A method for selecting a charging station site

Abstract

The present invention discloses a method for site selection of a charging station, comprising: constructing a data collector to collect charging demand data of k real charging stations and POI feature data within a preset range; obtaining weight information corresponding to different features by a hierarchical coefficient of variation method and a combined coefficient method; obtaining clusters of a target area by a clustering algorithm, treating them as charging demand points, and calculating the weight information corresponding to the features of the charging demand points to obtain the demand levels of the charging demand points; constructing a site selection model based on a white shark optimization algorithm, solving the site selection model by the white shark optimization algorithm, taking the charging demand level as input data of the site selection model, and obtaining the address of the charging station and the fixed capacity data of the charging station; the present invention obtains the optimal address and fixed capacity data of the charging station in combination with the charging demand in real scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle charging station planning, and more specifically, to a charging station site selection method. Background Art

[0002] In modern society, with the enhancement of environmental awareness and the development and utilization of renewable energy, electric vehicles, as representatives of new energy vehicles, are increasingly attracting widespread attention due to their clean and environmentally friendly characteristics. The rapid popularization of electric vehicles has put forward urgent demands for the construction of related infrastructure, especially charging stations. The site selection of charging stations, as an important part of the construction of charging infrastructure, is directly related to the convenience, economy and service quality of electric vehicle charging, and is one of the key factors in promoting the promotion and application of electric vehicles.

[0003] However, the current site selection problem of electric vehicle charging stations faces multiple challenges. First, due to the uneven distribution of electric vehicle users, how to scientifically and reasonably determine the location of charging stations to meet the charging needs of electric vehicle users in different regions has become a complex decision-making problem. Secondly, the construction cost of charging stations is relatively high, including land cost, construction cost, and operation and maintenance cost. How to achieve the optimal layout of charging infrastructure within a limited budget to maximize cost-effectiveness is also an important factor that must be considered in the site selection process. In addition, with the continuous changes in urban transportation networks and the development of electric vehicle technology, the site selection planning of charging stations also needs to be forward-looking and flexible to meet future development needs.

[0004] At present, although some studies have explored the site selection of electric vehicle charging stations, most of them focus on the use of traditional optimization algorithms and models, and rarely consider the complex constraints in actual site selection, such as geographical environment, user behavior patterns, grid load conditions, etc., which makes it difficult to directly apply these research results to the actual site selection planning of charging stations. In addition, with the continuous advancement of electric vehicle charging technology, such as the emergence of new technologies such as fast charging and wireless charging, it has also brought new challenges and opportunities to the site selection planning of charging stations.

[0005] In summary, conducting in-depth research on the site selection of electric vehicle charging stations, considering more practical application scenarios and constraints, and adopting more advanced and reasonable planning methods and technologies are of great theoretical significance and application value for promoting the scientific planning and rapid development of electric vehicle charging infrastructure. Summary of the invention

[0006] In order to solve the problem that it is difficult to determine the location of charging stations to meet the charging needs in different areas in the prior art, the present invention proposes a charging station site selection method, which combines the charging needs in real scenarios and applies the constraints in real scenarios to the charging station site selection planning. Through the site selection model based on the White Shark optimization algorithm, the optimal charging station address and fixed capacity data are obtained.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] The present invention discloses a charging station site selection method, comprising:

[0009] S1: Build a data collector to collect charging demand data of k real charging stations and POI feature data within a preset range;

[0010] S2: Obtain the weight information corresponding to different features through the hierarchical-coefficient of variation method combined with the coefficient method;

[0011] S3: Using a clustering algorithm, the clusters of the target area are obtained and regarded as charging demand points. The weight information corresponding to the characteristics of the charging demand points is calculated to obtain the demand level of the charging demand points.

[0012] S4: Construct a site selection model based on the White Shark optimization algorithm, solve the site selection model through the White Shark optimization algorithm, use the charging demand level as the input data of the site selection model, and obtain the charging station address and charging station capacity data.

[0013] Preferably, for each charging station x i First, collect the charging demand, and then collect the quantity based on M POI features;

[0014] POI features are set into 11 categories, including: catering services, retail stores, entertainment facilities, business services, scientific and educational cultural facilities, commercial residences, leisure and entertainment, companies and enterprises, transportation hubs, medical services, and public facilities;

[0015] After collecting the relevant data of k charging stations, the weight information corresponding to different features is obtained through the hierarchical-coefficient of variation method combined coefficient method.

[0016] Preferably, the hierarchical-coefficient of variation method combined coefficient method includes using the analytic hierarchy process and the coefficient of variation method to obtain the final feature weight w;

[0017] The process of using AHP includes:

[0018] Construct judgment matrix A n×n , calculate the judgment matrix A n×n The eigenvalues ​​and eigenvectors of the maximum eigenvalue λmax and the eigenvector T corresponding to the maximum eigenvalue are found, and the first weight vector W is obtained by calculation i;

[0019] The process of using the coefficient of variation method includes:

[0020] Constructing the data matrix R k×n , calculate the mean A of feature i j and standard deviation S j , and obtain the coefficient of variation V j , the second weight vector W is obtained by calculation j ;

[0021] The final feature weight is obtained by calculating two weight vectors.

[0022] Preferably, a clustering algorithm is used to obtain clusters of target areas as charging demand points, including:

[0023] To plan charging stations for electric vehicles using K-means clustering algorithm, we first collect data on points of interest in the area;

[0024] When processing POI data, the K-means clustering algorithm groups them according to the similarity of their geographical locations. During the clustering process, the algorithm minimizes the distance between each cluster point and the cluster center, forming a POI group centered on the cluster center in a multidimensional space. The cluster center is the geometric center of the cluster. POI is the POI data.

