Electric vehicle user portraying method, product, medium and portraying system

Abstract

The invention discloses an electric vehicle user portraying method, a product, a medium and a portraying system. The method comprises the following steps: S1, acquiring a charging behavior factor of an electric vehicle user; s2, aggregating the data by adopting an improved K-means clustering method, constructing an index system of an electric vehicle user portrait, and forming an index system library; in the improved K-means clustering method, aggregation centers with farther relative distances are obtained by setting different positions as the probabilities of the aggregation centers; and S3, acquiring an electrical load time sequence curve of the electric vehicle user, and comparing and fitting the electrical load time sequence curve with the electric vehicle user portrait in the index system library to obtain a final electric vehicle user portrait. The method can accurately capture the dynamic change of the user behavior of the electric vehicle, and has important significance for improving the electric energy quality of the system, reducing the network loss and improving the economical efficiency of system operation.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of electric vehicles, and in particular to an electric vehicle user profiling method, product, medium and profiling system. Background Art

[0002] With the improvement of global awareness of environmental protection and the use of renewable energy, electric vehicles (EVs), as an important part of green transportation, are experiencing rapid popularization and market penetration. However, the large-scale access of electric vehicles has brought unprecedented challenges to the power system, especially the randomness and uncertainty of charging load, which not only affects the stability and power quality of the power grid, but also increases the cost of power grid operation. In order to effectively manage the charging demand of electric vehicles, optimize the allocation of power resources, and improve the economic and environmental performance of the power grid, it is crucial to conduct in-depth analysis and accurate prediction of the charging behavior of electric vehicle users.

[0003] Traditional power user portraits mainly focus on static power consumption characteristics, such as power consumption and power consumption time, while ignoring dynamic user behavior patterns and personalized needs, especially in the field of electric vehicles. The charging behavior of electric vehicle users is affected by many factors, including but not limited to charging power, charging time preference, charging location selection, battery capacity, remaining state of charge (SOC), and personal travel needs. The complex interaction of these variables makes it difficult for traditional user portrait methods to fully reflect the real needs and behavior patterns of electric vehicle users.

[0004] In recent years, machine learning algorithms have made significant progress in the field of data analysis. Among them, the K-means clustering algorithm has been widely used in user grouping research due to its simplicity and efficiency. However, the standard K-means clustering algorithm has limitations when dealing with non-spherical distributions and data sets of different densities, and the selection of the initial cluster center has a great influence on the final result and is prone to fall into a local optimal solution. Summary of the invention

[0005] The technical problem to be solved by the present invention is: in response to the technical problems existing in the prior art, the present invention provides an electric vehicle user portrait method, product, medium and portrait system that can accurately capture the dynamic changes in electric vehicle user behavior.

[0006] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0007] A method for profiling electric vehicle users, comprising the steps of:

[0008] S1. Obtain charging behavior factors of electric vehicle users; the charging behavior factors include one or more of charging power, charging time, charging location, battery capacity and remaining state of charge;

[0009] S2. Use the improved K-means clustering method to aggregate the data, build an indicator system for electric vehicle user portraits, and form an indicator system library; in the improved K-means clustering method, the probability of setting different positions as aggregation centers is used to obtain aggregation centers with a relatively farther distance;

[0010] S3. Obtain the electric vehicle user's electricity load time series curve, compare and fit it with the electric vehicle user portrait in the indicator system library, and obtain the final electric vehicle user portrait.

