Clustering and grouping method and system for electric vehicle charging stations based on operation and maintenance consideration

Abstract

An electric vehicle charging station clustering and grouping method based on operation and maintenance consideration comprises the steps that S1, analysis processing is carried out based on historical charging station operation information, and an extracted reference feature sequence is obtained by adopting a time sequence feature extraction method of correlation analysis; s2, performing statistical analysis on the reference feature sequence, and extracting a statistical feature sequence; s3, calculating the weight of the operation information of each dimension by adopting a Sierphi method, and respectively calculating the weights of different statistical characteristic values of each dimension by adopting an entropy weight method to obtain a weight distribution set; s4, calculating a multi-dimensional weighted distance from each charging station to a clustering center based on the weight distribution set and the statistical feature sequence, and obtaining a clustering result of all the charging stations based on a K-means multi-dimensional time sequence clustering algorithm; and S5, according to the clustering result of the charging station, designing an operation and maintenance plan for each type in the clustering result. The design dynamically reflects the interaction relationship between the charging stations, is more flexible and has robustness.

Description

Technical Field

[0001] The present invention relates to a clustering grouping method and system for electric vehicle charging stations based on operation and maintenance considerations, and is specifically suitable for improving the efficiency of operation and maintenance planning. Background Art

[0002] In recent years, the global demand for electric vehicles (EVs) has risen significantly, becoming a key driver of the transformation of the transportation industry. Continuous breakthroughs in battery technology have greatly improved the range of electric vehicles and gradually shortened charging time, which has further promoted market acceptance. However, the rapid popularization of electric vehicles has also brought about a surge in charging demand. The surge in charging demand is not only reflected in the popularity of home charging piles, but also in the expansion of public charging networks. In order to meet the growing demand for charging, various regions have invested in the construction of charging stations and promoted the research and development of fast charging technology to improve charging efficiency and convenience. In this context, how to effectively meet charging needs will be the key to promoting the sustainable development of the electric vehicle industry.

[0003] By clustering the charging stations with similar operating status, we can dig out the internal rules of the actual charging status of the charging stations. Charging stations with similar charging behavior characteristics have mutually exclusive demands for power resources, and the clustering results often help guide the allocation of power resources in the power grid system. On the other hand, when one of the charging stations has an imbalance in supply and demand or a fault problem, clustering can quickly find the charging station with outliers, which helps to timely reallocate power resources and dynamically meet the charging needs of electric vehicles. However, most traditional charging station grouping clustering methods are based on the geographical location and type of charging stations, and cannot reflect the grouping information on the actual operating status of charging stations. Based on the historical operating information of charging stations, data mining and clustering methods can help improve the efficiency of power system operation and maintenance. Summary of the invention

[0004] The purpose of the present invention is to fill the technical gap of clustering operation and maintenance in the prior art, and to provide a clustering grouping method and system for electric vehicle charging stations based on operation and maintenance considerations.

[0005] To achieve the above objectives, the technical solution of the present invention is:

[0006] In a first aspect, the present invention provides a clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations, comprising the following steps:

[0007] S1. Analyze and process the historical operation information of charging stations, and use the time series feature extraction method of correlation analysis to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence;

[0008] S2, performing statistical analysis on the benchmark feature sequence and extracting statistical features as a statistical feature sequence;

[0009] S3. The Shelfie method is used to calculate the weight of the operation information of each dimension, and the entropy weight method is used to calculate the weight of the statistical features respectively to obtain a weight distribution set;

[0010] S4. Based on the weight distribution set and the statistical feature sequence, the multidimensional weighted distance from each charging station to the cluster center is calculated, and the clustering results of all charging stations are obtained based on the multidimensional time series clustering algorithm of K-means;

[0011] S5. According to the clustering results of the charging stations, an operation and maintenance plan is designed for each type in the clustering results.

[0012] In S1, within the scope of the urban system power system, there are I charging stations C at different geographical locations, each charging station includes K dimensions of time series operation information, and the collection of original data of each dimension is called the original sequence;

[0013] The time series feature extraction method of correlation analysis is used to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and the extracted benchmark feature sequence is obtained:

[0014] Through discrete Fourier transform, the original subsequences x(t) of each dimension of the original sequence are transformed from the time domain to the frequency domain to obtain discrete subsequences of each dimension in the frequency domain;

[0015] For each dimension, the absolute value of each group of data in the obtained frequency domain discrete subsequence is calculated according to the formula |A(r)*cos(2π*f(r)+P(r)*π)|, where the frequency vector of the rth group of data is f(r), the amplitude vector is A(k), and the phase vector is P(r). The absolute value calculation results are sorted from large to small, and the first m groups of data are selected to synthesize a new feature subsequence:

[0016]

[0017] Where: y(t) is the new feature subsequence; calculate the Pearson coefficient between the new feature subsequence and the original subsequence:

[0018]

[0019] Among them, x(t) is the original subsequence, is the mean of the original subsequence, is the mean of the new feature subsequence, T is the total number of time points;

[0020] If the Pearson coefficient value is less than the threshold δ, the value of m is increased to resynthesize a new feature subsequence and calculate the Pearson coefficient;

[0021] If the Pearson coefficient value is greater than or equal to the threshold δ, the new feature subsequence y(t) of this dimension can be used as the k-th dimension operation information sequence of the i-th charging station Add the benchmark feature sequence of the i-th charging station

[0022] When each charging station has completely collected the new feature subsequences of each dimension, the dimensionality reduction calculation is completed and the complete benchmark feature sequence F is obtained. base .

[0023] In S2, for each charging station, the baseline feature sequence Perform statistical analysis on the running information of each dimension and obtain the statistical feature subsequence:

[0024]

[0025] in, represents the k-th dimension operation information of the i-th charging station in the benchmark feature sequence, The time traversal characteristics of the k-th dimension of the running information for the i-th charging station, is the discrete fluctuation characteristic of the k-th dimension operation information of the i-th charging station, is the periodic feature of the k-th dimension operation information of the i-th charging station, Run the information peak feature for the kth dimension of the i-th charging station;

[0026] Using the calculated statistical characteristics of subsequences Synthetic statistical feature sequence F stas .

[0027] Extract the statistical eigenvalues ​​of the benchmark feature sequence and extract the time traversal features:

[0028]

[0029] Where: T is the total number of time points, f i k (t) is The value at time t;

[0030] Extract discrete fluctuation features:

[0031]

[0032] in: for The mean of

[0033] Extract periodic features:

[0034]

[0035] Where: n is the total number of days, d is the dth day, For The time series data within the d-th day intercepted in ; R() is the Pearson correlation coefficient between the d-th day and the d+1-th day;

[0036] Extract peak features:

[0037]

[0038] in: is the time when the peak value of the k-th dimension operation information of the ith charging station occurs on the d-th day.

[0039] In S3, the Shelfie method is used to calculate the weight of the operation information of each dimension: a number of power professionals are invited to score the importance of the operation information of each dimension to the charging station in the form of a questionnaire, and the weight value of each dimension {e1, e2, …, e k ,…,e K};

[0040] The entropy weight method is used to calculate the weights of statistical features respectively:

[0041] make For the k-th dimension operation information, the statistical feature sequence F of I charging stations stas The matrix is: The a-th statistical characteristic value at the i-th charging station is c ia ; The entropy weight method is used to calculate the weight of the statistical eigenvalue of the kth dimension, and the weight vector of the statistical eigenvalue is {w k1 ,w k2 ,…,w k4};

[0042] The weight of the operating information of each dimension is combined with the weight vector of the statistical eigenvalue to obtain a weight distribution set.

[0043] In S4, based on the weight distribution set and the statistical feature sequence F stas , calculate the multidimensional weighted distance from each charging station to the cluster center, charging station C i To the jth cluster center M j The distance Dist(C i ,M j )for:

[0044]

[0045] Among them, the cluster center M j ={H j,1 ,H j,2 ,…,H j,k ,…,H j,K}; H j,k is the running information value of the jth cluster center in the kth dimension, H j,k ={h jk1 ,h jk2 ,h jk3 ,h jk4};

[0046] Indicates the operation information dimension H from the i-th charging station to the j-th cluster center in the k-th j,k The distance is calculated as follows:

[0047]

[0048] Among them, c′ ika represents the normalized absolute value of the ath statistical characteristic value of the kth operation information of the i-th charging station, w ka h is the weight of the statistical eigenvalue of the k-th dimension running information at the a-th dimension; jka is the ath statistical eigenvalue of the jth cluster center in the kth running information, and |||| indicates the calculation of the Euclidean distance.

[0049] Cluster center M j Update:

[0050] New cluster center The calculation method for the ath statistical feature mean of the kth dimension running information is as follows:

[0051]

[0052] in, is the updated running information value of the j-th cluster center in the k-th dimension; S j Includes all charging stations belonging to the jth cluster, C i ∈S j .