[0025] Each cluster corresponds to a dense area of ​​POIs in a region, corresponding to a higher charging demand.

[0026] Furthermore, the mathematical representation of the K-means clustering algorithm is as follows:

[0027] For a given data set X = {x1, x2, ..., xn}, where each xi is a d-dimensional vector, divide it into k clusters;

[0028] Define k center points c j , j = 1, 2, ..., k, minimize the sum of the distances from each point to the center point of the cluster to which it belongs; this is achieved by minimizing the following objective function:

[0029]

[0030] In the iterative process, by moving the center point c j To optimize the objective function until the convergence condition is reached.

[0031] Furthermore, the specific steps to obtain the cluster of the target area as the charging demand point are as follows:

[0032] S301: Determine the number of clusters K, including: selecting K as initial center points, where K is the required number of clusters.

[0033] S302: Initializing cluster centers, including: randomly selecting K POIs as initial cluster centers C = {c1, c2, ..., c k}.

[0034] S303: assigning data points to the nearest cluster center, including: for each POI point p i , assign it to the nearest cluster center; the purpose is to minimize the distance from the POI point to its cluster center; specifically expressed as:

[0035] Assign p i to cluster k such that k=argmin j ||p i -c j || 2

[0036] Among them, ||p i -c j || 2 It is POI point p i To cluster center c j The square of the Euclidean distance.

[0037] S304: recalculating the cluster center by weighted centroid summation, including: updating the center of each cluster to make it the weighted centroid of all POI points belonging to the cluster, ensuring that the regenerated cluster center tends to be in the category with large weight; for each cluster k, the new center c k The calculation is as follows:

[0038]

[0039] Among them, w i Indicates the point of interest p i The weight value corresponding to the category to which it belongs;

[0040] S305: Repeating iterations, including: repeating S303 and S304 until k clusters are clustered;

[0041] S306: Perform weighted calculation on the relevant features of the k groups to obtain the corresponding demand level, which is calculated as follows:

[0042]

[0043] Among them, N i represents the number of feature categories i in the cluster to which the current cluster center q belongs, W i Represents the weight value of the current feature i, which is the weight value after combining the hierarchical analysis method and the coefficient of variation method.

[0044] Preferably, a site selection model based on the White Shark optimization algorithm is constructed, and the mathematical expression of the site selection model is:

[0045] min S=S1+S2 (1)

[0046]

[0047]

[0048]

[0049]

[0050] ∑ r∈R y qr =1 q∈Q (6)

[0051] N q γ max y qr ≤N r T e (7)

[0052] T qr +W r -T e ≥0 (8)

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] S1 is the waiting time cost of electric vehicle users charging at charging stations; S2 is the annual fixed cost of electric vehicle charging stations;

[0059] T qr is the average travel time from charging demand node q to candidate charging station r; W r is the average waiting time of electric vehicle users at candidate charging station r; T e Indicates the expected delay time before the user charges; q is the charging demand of charging demand node q, which is measured by the charging demand level output by the classification model; qr ∈{0,1} is a binary decision variable, indicating whether the user of charging demand node q is assigned to the candidate charging station at location j; P0 is the probability that there is no electric vehicle at the charging station; is the service intensity of charging station j, β r is the number of vehicles arriving at charging station j, μ is the average charging service rate of the charging station; d qr represents the distance from charging station q to charging demand point r; It represents the maximum charging distance that the charging user at the charging demand point r can accept, measured by Euclidean distance; represents the maximum service radius of charging station q; d qs Indicates the distance between two different charging stations; d min Indicates the minimum distance between two different charging stations.

[0060] g0 is the compound interest growth rate of the charging station investment; y k T is the estimated service life of the charging station; r and S T are the number of distribution transformers installed in charging station r and the unit price of distribution transformers; N r and C M are the number of chargers installed in charging station r and the unit price of chargers; A r and S A are the land area and land acquisition unit price of charging station r respectively; B r K is the infrastructure construction cost of charging station r; r To account for daily operating costs, including employee salaries, equipment maintenance, and safety inspections;

[0061] Q is the set of charging demand points, and q∈Q={1,2,…,q max}, q max is the total number of charging demand points (equivalent to k clusters); R is the set of electric vehicle charging stations, and r∈R={1,…,r max}, r max is the total number of charging stations; γmax is the maximum simultaneous charging rate of electric vehicles at the demand point.

[0062] Furthermore, formula (1) indicates that the total cost is composed of several different cost components;

[0063] Equations (2) to (3) define the specific methods for calculating different costs;

[0064] Formulas (4), (5), (6), (7), (8), (9), (10), (11), (12), and (13) are constraints. Formulas (4) and (5) are based on the assumption that the process of electric vehicle users arriving at the charging station follows the M / M / S queuing model under the Poisson distribution. Formula (4) represents the average waiting time of electric vehicle users at candidate charging station r, and formula (5) represents the probability that there are no electric vehicle users at candidate charging station r. Formula (6) indicates that each demand point goes to and only to one charging station for charging, and formula (7) represents the demand within the service range of charging station r. The number of charging vehicles is not greater than the service capacity of the station; Formula (8) indicates that the waiting time cost of electric vehicle users charging at the charging station must be greater than 0; Formula (9) indicates the necessary conditions for the establishment of Formulas (4) and (5); Formula (10) indicates the charging demand for access to charging station j; Formula (11) indicates that the distance from the user at the charging demand point to the charging station cannot exceed the maximum charging distance they can accept; Formula (12) indicates that the service range of the charging station cannot exceed its maximum value; Formula (13) indicates that the distance between two charging stations cannot be too close to prevent excessive competition between them.