[0011] Preferably, in step S2, the process of aggregating data using the improved K-means clustering method is:

[0012] S21. Determine the number of clusters K: According to each required data type, the number of clusters K is determined by using the elbow method; the core indicator of the elbow method is the sum of squared errors, that is:

[0013]

[0014] Where: n SSE is the sum of squared errors; p is all the points in each cluster; m i represents the aggregation center of each class;

[0015] S22, select the initial aggregation center: obtain an aggregation center with a relatively farther distance by setting the probability of different positions as the aggregation center; wherein N electric vehicle user feature vectors are used as samples, namely:

[0016]

[0017] Where: Represented as electric vehicle user feature vector;

[0018] The feature vector is composed of different aggregated features of electric vehicle users:

[0019]

[0020] Where: V represents the aggregated characteristic quantity of electric vehicle users;

[0021] S23. Calculate the Euclidean distance from each feature vector to the initial aggregation center, that is:

[0022]

[0023] Where: D represents the Euclidean distance from the feature vector to the initial aggregation center;

[0024] For each feature vector, assign it to the cluster corresponding to the aggregation center with the smallest Euclidean distance; in each cluster, update the aggregation center:

[0025]

[0026] S24. Repeat steps S21-S23 until the average error criterion function converges and clustering is completed.

[0027] Preferably, step S22 is specifically:

[0028] S221, randomly select a point as the initial aggregation center;

[0029] S222. For each point in the sample, calculate its Euclidean distance D to the nearest initial aggregation center:

[0030]

[0031] S223, increase the probability of the farthest point being the next aggregation center;

[0032] S224. Repeat S222-S223 until K initial aggregation centers are selected.

[0033] Preferably, the indicator system includes charging power, charging time, battery capacity and remaining state of charge;

[0034] According to the power size, the charging power is divided into 4 intervals, namely super charging, fast charging, normal charging, and slow charging;

[0035] According to the length of time, the charging time is divided into 4 intervals, namely super charging, fast charging, normal charging and slow charging;

[0036] According to the size of the battery capacity, the battery capacity is divided into 4 intervals, namely super large capacity, large capacity, normal capacity and small capacity;

[0037] The remaining state of charge (SOC) is divided into four intervals: sufficient charge, relatively sufficient charge, normal charge and insufficient charge.

[0038] Preferably, the specific process of step S3 is:

[0039] S31. Construct Gaussian mixture mathematical expressions of different indicators:

[0040] Let Ω={α k ,μ k ,σ k ; k = 1, 2...K} is the adjustable parameter set of GMM; the mathematical expression of GMM is as follows:

[0041]

[0042] Among them, f x (x) represents the joint probability density function of X, α k Represents the weight coefficient, N k (*) represents the multidimensional normal distribution, which is called the kth Gaussian component of GMM, K represents the total number of Gaussian components, D represents the dimension of X, det(*) represents the matrix determinant, μ and σ represent the mean vector and covariance matrix of the Gaussian distribution respectively;

[0043] The mean of a random vector X is:

[0044]

[0045] Where: E[X] is the mean of the random vector X, α k represents the weight coefficient, μ represents the mean vector of Gaussian distribution;

[0046] S32, input clustering result data x = (x1, x2, ... xk), joint distribution p (x, z; Ω), conditional distribution p (z|x, Ω), maximum number of iterations J;

[0047] S33, using the improved K-means clustering algorithm to determine the initial value Ω0 of the parameter set Ω, and respectively calculating the proportion of the number of data in each cluster data set to the total number, the mean and the variance, as the initial parameter set Ω0 of the model;

[0048] After obtaining the initial parameter set Ω0, the EM algorithm iteration begins, with the number of iterations j = 1, 2, .J;

[0049] S34, expectation step: calculate the conditional probability expectation of the joint distribution:

[0050] Q i (z i )=p(z i |x i ,Ω j );

[0051]

[0052] S35, maximum step: maximize l(Ω,Ω j ), we get Ω j+1 :

[0053] Ω j+1 =argmaxl(Ω,Ω j );

[0054] S36. If Ω+1 has converged, the algorithm ends and outputs the parameter set Ω; otherwise, continue to iterate through steps S34-S35;

[0055] S37, then respectively calculate the average values ​​of different electric vehicle users at 96 points in 24 hours every day to form a load matrix det(Am) representing a certain indicator of such users;

[0056] S38. According to the 365*24 load matrix det(Am) of each type of user, a probability density function is fitted for the power load at 24 moments, and the mean μ and standard deviation σ are determined by using the EM algorithm, k=1, 2, ..., 96, and the function is plotted as a time series curve;

[0057] S39. Compare the time series curves of electric vehicle users under different characteristics, compare and fit them with the indicator system of electric vehicle user portraits, and obtain the corresponding electric vehicle user portraits.