[0053] In a second aspect, the present invention provides a clustering grouping system for electric vehicle charging stations based on operation and maintenance considerations, which is used to execute the aforementioned clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations, and specifically includes: a data preprocessing module, a statistical analysis module, a weight allocation module, a weighted clustering module and an operation and maintenance optimization module;

[0054] Data preprocessing module: used to analyze and process the historical charging station operation information, using the time series feature extraction method of correlation analysis to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence;

[0055] Statistical analysis module: used to perform statistical analysis on the benchmark feature sequence and extract statistical features as a statistical feature sequence;

[0056] Weight distribution module: used to calculate the weight of each dimension of operation information using the Shelfie method, and to calculate the weight of statistical features using the entropy weight method to obtain a weight distribution set;

[0057] Weighted clustering module: used to calculate the multi-dimensional weighted distance from each charging station to the cluster center based on the weight distribution set and statistical feature sequence, and obtain the clustering results of all charging stations based on the multi-dimensional time series clustering algorithm of K-means;

[0058] Operation and maintenance optimization module: used to design operation and maintenance plans for each type in the clustering results based on the clustering results of charging stations.

[0059] In a third aspect, the present invention provides a clustering grouping device for electric vehicle charging stations based on operation and maintenance considerations, comprising a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor;

[0060] The processor is used to execute the aforementioned clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to the instructions in the computer program code.

[0061] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the aforementioned clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. In the clustering and grouping method of electric vehicle charging stations based on operation and maintenance considerations of the present invention, the operation information of the charging stations is collected for clustering, and the interaction relationship between the charging stations is dynamically reflected. Compared with the traditional clustering method based on location information, it is more flexible and has a certain robustness. The historical data of the charging station is mined by feature extraction, which can objectively mine the patterns and relationships hidden behind the data, and solve the problem of large amount of time series data and complex internal correlation. When clustering, the weights are determined by using the subjective Delphi method and the objective entropy weight method, and the weighted distance is used to measure the distance between the charging station and the cluster center. The complexity of the system is taken into account, and the clustering results are more scientific.

[0064] 2. A clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations in the present invention

[0065] Charging stations with similar characteristics are classified into one category through clustering. Charging stations with similar needs often have mutually exclusive demands for electric energy resources. Reasonable classification of them into different groups will help the power system complete fault detection and resource scheduling of charging stations. Traditional operation and maintenance are often based on the real-time operation information of a single charging station for detection. In the clustering process, this application considers the trend of the operating characteristics of a single charging station and the similarity of the operating characteristics of multiple charging stations; therefore, after clustering and grouping, charging stations with similar operating status are used as references, which improves the efficiency of monitoring and operation and improves the accuracy of charging station operation status evaluation and detection.

[0066] 3. A clustering grouping system for electric vehicle charging stations based on operation and maintenance considerations of the present invention includes a data preprocessing module, a statistical analysis module, a weight distribution module, a weighted clustering module and an operation and maintenance optimization module; the system is used to implement the steps of the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions. Therefore, the system also includes all the beneficial effects of the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions, which will not be repeated here.

[0067] 4. A clustering grouping device for electric vehicle charging stations based on operation and maintenance considerations of the present invention includes a processor and a memory, the memory is used to store computer program codes, and transmit the computer program codes to the processor, and the processor is used to execute the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions according to the instructions in the computer program codes. Therefore, the device also includes all the beneficial effects of the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions, which will not be repeated here.

[0068] 5. A computer program product of the present invention, when executed by a processor, implements the steps of the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions. Therefore, the computer program product also includes all the beneficial effects of the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations provided in any of the above technical solutions, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a flow chart of the method of the present invention.

[0070] Figure 2 It is a system diagram of the present invention.

[0071] Figure 3 It is a diagram of the equipment of the present invention.

[0072] Figure 4 It is a relationship diagram of the data processing process in Example 1.

[0073] Figure 5 It is a weighted distance process diagram from k dimensions to the cluster center in Example 1.

[0074] Figure 6 It is a weighted distance process diagram from the charging station to the cluster center in Example 1. DETAILED DESCRIPTION

[0075] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0076] Embodiment 1:

[0077] See also Figure 1 , Figures 4 to 6 ,A clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations,S1.,Based on the historical operation information of charging stations, a time series feature extraction method of correlation analysis is adopted,to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence;

[0078] Within the scope of the urban power system, there are multiple charging stations C at different geographical locations. i ,Each charging station includes K dimensions of time series operation information, and the set of raw data in each dimension is called the raw sequence;

[0079] The time series feature extraction method of correlation analysis is used to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and the extracted benchmark feature sequence is obtained:

[0080] Through discrete Fourier transform, the original subsequences x(t) of each dimension of the original sequence are transformed from the time domain to the frequency domain to obtain discrete subsequences of each dimension in the frequency domain;

[0081] For each dimension, the absolute value of each group of data in the obtained frequency domain discrete subsequence is calculated according to the formula |A(r)*cos(2π*f(r)+P(r)*π)|, where the frequency vector of the rth group of data is f(r), the amplitude vector is A(k) and the phase vector is P(r). The absolute value calculation results are sorted from large to small, and the first m groups of data are selected to synthesize a new feature subsequence: x(t)