[0065] Furthermore, the site selection model is solved by the White Shark optimization algorithm, including:

[0066] S401: Population coding, including: each particle represents a charging station plan, where the position of the particle corresponds to the location and scale of the charging station; the particle coding format is as follows:

[0067] X=[x1,x2,...,x n ,y1,y2,...,y n ]n∈(n min ,n max )

[0068] x n is a non-repeating positive integer, indicating that the nth charging station is selected at the candidate location x n y n is a positive integer, indicating the scale level of the nth charging station;

[0069] S402: Initializing the population, including: randomly generating an initial particle group including m examples and particle movement speed, and setting algorithm parameters; the algorithm parameters include the number of particles, initial speed, intensity factor, inertia weight, learning factor, contraction factor, and acceleration coefficient;

[0070] S403: motion speed update, including:

[0071]

[0072] is the global optimal position within the population at the kth iteration, is the optimal position obtained by individual i at the kth iteration, c1 and c2 are both random values ​​in (0, 1), is the velocity value of the i-th population sample at the k-th iteration, is the velocity value of the i-th population sample at the k+1-th iteration;

[0073] For p1 and p2:

[0074]

[0075]

[0076] where p max and p min are the maximum and minimum values ​​of the intensity factor, respectively; k is the current number of iterations, and K is the maximum number of iterations; by setting this, the balance between global search and local development is controlled; μ is the contraction factor:

[0077]

[0078] Here, τ is the acceleration coefficient;

[0079] S404: Location update, including:

[0080]

[0081]

[0082]

[0083]

[0084] in, is the negation operator; is a bitwise operator; sgn(-) is a sign-returning function; u and l represent random numbers, and rand represents a random number uniformly distributed in the interval [0,1];

[0085] In addition, f is the frequency of wave motion:

[0086]

[0087] where f min is the minimum frequency of wave motion, f max is the maximum frequency of wave motion; mv is the driving force of the white shark that increases with the number of iterations:

[0088]

[0089] Where a0 and a1 are random numbers;

[0090] S406: The formula for updating the position of the optimal white shark is as follows:

[0091]

[0092] in, is the relative position of the current individual relative to the best individual in the population, r1, r2, r3 are all random numbers in (0,1); D ω For the relative distance:

[0093]

[0094] S S The strength of the shark's sense of smell and hearing in guiding its position update:

[0095]

[0096] S407: The final location of the individual is updated as follows:

[0097]

[0098] Where rand represents a random number uniformly distributed in the interval [0,1];

[0099] Through the White Shark optimization algorithm, the speed update and global optimal update are executed cyclically until the stopping condition is met. The stopping condition is the preset number of iterations; the optimal fitness value and site selection and capacity determination scheme for different numbers of sites n are obtained.

[0100] Furthermore, the White Shark optimization algorithm aims to optimize the objective function representing the total cost, i.e., the function defined by equation (1). In the iterative search process, the White Shark optimization algorithm guides the search direction by calculating the speed of each iteration, gradually approaching the solution that minimizes the total cost.

[0101] The White Shark optimization algorithm maintains a strong global search capability in the later stage, allowing the algorithm to jump out of the local optimal solution and find the true global optimal solution; achieving the optimal charging station layout and capacity configuration.

[0102] Compared with the prior art, the present invention has the following beneficial effects:

[0103] 1. Collect demand data of charging stations and surrounding POI data through data collectors, obtain weight information corresponding to different features through hierarchical-coefficient of variation method and combined coefficient method, and then obtain charging demand points through clustering algorithm. Combined with weight information, the demand level is obtained, which provides data input for the site selection model;

[0104] 2. Through the site selection model based on the White Shark optimization algorithm, the site selection model is solved by the White Shark optimization algorithm to obtain the optimal charging station address and charging station capacity data. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 A schematic diagram of a flow chart of a charging station site selection method provided in an embodiment.

[0106] Figure 2 A schematic diagram of data collected by a data collector provided in an embodiment. DETAILED DESCRIPTION

[0107] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0108] Example 1

[0109] In this embodiment, if Figure 1 As shown, a charging station site selection method is characterized by comprising:

[0110] S1: Build a data collector to collect charging demand data of k real charging stations and POI feature data within a preset range; the preset range is two kilometers nearby

[0111] S2: Obtain the weight information corresponding to different features through the hierarchical-coefficient of variation method combined with the coefficient method;

[0112] S3: Using a clustering algorithm, the clusters of the target area are obtained and regarded as charging demand points. The weight information corresponding to the characteristics of the charging demand points is calculated to obtain the demand level of the charging demand points.

[0113] S4: Construct a site selection model based on the White Shark optimization algorithm, solve the site selection model through the White Shark optimization algorithm, use the charging demand level as the input data of the site selection model, and obtain the charging station address and charging station capacity data.

[0114] In this embodiment, for each charging station x i First, collect the charging demand, and then collect the quantity based on M POI features;

[0115] POI features are set into 11 categories, including: catering services, retail stores, entertainment facilities, business services, scientific and educational cultural facilities, commercial residences, leisure and entertainment, companies and enterprises, transportation hubs, medical services, and public facilities;

[0116] After collecting the relevant data of k charging stations, the weight information corresponding to different features is obtained through the hierarchical-coefficient of variation method combined coefficient method.