[0058] Preferably, in step S33, the specific calculation formulas for the proportion, mean and variance are:

[0059]

[0060] where u n represents the number of data belonging to the nth class, U represents the number of all data, x n Represents all data values ​​belonging to the nth class. After obtaining the initial parameter set Ω0, the EM algorithm iteration begins, with the number of iterations j=1,2,.J.

[0061] Preferably, in step S37, the calculation formula of the load matrix det(Am) is:

[0062]

[0063] The present invention also discloses a computer program product, comprising a computer program, wherein the computer program executes the steps of the method described above when executed by a processor.

[0064] The present invention further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0065] The present invention also discloses an electric vehicle user portrait system, comprising a memory and a processor connected to each other, wherein a computer program is stored on the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0066] Compared with the prior art, the advantages of the present invention are:

[0067] The present invention samples different data of electric vehicle users, such as charging power, charging time, charging location, battery capacity, remaining state of charge (SOC), travel demand, etc., cleans the data, and then uses an improved K-means clustering method to aggregate the data, constructs an index system for electric vehicle user portraits, and forms an index system library; finally, an improved Gaussian mixture model is used to fit the electric vehicle user portrait to achieve user portraits. The electric vehicle user portrait method based on the improved K-means clustering and Gaussian mixture model of the present invention aims to overcome the limitations of the existing technology (traditional single power user portrait), and accurately captures the dynamic changes of electric vehicle user behavior by mining the charging behavior rules of a large number of electric vehicle users, which is of great significance for improving the power quality of the system, reducing network losses, and improving the economy of system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The present invention is a flowchart of an electric vehicle user profiling method in an embodiment. DETAILED DESCRIPTION

[0069] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0070] like Figure 1 As shown, the electric vehicle user profiling method according to an embodiment of the present invention comprises the following steps:

[0071] S1. Sampling different data of electric vehicle users, such as charging power, charging time, charging location, battery capacity, remaining state of charge (SOC), travel demand, etc.; such data sampling is conducted every 15 minutes; travel demand is reported by electric vehicle users through mobile phone APP or mini program;

[0072] S2. Use the improved K-means clustering method to aggregate the data, build an indicator system for electric vehicle user portraits, and form an indicator system library. Among them, the improved K-means clustering method used is:

[0073] S21. Determine the number of clusters K. According to each required data type, the number of clusters is determined by the elbow method. The elbow method is a simple and effective method for determining the number of clusters K. The core indicator is the error square sum n SSE ,Right now:

[0074]

[0075] Where: n SSE is the sum of squared errors; p is all the points in each cluster; m iRepresents the aggregation center of each class. As K increases, the aggregation degree of each aggregation cluster will gradually increase, and the error sum of squares will gradually decrease; when K reaches the actual number of clusters, the aggregation degree return decreases sharply, and the corresponding maximum value of SSE decreases, which is similar to the elbow of the SSE and K relationship graph. The corresponding number can be determined as the size of the cluster number K.

[0076] S22, select the initial cluster center. The improved K-means clustering algorithm obtains the cluster center with a relatively farther distance by setting the probability of different positions as the cluster center, so as to achieve a better clustering effect. The present invention uses N electric vehicle user feature data vectors as samples, namely:

[0077]

[0078] Where: Represented as electric vehicle user feature vector;

[0079] The feature data vector is composed of different aggregated feature quantities of electric vehicle users:

[0080]

[0081] Where: V represents the aggregated characteristic quantity of electric vehicle users;

[0082] The specific steps include:

[0083] S221, randomly select a point as the initial aggregation center;

[0084] S222. For each point in the sample, calculate the Euclidean distance between it and the nearest initial aggregation center;

[0085]

[0086] Where: D represents the feature vector To the initial aggregation center The Euclidean distance of

[0087] S223, increase the probability of the farthest point being the next aggregation center;

[0088] S224. Repeat steps S222-S223 until K initial aggregation centers are selected.

