging pile; S32, loop iteration process The number of iterations M is the number of weak learners; the model of iteration m is denoted as F m (x), when m = 1, 2, ... M, M is the number of iterations, perform the following operations; for i = 1, 2, ..., n, calculate the negative gradient of the loss function L as: γ m is the weighting coefficient of the fitting residual of the training decision tree, specifically: By h m (x) The model obtained after iteration is as follows: F m (x)=F m-1 (x)+γ m h m (x) After M iterations, the final strong classification model F is obtained M (x).
10. An electric vehicle user classification system, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 9.