[0117] In this embodiment, the hierarchical-coefficient of variation method combined coefficient method includes using the hierarchical analysis method and the coefficient of variation method to obtain the final feature weight w;

[0118] The process of using AHP includes:

[0119] Construct judgment matrix A n×n , calculate the judgment matrix A n×n The eigenvalues ​​and eigenvectors of the maximum eigenvalue λmax and the eigenvector T corresponding to the maximum eigenvalue are found, and the first weight vector W is obtained by calculation i ;

[0120] The process of using the coefficient of variation method includes:

[0121] Constructing the data matrix R k×n , calculate the mean A of feature i j and standard deviation S j , and obtain the coefficient of variation V j , the second weight vector W is obtained by calculation j ;

[0122] The final feature weight is obtained by calculating two weight vectors.

[0123] In this embodiment, clusters of target areas are obtained through a clustering algorithm and are used as charging demand points, including:

[0124] To plan charging stations for electric vehicles using K-means clustering algorithm, we first collect data on points of interest in the area;

[0125] When processing POI data, the K-means clustering algorithm groups them according to the similarity of their geographical locations. During the clustering process, the algorithm minimizes the distance between each cluster point and the cluster center, forming a POI group centered on the cluster center in a multidimensional space. The cluster center is the geometric center of the cluster. POI is the POI data.

[0126] Each cluster corresponds to a dense area of ​​POIs in a region, corresponding to a higher charging demand.

[0127] In this embodiment, the mathematical representation of the K-means clustering algorithm is as follows:

[0128] For a given data set X = {x1, x2, ..., xn}, where each xi is a d-dimensional vector, divide it into k clusters;

[0129] Define k center points c j , j = 1, 2, ..., k, minimize the sum of the distances from each point to the center point of the cluster to which it belongs; this is achieved by minimizing the following objective function:

[0130]

[0131] In the iterative process, by moving the center point c j To optimize the objective function until the convergence condition is reached.

[0132] More specifically, the specific steps to obtain the cluster of the target area as the charging demand point are as follows:

[0133] S301: Determine the number of clusters K, including: selecting K as initial center points, where K is the required number of clusters;

[0134] S302: Initializing cluster centers, including: randomly selecting K POIs as initial cluster centers C = {c1, c2, ..., c k};

[0135] S303: assigning data points to the nearest cluster center, including: for each POI point p i , assign it to the nearest cluster center; the purpose is to minimize the distance from the POI point to its cluster center; specifically expressed as:

[0136] Assign p i to cluster k such that k=argmin j ||p i -c j || 2

[0137] Among them, ||p i -c j || 2 It is POI point p i To cluster center c j The square of the Euclidean distance;

[0138] S304: recalculating the cluster center by weighted centroid summation, including: updating the center of each cluster to make it the weighted centroid of all POI points belonging to the cluster, ensuring that the regenerated cluster center tends to be in the category with large weight; for each cluster k, the new center c k The calculation is as follows:

[0139]

[0140] Among them, w i Indicates the point of interest p i The weight value corresponding to the category to which it belongs;

[0141] S305: Repeating iterations, including: repeating S303 and S304 until k clusters are clustered;

[0142] S306: Perform weighted calculation on the relevant features of the k groups to obtain the corresponding demand level, which is calculated as follows:

[0143]

[0144] Among them, N i represents the number of feature categories i in the cluster to which the current cluster center q belongs, W i Represents the weight value of the current feature i, which is the weight value after combining the hierarchical analysis method and the coefficient of variation method.

[0145] In this embodiment, a site selection model based on the White Shark optimization algorithm is constructed, and the mathematical expression of the site selection model is:

[0146] min S=S1+S2 (1)

[0147]

[0148]

[0149]

[0150]

[0151] ∑ r∈R y qr =1 q∈Q (6)

[0152] N q γ max y qr ≤N r T e (7)

[0153] T qr +W r -T e ≥0 (8)

[0154]

[0155]

[0156]

[0157]

[0158]

[0159] S1 is the waiting time cost of electric vehicle users charging at charging stations; S2 is the annual fixed cost of electric vehicle charging stations;

[0160] T qris the average travel time from charging demand node q to candidate charging station r; W r is the average waiting time of electric vehicle users at candidate charging station r; T e Indicates the expected delay time before the user charges; q is the charging demand of charging demand node q, which is measured by the charging demand level output by the classification model; qr ∈{0,1} is a binary decision variable, indicating whether the user of charging demand node q is assigned to the candidate charging station at location j; P0 is the probability that there is no electric vehicle at the charging station; is the service intensity of charging station j, β r is the number of vehicles arriving at charging station j, μ is the average charging service rate of the charging station; d qr represents the distance from charging station q to charging demand point r; It represents the maximum charging distance that the charging user at the charging demand point r can accept, measured by Euclidean distance; represents the maximum service radius of charging station q; d qs Indicates the distance between two different charging stations; d min Indicates the minimum distance between two different charging stations;

[0161] g0 is the compound interest growth rate of the charging station investment; y k T is the estimated service life of the charging station; r and S T are the number of distribution transformers installed in charging station r and the unit price of distribution transformers; N r and C M are the number of chargers installed in charging station r and the unit price of chargers; A r and S A are the land area and land acquisition unit price of charging station r respectively; B r K is the infrastructure construction cost of charging station r; r To account for daily operating costs, including employee salaries, equipment maintenance, and safety inspections;

[0162] Q is the set of charging demand points, and q∈Q={1,2,…,q max}, q max is the total number of charging demand points (equivalent to k clusters); R is the set of electric vehicle charging stations, and r∈R={1,…,r max}, r max is the total number of charging stations; γmax is the maximum simultaneous charging rate of electric vehicles at the demand point.