[0089] S23, standard K-means algorithm. Calculate the Euclidean distance from each feature vector to the aggregation center, that is:

[0090]

[0091] Where: D represents the feature vector To the initial aggregation center The Euclidean distance of

[0092] For each feature vector, assign it to the cluster corresponding to the aggregation center with the smallest Euclidean distance. In each cluster, update the aggregation center:

[0093]

[0094] Repeat the above steps until the average error criterion function converges and clustering is completed.

[0095] Among them, according to the clustering results, an indicator system for electric vehicle user portraits is formed, specifically:

[0096]

[0097]

[0098] S3. Use the improved Gaussian mixture model to fit the portrait of electric vehicle users and realize user portrait. The specific process of the improved Gaussian mixture model is as follows:

[0099] S31. Based on the clustering results and indicator system formed in the above steps, construct Gaussian mixture mathematical expressions of different indicators:

[0100] Let Ω={α k ,μ k ,σ k ; k = 1, 2...K} is the set of adjustable parameters of the Gaussian mixture model (denoted as GMM). The mathematical expression of GMM is as follows:

[0101]

[0102] Among them, f x (x) represents the joint probability density function of X, α k Represents the weight coefficient, N k (*) represents a multidimensional normal distribution, which is called the kth Gaussian component of GMM, K represents the total number of Gaussian components, D represents the dimension of X, det(*) represents the matrix determinant, μ and σ represent the mean vector and covariance matrix of the Gaussian distribution, respectively.

[0103] When the joint probability density function of the random vector X is represented by equations (7), (8), and (9), the mean of X is as shown in equation (10):

[0104]

[0105] Where: E[X] is the mean of the random vector X, α k represents the weight coefficient, and μ represents the mean vector of the Gaussian distribution.

[0106] S32. Input clustering result data x = (x1, x2, ... xk), joint distribution p(x, z; Ω), conditional distribution p(z|x, Ω), and maximum number of iterations J.

[0107] S33. Determine the initial value Ω0 of the parameter set Ω using the improved K-means clustering algorithm.

[0108] The proportion, mean and variance of the number of data in each cluster data set are calculated respectively as the initial parameter set Ω0 of the model. The parameter values ​​obtained in this way can reflect the prior information of the observed data to a large extent, so it can effectively reduce the number of iterations and improve the training efficiency of GMM. The specific calculation formula is:

[0109]

[0110]

[0111] where u n represents the number of data belonging to the nth class, U represents the number of all data, x n Represents all data values ​​belonging to the nth class. After obtaining the initial parameter set Ω0, the EM algorithm iteration begins, with the number of iterations j = 1, 2, .J.

[0112] S34, Expectation step (E-step): Calculate the conditional probability expectation of the joint distribution:

[0113] Q i (z i )=p(z i |x i ,Ω j ) (14)

[0114]

[0115] S35, M-step: Maximize l(Ω,Ω j ), we get Ω j+1 :

[0116] Ω j+1 =argmaxl(Ω,Ω j ) (16)

[0117] S36, if Ω j+1 If convergence has been achieved, the algorithm ends and the parameter set Ω is output; otherwise, the E-step and M-step are continued to iterate.

[0118] S37, then respectively calculate the average of 96 points of different electric vehicle users at 24 times a day to form a matrix det representing a certain indicator of such users, and the calculation formula is:

[0119]

[0120] S38. According to the 365*24 load matrix det(Am) of each type of user, refer to GMM, fit the probability density function of the power load at 24 moments, use the EM algorithm to determine the mean μ and standard deviation σ, k=1,2,...,96, and draw it into a time series curve.