[0082]

[0083] Where: y(t) is the new feature subsequence; calculate the Pearson coefficient between the new feature subsequence and the original subsequence:

[0084]

[0085] Among them, x(t) is the original subsequence, is the mean of the original subsequence, is the mean of the new feature subsequence, T is the total number of time points;

[0086] If the Pearson coefficient value is less than the threshold δ, the value of m is increased to resynthesize a new feature subsequence and calculate the Pearson coefficient;

[0087] If the Pearson coefficient value is greater than or equal to the threshold δ, the new feature subsequence y(t) of this dimension can be used as the k-th dimension operation information sequence of the i-th charging station Add the benchmark feature sequence of the i-th charging station

[0088] When each charging station has completely collected the new feature subsequences of each dimension, the dimensionality reduction calculation is completed and the complete benchmark feature sequence F is obtained. base .

[0089] S2, performing statistical analysis on the benchmark feature sequence and extracting statistical features as a statistical feature sequence;

[0090] In S2, for each charging station, the baseline feature sequence Perform statistical analysis on the running information of each dimension and obtain the statistical feature subsequence:

[0091]

[0092] in, represents the k-th dimension operation information of the i-th charging station in the benchmark feature sequence, The time traversal characteristics of the k-th dimension of the running information for the i-th charging station, is the discrete fluctuation characteristic of the k-th dimension operation information of the i-th charging station, is the periodic feature of the k-th dimension operation information of the i-th charging station, Run the information peak feature for the kth dimension of the i-th charging station;

[0093] Each statistical characteristic value reflects the different changing trends of the time series. Synthetic statistical feature sequence F stas .

[0094] Statistical characteristics include mean, variance, periodic characteristics and peak value, which respectively represent the overall trend, fluctuation trend, periodic trend and sudden trend of charging station operation information. Fully considering different statistical characteristic values ​​will better extract the characteristics of charging station operation information on the time scale, and the calculation of clustering distance will be more comprehensive and truly reflect the time characteristics.

[0095] Extract the statistical eigenvalues ​​of the benchmark feature sequence and extract the time traversal features:

[0096]

[0097] Where: T is the total number of time points, f i k (t) is The value at time t;

[0098] It represents the mean value of the kth operation information of the i-th charging station in the time series time T. Charging stations with similar operation status have consistency in their means.

[0099] Extract discrete fluctuation features:

[0100]

[0101] in: for The mean of

[0102] It indicates the fluctuation degree of the kth operating index of charging station i within the time series time T. Charging stations with similar operating status have consistent fluctuation trends.

[0103] Extract periodic features:

[0104]

[0105] Where: n is the total number of days, d is the dth day, For The time series data within the d-th day intercepted in ; R() is the Pearson correlation coefficient between the d-th day and the d+1-th day;

[0106] It represents the similarity of the operation information of the kth dimension of power station i in the two consecutive days within the time series time T.

[0107] Extract peak features:

[0108]

[0109] in: is the time when the peak value of the k-th dimension operation information of the i-th charging station occurs on the d-th day;

[0110] It indicates the peak value of the kth operating index of charging station i in the time series time T. When the peaks of two time series occur at similar times every day, the two charging stations have similar operating information.

[0111] S3. The Shelfie method is used to calculate the weight of the operation information of each dimension, and the entropy weight method is used to calculate the weight of the statistical features respectively to obtain a weight distribution set;

[0112] The Shelfie method is used to calculate the weight of each dimension of operation information: a number of power professionals are invited to score the importance of each dimension of operation information to the charging station in the form of a questionnaire. After statistical analysis, the weight values ​​of each dimension {e1, e2, …, e k ,…,e K};

[0113] The weights of the operating information in each dimension are combined with the weight vector of the statistical eigenvalue to obtain the weight distribution set. The Delphi method is used to calculate and 10 experts in the power industry are invited to design a questionnaire to score the importance of each dimension of the operating information to the charging station from 1 to 10. The 10 experts score to determine the scores of different dimensions, and each expert scores {s z1 ,s z2 ,…,s zk …,s zK}, calculate the weight value of each dimension: After statistical analysis, we can get the weight values ​​of different dimensions {e1,e2,…,e k ,…,e K}.