[0163] Furthermore, formula (1) indicates that the total cost is composed of several different cost components;

[0164] Equations (2) to (3) define the specific methods for calculating different costs;

[0165] Formulas (4), (5), (6), (7), (8), (9), (10), (11), (12), and (13) are constraints. Formulas (4) and (5) are based on the assumption that the process of electric vehicle users arriving at the charging station follows the M / M / S queuing model under the Poisson distribution. Formula (4) represents the average waiting time of electric vehicle users at candidate charging station r, and formula (5) represents the probability that there are no electric vehicle users at candidate charging station r. Formula (6) indicates that each demand point goes to and only to one charging station for charging, and formula (7) represents the demand within the service range of charging station r. The number of charging vehicles is not greater than the service capacity of the station; Formula (8) indicates that the waiting time cost of electric vehicle users charging at the charging station must be greater than 0; Formula (9) indicates the necessary conditions for the establishment of Formulas (4) and (5); Formula (10) indicates the charging demand for access to charging station j; Formula (11) indicates that the distance from the user at the charging demand point to the charging station cannot exceed the maximum charging distance they can accept; Formula (12) indicates that the service range of the charging station cannot exceed its maximum value; Formula (13) indicates that the distance between two charging stations cannot be too close to prevent excessive competition between them.

[0166] More specifically, the site selection model is solved by the White Shark optimization algorithm, including:

[0167] S401: Population coding, including: each particle represents a charging station plan, where the position of the particle corresponds to the location and scale of the charging station; the particle coding format is as follows:

[0168] X=[x1,x2,...,x n ,y1,y2,...,y n ]n∈(n min ,n max )

[0169] x n is a non-repeating positive integer, indicating that the nth charging station is selected at the candidate location x n y n is a positive integer, indicating the scale level of the nth charging station;

[0170] S402: Initializing the population, including: randomly generating an initial particle group including m examples and particle movement speed, and setting algorithm parameters; the algorithm parameters include the number of particles, initial speed, intensity factor, inertia weight, learning factor, contraction factor, and acceleration coefficient;

[0171] S403: motion speed update, including:

[0172]

[0173] is the global optimal position within the population at the kth iteration, is the optimal position obtained by individual i at the kth iteration, c1 and c2 are both random values ​​in (0, 1), is the velocity value of the i-th population sample at the k-th iteration, is the velocity value of the i-th population sample at the k+1-th iteration;

[0174] For p1 and p2:

[0175]

[0176]

[0177] where p max and p min are the maximum and minimum values ​​of the intensity factor, respectively; k is the current number of iterations, and K is the maximum number of iterations; by setting this, the balance between global search and local development is controlled; μ is the contraction factor:

[0178]

[0179] Here, τ is the acceleration coefficient;

[0180] S404: Location update, including:

[0181]

[0182]

[0183]

[0184]

[0185] in, is the negation operator; is a bitwise operator; sgn(-) is a sign-returning function; u and l represent random numbers, and rand represents a random number uniformly distributed in the interval [0,1];

[0186] In addition, f is the frequency of wave motion:

[0187]

[0188] where f min is the minimum frequency of wave motion, f max is the maximum frequency of wave motion; mv is the driving force of the white shark that increases with the number of iterations:

[0189]

[0190] Where a0 and a1 are random numbers;

[0191] S406: The formula for updating the position of the optimal white shark is as follows:

[0192]

[0193] in, is the relative position of the current individual relative to the best individual in the population, r1, r2, r3 are all random numbers in (0,1); D ω For the relative distance:

[0194]

[0195] S s The strength of the shark's sense of smell and hearing in guiding its position update:

[0196]

[0197] S407: The final location of the individual is updated as follows:

[0198]

[0199] Where rand represents a random number uniformly distributed in the interval [0,1];

[0200] Through the White Shark optimization algorithm, the speed update and global optimal update are executed cyclically until the stopping condition is met. The stopping condition is the preset number of iterations; the optimal fitness value and site selection and capacity determination scheme for different numbers of sites n are obtained.

[0201] More specifically, the White Shark optimization algorithm aims to optimize the objective function representing the total cost, i.e., the function defined by equation (1). During the iterative search process, the White Shark optimization algorithm guides the search direction by calculating the speed of each iteration, gradually approaching the solution that minimizes the total cost.

[0202] The White Shark optimization algorithm maintains a strong global search capability in the later stage, allowing the algorithm to jump out of the local optimal solution and find the true global optimal solution; achieving the optimal charging station layout and capacity configuration.

[0203] Example 2

[0204] In this embodiment, if Figure 2As shown, for each charging station xi, its charging demand is first collected, and then the quantity is collected on M POI features, including but not limited to: catering services, retail stores, entertainment facilities, business services, scientific and educational cultural facilities, commercial residences, leisure and entertainment, companies and enterprises, transportation hubs, medical services, public facilities, and a charging station data collector containing rich information is constructed.

[0205] In this embodiment, POI sets 11 specific types:

[0206] Catering services: Restaurants, cafes, fast food restaurants, etc. provide catering services for drivers waiting to charge.

[0207] Retail stores: convenience stores, supermarkets, etc., so that users can shop while waiting for charging.

[0208] Entertainment facilities: such as cinemas, amusement parks, game halls, etc., providing leisure and entertainment services.

[0209] Business services: such as hotels, conference rooms, business centers, etc., to provide convenience for business people.