[0121] S39. Compare the time series curves of electric vehicle users under different characteristics, compare and fit them with the indicator system of electric vehicle user portraits, and obtain different electric vehicle user portraits.

[0122] The electric vehicle user profiling method based on improved K-means clustering and Gaussian mixture model aims to overcome the limitations of the existing technology and accurately capture the dynamic changes of electric vehicle user behavior by mining the charging behavior patterns of electric vehicle users. It is of great significance to improve the power quality of the system, reduce network losses, and improve the economy of system operation.

[0123] The present invention samples different data of electric vehicle users, such as charging power, charging time, charging location, battery capacity, remaining state of charge (SOC), travel demand, etc., cleans the data, and then uses an improved K-means clustering method to aggregate the data, constructs an index system for electric vehicle user portraits, and forms an index system library; finally, an improved Gaussian mixture model is used to fit the portraits of electric vehicle users to achieve user portraits. Different from the traditional single power user portrait, the present invention can accurately capture the dynamic changes in the behavior of electric vehicle users by mining the charging behavior laws of a large number of electric vehicle users, which is of great significance for improving the power quality of the system, reducing network losses, and improving the economy of system operation.

[0124] The present invention also discloses a computer program product, including a computer program, which executes the steps of the above method when executed by a processor. The present invention further discloses a computer-readable storage medium, on which a computer program is stored, which executes the steps of the above method when executed by a processor. The present invention also discloses an electric vehicle user portrait system, including an interconnected memory and a processor, on which a computer program is stored, which executes the steps of the above method when executed by a processor. The product, medium and system of the present invention correspond to the above method and also have the advantages described in the above method.

[0125] The present invention implements all or part of the processes in the above-mentioned 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-mentioned method embodiment when 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 that can carry computer program code, recording medium, U disk, mobile hard disk, 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. 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.

[0126] The above are only preferred embodiments of the present invention, and 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 profiling electric vehicle users, characterized in that: Includes steps: S1. Obtain charging behavior factors of electric vehicle users; the charging behavior factors include one or more of charging power, charging time, charging location, battery capacity and remaining state of charge; S2. Use the improved K-means clustering method to aggregate the data, build an indicator system for electric vehicle user portraits, and form an indicator system library; in the improved K-means clustering method, the probability of setting different positions as aggregation centers is used to obtain aggregation centers with a relatively farther distance; S3. Obtain the electric vehicle user's electricity load time series curve, compare and fit it with the electric vehicle user portrait in the indicator system library, and obtain the final electric vehicle user portrait.

2. The electric vehicle user profiling method according to claim 1, characterized in that: In step S2, the process of aggregating data using the improved K-means clustering method is as follows: S21. Determine the number of clusters K: According to each required data type, the number of clusters K is determined by using the elbow method; the core indicator of the elbow method is the sum of squared errors, that is: Where: n SSE is the sum of squared errors; p is all the points in each cluster; m i represents the aggregation center of each class; S22, select the initial aggregation center: obtain an aggregation center with a relatively farther distance by setting the probability of different positions as the aggregation center; wherein N electric vehicle user feature vectors are used as samples, namely: Where: Represented as electric vehicle user feature vector; The feature vector is composed of different aggregated features of electric vehicle users: Where: V represents the aggregated characteristic quantity of electric vehicle users; S23. Calculate the Euclidean distance from each feature vector to the initial aggregation center, that is: Where: D represents the feature vector To the initial aggregation center The Euclidean distance of For each feature vector, assign it to the cluster corresponding to the aggregation center with the smallest Euclidean distance; in each cluster, update the aggregation center: Where: D represents the updated aggregation center S24. Repeat steps S21-S23 until the average error criterion function converges and clustering is completed.