[0114] The entropy weight method is used to calculate the weights of statistical features respectively:

[0115] For the k-th dimension operation information, the matrix composed of the statistical feature sequences of all charging stations is: Each row represents the four statistical characteristic values ​​in the statistical characteristic sequence of the charging station, and each column represents the different values ​​of the same statistical characteristic value for m charging stations. The a-th statistical characteristic value at the i-th charging station is c ia ; The entropy weight method is used to calculate the weight of the statistical eigenvalue of the kth dimension, and the weight vector of the statistical eigenvalue is {w k1 ,w k2 ,…,w k4};

[0116] c ia represent The nth element in ; each column represents different values ​​of the same statistical characteristic value for m charging stations.

[0117] Normalize each element of the matrix:

[0118]

[0119] Calculate the entropy of each feature:

[0120]

[0121] in, I is the total number of charging stations.

[0122] Step C3: Calculate the weight of each feature value: The weight vectors of different feature values ​​are obtained as {w k1 ,w k2 ,…,w k4}.

[0123] S4. Based on the weight distribution set and the statistical feature sequence, the multidimensional weighted distance from each charging station to the cluster center is calculated, and the clustering results of all charging stations are obtained based on the multidimensional time series clustering algorithm of K-means;

[0124] Based on weight distribution set and statistical feature sequence F stas , calculate the multidimensional weighted distance from each charging station to the cluster center, charging station C i To the jth cluster center M j The distance Dist(C i ,M j )for:

[0125]

[0126] Among them, the cluster center M j ={H j,1 ,Hj,2 ,…,H j,k ,…,H j,K}; H j,k is the running information value of the jth cluster center in the kth dimension, H j,k ={h jk1 ,h jk2 ,h jk3 ,h jk4};

[0127] Indicates the operation information dimension H from the i-th charging station to the j-th cluster center in the k-th j,k The distance is calculated as follows:

[0128]

[0129] Among them, c′ ika represents the normalized absolute value of the ath statistical characteristic value of the kth operation information of the i-th charging station, w ka h is the weight of the statistical eigenvalue of the k-th dimension running information at the a-th dimension; jka is the ath statistical eigenvalue of the jth cluster center in the kth running information, and |||| indicates the calculation of the Euclidean distance.

[0130] New cluster center The calculation method for the ath statistical feature mean of the kth dimension running information is as follows:

[0131]

[0132] in, is the updated running information value of the j-th cluster center in the k-th dimension; S j Includes all charging stations belonging to the jth cluster, C i ∈S j .

[0133] The charging stations with similar overall trend, fluctuation trend, periodic trend and sudden trend of running status are classified into one category. Step D1: Select the number of clusters required in the end, set to K, and randomly select K charging stations as cluster centers. Denoted as M1, M2, ..., M K .

[0134] Step 1: Select the number of clusters that need to be clustered, set to J, and randomly select J charging stations as cluster centers. Denoted as M1, M2, ..., M J .

[0135] Step 2: For each charging station C i , calculate the distance between it and each cluster center as Dist(C i ,Mj ),

[0136] Assign the charging station to the nearest cluster center: Cluster (C i ) = argmin j Dist(C i ,M j );

[0137] Step 3: After clustering all charging stations, update the cluster center; for the jth cluster S j All charging stations C i ∈S j , respectively calculate the cluster S j The mean of all dimensional parameters of the charging station in the cluster constitutes the new cluster center For each cluster S j For all charging stations, the mean of all four statistical eigenvalues ​​of each dimension under all dimensional operation information is calculated. Specifically, the new cluster center The calculation method for the a-th statistical feature mean of the k-th running information is as follows:

[0138]

[0139] in, is the updated cluster center, C i ∈S j belongs to cluster S j Charging station.

[0140] Calculate the mean of all statistical features for each dimension Get the cluster center of the K-th dimension running information Calculate all Get new cluster centers

[0141] 4: Repeat steps 2 to 4 until the cluster center no longer changes and satisfies Where ∈ is the preset threshold, and finally the final clustering results M1, M2, …, M K .

[0142] S5. According to the clustering results of the charging stations, an operation and maintenance plan is designed for each type in the clustering results.

[0143] Embodiment 2:

[0144] See also Figure 2, a clustering grouping system for electric vehicle charging stations based on operation and maintenance considerations, used to execute the aforementioned clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations, specifically comprising: a data preprocessing module, a statistical analysis module, a weight allocation module, a weighted clustering module and an operation and maintenance optimization module;

[0145] Data preprocessing module: used to analyze and process the historical charging station operation information, using the time series feature extraction method of correlation analysis to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence;

[0146] Within the scope of the urban power system, there are I charging stations C at different geographical locations. Each charging station includes K dimensions of time series operation information. The collection of raw data in each dimension is called the raw sequence.