[0210] Scientific, educational and cultural facilities: schools, libraries, museums, etc. These places are usually crowded and may require charging services.

[0211] Commercial and residential: Charging stations near residential areas can provide residents with convenient charging services.

[0212] Recreation and entertainment: such as parks, stadiums, etc., which may attract a large number of tourists and vehicles.

[0213] Enterprises: Industrial parks or business offices where employees or visitors may need to use charging stations.

[0214] Transportation hubs: such as railway stations, subway stations, airports, etc., providing charging services for passengers taking public transportation.

[0215] Medical services: hospitals, clinics, etc., providing convenience for medical staff and patients.

[0216] Public facilities: such as public parking lots, government agencies, etc., provide charging services while providing public services.

[0217] After collecting the relevant data of k charging stations, the hierarchical-coefficient of variation method and combined coefficient method are used to obtain the weight information corresponding to different features. The specific steps are as follows:

[0218] First, construct the judgment matrix a nn The scaling method is shown in Table 1.

[0219] Table 1

[0220]

[0221] In this embodiment, the number of n is 12, which are the charging demand and the number of 11 POI types within 2 kilometers of the charging station xi. n×n Then calculate its eigenvalues ​​and eigenvectors, find the maximum eigenvalue λmax and the eigenvector corresponding to the maximum eigenvalue T = [t1, t2, t3 … ,t n ].

[0222] Then we can get the weight of feature i Thus, we get the subjective weight obtained by the hierarchical analysis method. Then we use the coefficient of variation method to get the objective law of the data itself to assign weights. The specific method is to construct a data matrix Calculate the mean of feature i:

[0223] Then calculate the standard deviation of feature i Then we get the coefficient of variation So as to obtain the weight

[0224] Where R k×n represents the n features of k real charging stations (in this case, n is 12, which are charging demand, charging station x i The number of 11 POI types within 2 kilometers) matrix, r kn represents the nth feature number of the kth charging station; j∈(1,n).

[0225] Finally, through The final feature weights corresponding to n dimensions can be obtained.

[0226] Example 3

[0227] In this embodiment, a charging station site selection method includes:

[0228] The relevant characteristics of the charging demand point are weighted and calculated to obtain the demand level of the charging demand point.

[0229] The clusters obtained by clustering the study area are regarded as charging demand points. The dimensional characteristics of the corresponding charging demand points are brought into the weights obtained by the hierarchical-coefficient of variation method and the combined coefficient method, and the demand level of the demand points can be obtained, which can be used for the subsequent site selection model.

[0230] The charging demand level and the remaining information required in the mathematical model for optimizing the location and capacity of charging stations are input into the White Shark optimization algorithm model to determine the final charging station address and the specific type and number of charging guns.

[0231] K-means clustering algorithm plays a vital role in the planning of electric vehicle charging stations. The planning process starts with the collection of POI data in a specific area, which identifies possible high-demand charging locations in the area, such as restaurants, shopping malls, parks, office buildings, etc. These locations are usually active areas with concentrated traffic in the city, and thus become potential hotspots for electric vehicle charging demand.

[0232] Determine the number of clusters K: Select K as the initial center points, where K is the number of clusters you want.

[0233] Initialize cluster centers: randomly select K POIs as initial cluster centers C = {c1, c2, ..., c k}

[0234] Assign data points to the nearest cluster center: For each POI point pi, assign it to the nearest cluster center. The purpose of this step is to minimize the distance between the POI point and its cluster center. The specific operation can be expressed as:

[0235] Assign p i to cluster k such that k=argmin j ||p i -c j || 2

[0236] Among them, ||p i -c j || 2 It is POI point p i To cluster center c j The square of the Euclidean distance.

[0237] Recalculate cluster centers: Update the center of each cluster so that it becomes the mean of all POIs belonging to that cluster. For each cluster k, the new center c k The calculation is as follows:

[0238] The cluster centers are recalculated by weighted centroid summation, including: updating the center of each cluster to make it the weighted centroid of all POI points belonging to the cluster, ensuring that the regenerated cluster centers tend to be in the category with large weights; for each cluster k, the new center c k The calculation is as follows:

[0239]

[0240] Among them, w i Indicates the point of interest p i The weight value corresponding to the category to which it belongs;

[0241] Repeat the iteration: Repeat the steps of assigning data points to the nearest cluster center and recalculating the cluster center until k clusters are clustered.

[0242] The scoring criteria of k groups are counted and made into a data set to input the classification model, and their charging demand levels are output.

[0243] In the planning of the location and capacity of electric vehicle charging stations, we propose a mathematical model that takes into account the interests of multiple parties based on a comprehensive perspective. This includes the economic benefits of charging station operators, the rational layout of power infrastructure, and the convenience of electric vehicle users. Our goal is to form an optimization framework that can minimize the total cost to the entire society.

[0244] The model is formalized as a series of formulas (from (1) to (13)) that capture the key economic and technical parameters of charging station planning. These formulas include construction and operation costs, travel costs for users to charging stations, time costs for users to charging stations, and other necessary constraints, such as the balance between electricity demand and supply, and the relationship between charging station capacity and car owner demand. By solving this optimization problem, we aim to provide a scientific decision support for the rational layout of charging stations, so as to achieve the optimization of social resource allocation while meeting the development needs of electric vehicles.