3. The electric vehicle user portrait method according to claim 2, characterized in that: Step S22 is specifically as follows: S221, randomly select a point as the initial aggregation center; S222. For each point in the sample, calculate its Euclidean distance D to the nearest initial aggregation center: S223, increase the probability of the farthest point being the next aggregation center; S224. Repeat S222-S223 until K initial aggregation centers are selected.

4. The electric vehicle user profiling method according to claim 3, characterized in that: The indicator system includes charging power, charging time, battery capacity and remaining state of charge; According to the power size, the charging power is divided into 4 intervals, namely super charging, fast charging, normal charging, and slow charging; According to the length of time, the charging time is divided into 4 intervals, namely super charging, fast charging, normal charging and slow charging; According to the size of the battery capacity, the battery capacity is divided into 4 intervals, namely super large capacity, large capacity, normal capacity and small capacity; The remaining state of charge (SOC) is divided into four intervals: sufficient charge, relatively sufficient charge, normal charge and insufficient charge.

5. The electric vehicle user profiling method according to any one of claims 1 to 4, characterized in that: The specific process of step S3 is: S31. Construct Gaussian mixture mathematical expressions of different indicators: Let Ω={α k ,μ k ,σ k ; k = 1, 2...K} is the adjustable parameter set of Gaussian model GMM; the mathematical expression of GMM is as follows: Among them, f x (x) represents the joint probability density function of X, α k Represents the weight coefficient, N k (*) represents the multidimensional normal distribution, which is called the kth Gaussian component of GMM, K represents the total number of Gaussian components, D represents the dimension of X, det(*) represents the matrix determinant, μ and σ represent the mean vector and covariance matrix of the Gaussian distribution respectively; The mean of a random vector X is: Where: E[X] is the mean of the random vector X, α k represents the weight coefficient, μ represents the mean vector of Gaussian distribution; S32, input clustering result data x = (x1, x2, ... xk), joint distribution p (x, z; Ω), conditional distribution p (z|x, Ω), maximum number of iterations J; S33, using the improved K-means clustering algorithm to determine the initial value Ω0 of the parameter set Ω, and respectively calculating the proportion of the number of data in each cluster data set to the total number, the mean and the variance, as the initial parameter set Ω0 of the model; After obtaining the initial parameter set Ω0, the EM algorithm iteration begins, with the number of iterations j = 1, 2, .J; S34, expectation step: calculate the conditional probability expectation of the joint distribution: Q i (z i )=p(z i |x i ,Ω j ); S35, maximum step: maximize l(Ω,Ω j ), we get Ω j+1 : Oh j+1 =argmaxl(Ω,Ω j ); S36, if Ω j+1 If convergence has been achieved, the algorithm ends and the parameter set Ω is output; otherwise, continue to iterate through steps S34-S35; S37, then respectively calculate the average values ​​of different electric vehicle users at 96 points in 24 hours every day to form a load matrix det(Am) representing a certain indicator of such users; S38. According to the 365*24 load matrix det(Am) of each type of user, a probability density function is fitted for the power load at 24 moments, and the mean μ and standard deviation σ are determined by using the EM algorithm, k=1, 2, ..., 96, and the function is plotted as a time series curve; S39. Compare the time series curves of electric vehicle users under different characteristics, compare and fit them with the indicator system of electric vehicle user portraits, and obtain the corresponding electric vehicle user portraits.

6. The electric vehicle user profiling method according to claim 5, characterized in that: In step S33, the specific calculation formulas for proportion, mean and variance are: where u n represents the number of data belonging to the nth class, U represents the number of all data, x n Represents all data values ​​belonging to the nth class. After obtaining the initial parameter set Ω0, the EM algorithm iteration begins, with the number of iterations j=1,2,.J.

7. The electric vehicle user portrait method according to claim 5, characterized in that: In step S37, the calculation formula of the load matrix det(Am) is:

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium having a computer program stored thereon, 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 7.

10. An electric vehicle user portrait 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 7.

Images