[0147] The time series feature extraction method of correlation analysis is used to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and the extracted benchmark feature sequence is obtained:

[0148] Through discrete Fourier transform, the original subsequences x(t) of each dimension of the original sequence are transformed from the time domain to the frequency domain to obtain discrete subsequences of each dimension in the frequency domain;

[0149] For each dimension, the absolute value of each group of data in the obtained frequency domain discrete subsequence is calculated according to the formula |A(r)*cos(2π*f(r)+P(r)*π)|, where the frequency vector of the rth group of data is f(r), the amplitude vector is A(k), and the phase vector is P(r). The absolute value calculation results are sorted from large to small, and the first m groups of data are selected to synthesize a new feature subsequence:

[0150]

[0151] Where: y(t) is the new feature subsequence; calculate the Pearson coefficient between the new feature subsequence and the original subsequence:

[0152]

[0153] Among them, x(t) is the original subsequence, is the mean of the original subsequence, is the mean of the new feature subsequence, T is the total number of time points;

[0154] If the Pearson coefficient value is less than the threshold δ, the value of m is increased to resynthesize a new feature subsequence and calculate the Pearson coefficient;

[0155] If the Pearson coefficient value is greater than or equal to the threshold δ, the new feature subsequence y(t) of this dimension can be used as the k-th dimension operation information sequence of the i-th charging station Add the benchmark feature sequence of the i-th charging station

[0156] When each charging station has completely collected the new feature subsequences of each dimension, the dimensionality reduction calculation is completed and the complete benchmark feature sequence F is obtained. base .

[0157] Statistical analysis module: used to perform statistical analysis on the benchmark feature sequence and extract statistical features as the statistical feature sequence; for the benchmark feature sequence of each charging station Perform statistical analysis on the running information of each dimension and obtain the statistical feature subsequence:

[0158]

[0159] in, represents the k-th dimension operation information of the i-th charging station in the benchmark feature sequence, The time traversal characteristics of the k-th dimension of the running information for the i-th charging station, is the discrete fluctuation characteristic of the k-th dimension operation information of the i-th charging station, is the periodic feature of the k-th dimension operation information of the i-th charging station, Run the information peak feature for the kth dimension of the i-th charging station;

[0160] Using the calculated statistical characteristics of subsequences Synthetic statistical feature sequence F stas .

[0161] Extract the statistical eigenvalues ​​of the benchmark feature sequence and extract the time traversal features:

[0162]

[0163] Where: T is the total number of time points, f i k (t) is The value at time t;

[0164] Extract discrete fluctuation features:

[0165]

[0166] in: for The mean of

[0167] Extract periodic features:

[0168]

[0169] Where: n is the total number of days, d is the dth day, For The time series data within the d-th day intercepted in ; R() is the Pearson correlation coefficient between the d-th day and the d+1-th day;

[0170] Extract peak features:

[0171]

[0172] in: is the time when the peak value of the k-th dimension operation information of the ith charging station occurs on the d-th day.

[0173] Weight distribution module: used to calculate the weight of each dimension of operation information using the Shelfie method, and to calculate the weight of statistical features using the entropy weight method to obtain a weight distribution set;

[0174] In S3, the Shelfie method is used to calculate the weight of the operation information of each dimension: a number of power professionals are invited to score the importance of the operation information of each dimension to the charging station in the form of a questionnaire, and the weight value of each dimension {e1, e2, …, e k ,…,e K};

[0175] The entropy weight method is used to calculate the weights of statistical features respectively:

[0176] make For the k-th dimension operation information, the statistical feature sequence F of I charging stations stas The matrix is: The a-th statistical characteristic value at the i-th charging station is c ia ; The entropy weight method is used to calculate the weight of the statistical eigenvalue of the kth dimension, and the weight vector of the statistical eigenvalue is {w k1 ,w k2 ,…,w k4};

[0177] The weight of the operating information of each dimension is combined with the weight vector of the statistical eigenvalue to obtain a weight distribution set.

[0178] Weighted clustering module: used to calculate the multi-dimensional weighted distance from each charging station to the cluster center based on the weight distribution set and statistical feature sequence, and obtain the clustering results of all charging stations based on the multi-dimensional time series clustering algorithm of K-means;

[0179] S4, based on the weight distribution set and the statistical feature sequence F stas, calculate the multidimensional weighted distance from each charging station to the cluster center, charging station C i To the jth cluster center M j The distance Dist(C i ,M j )for:

[0180]

[0181] Among them, the cluster center M j ={H j,1 ,H j,2 ,…,H j,k ,…,H j,K}; H j,k is the running information value of the jth cluster center in the kth dimension, H j,k ={h jk1 ,h jk2 ,h jk3 ,h jk4};

[0182] Indicates the operation information dimension H from the i-th charging station to the j-th cluster center in the k-th j,k The distance is calculated as follows:

[0183]

[0184] Among them, c′ ika represents the normalized absolute value of the ath statistical characteristic value of the kth operation information of the i-th charging station, w ka h is the weight of the statistical eigenvalue of the k-th dimension running information at the a-th dimension; jka is the ath statistical eigenvalue of the jth cluster center in the kth running information, and |||| indicates the calculation of the Euclidean distance.