[0245] min S=S1+S2 (1)

[0246]

[0247]

[0248]

[0249]

[0250] ∑ r∈R y qr =1 q∈Q (6)

[0251] N q γ max y qr ≤N r T e (7)

[0252] T qr +W r -T e ≥0 (8)

[0253]

[0254]

[0255]

[0256]

[0257]

[0258] S1 is the waiting time cost of electric vehicle users charging at charging stations; S2 is the annual fixed cost of electric vehicle charging stations;

[0259] T qr is the average travel time from charging demand node q to candidate charging station r; W r is the average waiting time of electric vehicle users at candidate charging station r; T e Indicates the expected delay time before the user charges; q is the charging demand of charging demand node q, which is measured by the charging demand level output by the classification model; qr ∈{0,1} is a binary decision variable, indicating whether the user of charging demand node q is assigned to the candidate charging station at location j; P0 is the probability that there is no electric vehicle at the charging station; is the service intensity of charging station j, β r is the number of vehicles arriving at charging station j, μ is the average charging service rate of the charging station; d qr represents the distance from charging station q to charging demand point r; It represents the maximum charging distance that the charging user at the charging demand point r can accept, measured by Euclidean distance; represents the maximum service radius of charging station q; d qs Indicates the distance between two different charging stations; d min Indicates the minimum distance between two different charging stations;

[0260] g0 is the compound interest growth rate of the charging station investment; y k T is the estimated service life of the charging station; r and S T are the number of distribution transformers installed in charging station r and the unit price of distribution transformers; N r and C M are the number of chargers installed in charging station r and the unit price of chargers; A r and S A are the land area and land acquisition unit price of charging station r respectively; B r K is the infrastructure construction cost of charging station r; r To account for daily operating costs, including employee salaries, equipment maintenance, and safety inspections;

[0261] Q is the set of charging demand points, and q∈Q={1,2,…,q max}, q max is the total number of charging demand points (equivalent to k clusters); R is the set of electric vehicle charging stations, and r∈R={1,…,r max}, r max is the total number of charging stations; γmax is the maximum simultaneous charging rate of electric vehicles at the demand point.

[0262] The following are the steps for site selection and capacity determination of electric vehicle charging stations based on the White Shark optimization algorithm:

[0263] In the data preparation stage, traditional methods usually use simulated data to generate CAD drawings of urban roads, and estimate the demand for each charging demand point based on this. In contrast, this embodiment introduces real data. By inputting these data, the weights of different features are first obtained through the hierarchical-coefficient of variation method, and then the charging demand points are obtained based on clustering. Then, based on the weights of different features, a more accurate and practical charging demand point level is obtained. This method of using real data not only enhances the practicality of the model, but also improves the accuracy of charging demand prediction.

[0264] S401: Population coding, including: each particle represents a charging station plan, where the position of the particle corresponds to the location and scale of the charging station; the particle coding format is as follows:

[0265] X=[x1,x2,...,x n ,y1,y2,...,y n ]n∈(n min ,n max )

[0266] x n is a non-repeating positive integer, indicating that the nth charging station is selected at the candidate location x n y n is a positive integer, indicating the scale level of the nth charging station.

[0267] S402: Initializing the population, including: randomly generating an initial particle group containing m examples and particle movement speed, and setting algorithm parameters; the algorithm parameters include the number of particles, initial speed, intensity factor, inertia weight, learning factor, contraction factor, and acceleration coefficient.

[0268] S403: motion speed update, including:

[0269]

[0270] is the global optimal position within the population at the kth iteration, is the optimal position obtained by individual i at the kth iteration, c1 and c2 are both random values ​​in (0, 1), is the velocity value of the i-th population sample at the k-th iteration, is the velocity value of the i-th population sample at the k+1-th iteration.

[0271] For p1 and p2:

[0272]

[0273]

[0274] where p max and p min are the maximum and minimum values ​​of the intensity factor, respectively; k is the current number of iterations, and K is the maximum number of iterations; by setting this, the balance between global search and local development is controlled; μ is the contraction factor:

[0275]

[0276] Here, τ is the acceleration coefficient;

[0277] S404: Location update, including:

[0278]

[0279]

[0280]

[0281]

[0282] in, is the negation operator; is a bitwise operator; sgn(-) is a sign-returning function; u and l represent random numbers, and rand represents a random number uniformly distributed in the interval [0,1];

[0283] In addition, f is the frequency of wave motion:

[0284]

[0285] where f min is the minimum frequency of wave motion, f max is the maximum frequency of wave motion; mv is the driving force of the white shark that increases with the number of iterations:

[0286]

[0287] Where a0 and a1 are random numbers;

[0288] S406: The formula for updating the position of the optimal white shark is as follows:

[0289]

[0290] in, is the relative position of the current individual relative to the best individual in the population, r1, r2, r3 are all random numbers in (0,1); D ω For the relative distance:

[0291]

[0292] S S The strength of the shark's sense of smell and hearing in guiding its position update:

[0293]

[0294] S407: The final location of the individual is updated as follows:

[0295]

[0296] Where rand represents a random number uniformly distributed in the interval [0,1];

[0297] Through the White Shark optimization algorithm, the speed update and global optimal update are executed cyclically until the stopping condition is met. The stopping condition is the preset number of iterations; the optimal fitness value and site selection and capacity determination scheme for different numbers of sites n are obtained.