[0185] Cluster center M j Update:

[0186] New cluster center The calculation method for the ath statistical feature mean of the kth dimension running information is as follows:

[0187]

[0188] in, is the updated running information value of the j-th cluster center in the k-th dimension; S j Includes all charging stations belonging to the jth cluster, C i ∈S j .

[0189] Operation and maintenance optimization module: used to design operation and maintenance plans for each type in the clustering results based on the clustering results of charging stations.

[0190] The purpose of the operation and maintenance optimization module is to determine the long-term operation and maintenance plan and the short-term charging station fault detection within the group. After the clustering and grouping of charging stations is completed by the present invention, it can guide the formulation of a long-term operation and maintenance plan: charging stations of the same category will have similar operating states in time. According to the operating characteristics of charging stations of different categories, a fixed time is arranged for each category of charging stations to reduce the impact of operation and maintenance on the normal operation of the charging station. Clustering and grouping improves the efficiency of monitoring and operation and maintenance.

[0191] At the same time, it can guide the short-term detection of charging station faults: the method of the present invention obtains the category of each charging station, collects the real-time operation data of all charging stations, and updates the cluster center of all charging stations in each cluster group. The weighted distance Dist(i, j) from a charging station to the cluster center is calculated. If the value is greater than the threshold, the operation status of the charging station during this period has been out of the group, and there may be anomalies, which requires special attention, and personnel should be dispatched to the charging station for operation and maintenance when necessary. By using cluster grouping data for reference, the accuracy of operation and maintenance status fault detection is improved.

[0192] Embodiment 3:

[0193] See also Figure 3 A clustering grouping device for electric vehicle charging stations based on operation and maintenance considerations includes a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor; the processor is used to execute the aforementioned clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations according to instructions in the computer program code.

[0194] Embodiment 4:

[0195] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the aforementioned clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations.

Claims

1. A clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations, characterized in that: The steps include: S1. Analyze and process the historical operation information of charging stations, and use the time series feature extraction method of correlation analysis to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence; S2, performing statistical analysis on the benchmark feature sequence and extracting statistical features as a statistical feature sequence; S3. The Shelfie method is used to calculate the weight of the operation information of each dimension, and the entropy weight method is used to calculate the weight of the statistical features respectively to obtain a weight distribution set; S4. Based on the weight distribution set and the statistical feature sequence, the multidimensional weighted distance from each charging station to the cluster center is calculated, and the clustering results of all charging stations are obtained based on the multidimensional time series clustering algorithm of K-means; S5. According to the clustering results of the charging stations, an operation and maintenance plan is designed for each type in the clustering results.

2. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 1 is characterized by: In S1, within the scope of the urban system power system, there are I charging stations C at different geographical locations, each charging station includes K dimensions of time series operation information, and the collection of original data of each dimension is called the original sequence; The time series feature extraction method of correlation analysis is used to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and the extracted benchmark feature sequence is obtained: pass Discrete Fourier transform, transforming the original subsequence x(t) of each dimension of the original sequence from the time domain to the frequency domain, and obtaining the discrete subsequence of each dimension in the frequency domain; For each dimension, the absolute value of each group of data in the obtained frequency domain discrete subsequence is calculated according to the formula |A(r)*cos(2π*f(r)+P(r)*π)|, where the frequency vector of the rth group of data is f(r), the amplitude vector is A(k), and the phase vector is P(r). The absolute value calculation results are sorted from large to small, and the first m groups of data are selected to synthesize a new feature subsequence: Where: y(t) is the new feature subsequence; calculate the Pearson coefficient between the new feature subsequence and the original subsequence: Among them, x(t) is the original subsequence, is the mean of the original subsequence, is the mean of the new feature subsequence, T is the total number of time points; If the Pearson coefficient value is less than the threshold δ, the value of m is increased to resynthesize a new feature subsequence and calculate the Pearson coefficient; If the Pearson coefficient value is greater than or equal to the threshold δ, the new feature subsequence y(t) of this dimension can be used as the k-th dimension operation information sequence of the i-th charging station Add the benchmark feature sequence of the i-th charging station When each charging station has completely collected the new feature subsequences of each dimension, the dimensionality reduction calculation is completed and the complete benchmark feature sequence F is obtained. base .

3. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 1 is characterized by: In S2, for each charging station, the baseline feature sequence Perform statistical analysis on the running information of each dimension and obtain the statistical feature subsequence: in, represents the k-th dimension operation information of the i-th charging station in the benchmark feature sequence, The time traversal characteristics of the k-th dimension of the running information for the i-th charging station, is the discrete fluctuation characteristic of the k-th dimension operation information of the i-th charging station, is the periodic feature of the k-th dimension operation information of the i-th charging station, Run the information peak feature for the kth dimension of the i-th charging station; Using the calculated statistical characteristics of subsequences Synthetic statistical feature sequence F stas .

4. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 3 is characterized by: Extract the statistical eigenvalues ​​of the benchmark feature sequence and extract the time traversal features: Where: T is the total number of time points, for The value at time t; Extract discrete fluctuation features: in: for The mean of Extract periodic features: Where: n is the total number of days, d is the dth day, For The time series data within the d-th day intercepted in ; R() is the Pearson correlation coefficient between the d-th day and the d+1-th day; Extract peak features: in: is the time when the peak value of the k-th dimension operation information of the ith charging station occurs on the d-th day.

5. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 1 is characterized by: In S3, the Shelfie method is used to calculate the weight of the operation information of each dimension: a number of power professionals are invited to score the importance of the operation information of each dimension to the charging station in the form of a questionnaire, and the weight value of each dimension {e1, e2, …, e k ,…,e K }; The entropy weight method is used to calculate the weights of statistical features respectively: make For the k-th dimension operation information, the statistical feature sequence F of I charging stations stas The matrix is: The a-th statistical characteristic value at the i-th charging station is c ia ; The entropy weight method is used to calculate the weight of the statistical eigenvalue of the kth dimension, and the weight vector of the statistical eigenvalue is {w k1 ,w k2 ,…,w k4 }; The weight of the operating information of each dimension is combined with the weight vector of the statistical eigenvalue to obtain a weight distribution set.

6. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 1 is characterized by: In S4, based on the weight distribution set and the statistical feature sequence F stas , calculate the multidimensional weighted distance from each charging station to the cluster center, charging station C i To the jth cluster center M j The distance Dist(C i ,M j )for: Among them, the cluster center M j ={H j,1 ,H j,2 ,…,H j,k ,…,H j,K }; H j,k is the running information value of the jth cluster center in the kth dimension, H j,k ={h jk1 ,h jk2 ,h jk3 ,h jk4 }; Indicates the operation information dimension H from the i-th charging station to the j-th cluster center in the k-th j,k The distance is calculated as follows: Among them, c′ ika represents the normalized absolute value of the ath statistical characteristic value of the kth operation information of the i-th charging station, w ka h is the weight of the statistical eigenvalue of the k-th dimension running information at the a-th dimension; jka is the ath statistical eigenvalue of the jth cluster center in the kth running information, and |||| indicates the calculation of the Euclidean distance.

7. The clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations according to claim 6 is characterized by: Cluster center M j Update: New cluster center The calculation method for the ath statistical feature mean of the kth dimension running information is as follows: in, is the updated running information value of the j-th cluster center in the k-th dimension; S j Includes all charging stations belonging to the jth cluster, C i ∈S j .

8. A clustering grouping system for electric vehicle charging stations based on operation and maintenance considerations, characterized in that: The system is used to execute the clustering and grouping method of electric vehicle charging stations based on operation and maintenance considerations as described in any one of claims 1 to 7, specifically comprising: a data preprocessing module, a statistical analysis module, a weight distribution module, a weighted clustering module and an operation and maintenance optimization module; Data preprocessing module: used to analyze and process the historical charging station operation information, using the time series feature extraction method of correlation analysis to extract the main frequency domain features of the corresponding charging station from the time series operation information of each charging station, and obtain the extracted benchmark feature sequence; Statistical analysis module: used to perform statistical analysis on the benchmark feature sequence and extract statistical features as a statistical feature sequence; Weight distribution module: used to calculate the weight of each dimension of operation information using the Shelfie method, and to calculate the weight of statistical features using the entropy weight method to obtain a weight distribution set; Weighted clustering module: used to calculate the multi-dimensional weighted distance from each charging station to the cluster center based on the weight distribution set and statistical feature sequence, and obtain the clustering results of all charging stations based on the multi-dimensional time series clustering algorithm of K-means; Operation and maintenance optimization module: used to design operation and maintenance plans for each type in the clustering results based on the clustering results of charging stations.

9. A clustering grouping device for electric vehicle charging stations based on operation and maintenance considerations, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor; The processor is used to execute the clustering grouping method of electric vehicle charging stations based on operation and maintenance considerations as described in any one of claims 1 to 7 according to the instructions in the computer program code.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor as the clustering grouping method for electric vehicle charging stations based on operation and maintenance considerations as described in any one of claims 1 to 7.

Images