[0298] Obviously, the above embodiments of the present invention are only examples for clearly explaining the present invention, and are not intended to limit the implementation methods of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A method for selecting a charging station site, characterized in that: include: S1: Build a data collector to collect charging demand data of k real charging stations and POI feature data within a preset range; S2: Obtain the weight information corresponding to different features through the hierarchical-coefficient of variation method combined with the coefficient method; S3: Using a clustering algorithm, the clusters of the target area are obtained and regarded as charging demand points. The weight information corresponding to the characteristics of the charging demand points is calculated to obtain the demand level of the charging demand points. S4: Construct a site selection model based on the White Shark optimization algorithm, solve the site selection model through the White Shark optimization algorithm, use the charging demand level as the input data of the site selection model, and obtain the charging station address and charging station capacity data; the mathematical expression of the site selection model is: min S=S1+S2 (1) ∑ r∈R and qr =1 q∈Q (6) N q γ max y qr ≤N r T e (7) T qr +W r -T e ≥0 (8) S1 is the waiting time cost of electric vehicle users charging at charging stations; S2 is the annual fixed cost of electric vehicle charging stations; T qr is the average travel time from charging demand node q to candidate charging station r; W r is the average waiting time of electric vehicle users at candidate charging station r; T e Indicates the expected delay time before the user charges; q is the charging demand of charging demand node q, which is measured by the charging demand level output by the classification model; qr ∈{0,1} is a binary decision variable, indicating whether the user of charging demand node q is assigned to the candidate charging station at location j; P0 is the probability that there is no electric vehicle at the charging station; is the service intensity of charging station j, β r is the number of vehicles arriving at charging station j, μ is the average charging service rate of the charging station; d qr represents the distance from charging station q to charging demand point r; It represents the maximum charging distance that the charging user at the charging demand point r can accept, measured by Euclidean distance; represents the maximum service radius of charging station q; d qs Indicates the distance between two different charging stations; d min Indicates the minimum distance between two different charging stations; g0 is the compound interest growth rate of the charging station investment; y k T is the estimated service life of the charging station; r and S T are the number of distribution transformers installed in charging station r and the unit price of distribution transformers; N r and C M are the number of chargers installed in charging station r and the unit price of chargers; A r and S A are the land area and land acquisition unit price of charging station r respectively; B r K is the infrastructure construction cost of charging station r; r To account for daily operating costs, including employee salaries, equipment maintenance, and safety inspections; Q is the set of charging demand points, and q∈Q={1,2,…,q max }, q max is the total number of charging demand points, i.e., k clusters; R is the set of electric vehicle charging stations, and r∈R={1,…,r max }, r max is the total number of charging stations; γmax is the maximum simultaneous charging rate of electric vehicles at the demand point; Formula (1) indicates that the total cost is composed of several different cost components; Equations (2) to (3) define the specific methods for calculating different costs; Formulas (4), (5), (6), (7), (8), (9), (10), (11), (12), and (13) are constraints. Formulas (4) and (5) are based on the assumption that the process of electric vehicle users arriving at the charging station follows the M / M / S queuing model under the Poisson distribution. Formula (4) represents the average waiting time of electric vehicle users at candidate charging station r, and formula (5) represents the probability that there are no electric vehicle users at candidate charging station r. Formula (6) indicates that each demand point goes to and only to one charging station for charging, and formula (7) represents the demand within the service range of charging station r. The number of charging vehicles is not greater than the service capacity of the station; Formula (8) indicates that the waiting time cost of electric vehicle users charging at the charging station must be greater than 0; Formula (9) indicates the necessary conditions for the establishment of Formulas (4) and (5); Formula (10) indicates the charging demand for access to charging station j; Formula (11) indicates that the distance from the user at the charging demand point to the charging station cannot exceed the maximum charging distance they can accept; Formula (12) indicates that the service range of the charging station cannot exceed its maximum value; Formula (13) indicates that the distance between two charging stations cannot be too close to prevent excessive competition between them; The site selection model is solved by the White Shark optimization algorithm, including: S401: Population coding, including: each particle represents a charging station plan, where the position of the particle corresponds to the location and scale of the charging station; the particle coding format is as follows: X=[x1,x2,…,x n ,y1,y2,…,y n ]n∈(n min ,n max ) x n is a non-repeating positive integer, indicating that the nth charging station is selected at the candidate location x n y n is a positive integer, indicating the scale level of the nth charging station; S402: Initializing the population, including: randomly generating an initial particle group including m examples and particle movement speed, and setting algorithm parameters; the algorithm parameters include the number of particles, initial speed, intensity factor, inertia weight, learning factor, contraction factor, and acceleration coefficient; S403: motion speed update, including: is the global optimal position within the population at the kth iteration, is the optimal position obtained by individual i at the kth iteration, c1 and c2 are both random values ​​in (0, 1), is the velocity value of the i-th population sample at the k-th iteration, is the velocity value of the i-th population sample at the k+1-th iteration; For p1 and p2: where p max and p min are the maximum and minimum values ​​of the intensity factor, respectively; k is the current number of iterations, K is the maximum number of iterations; μ is the shrinkage factor: Here, τ is the acceleration coefficient; S404: Location update, including: in, is the negation operator; is a bitwise operator; sgn(-) is a sign-returning function; u and l represent random numbers, and rand represents a random number uniformly distributed in the interval [0,1]; In addition, f is the frequency of wave motion: where f min is the minimum frequency of wave motion, f max is the maximum frequency of wave motion; mv is the driving force of the white shark that increases with the number of iterations: Where a0 and a1 are random numbers; S406: The formula for updating the position of the optimal white shark is as follows: in, is the relative position of the current individual relative to the best individual in the population, r1, r2, r3 are all random numbers in (0,1); D ω For the relative distance: S S The strength of the shark's sense of smell and hearing in guiding its position update: S407: The final location of the individual is updated as follows: Where rand represents a random number uniformly distributed in the interval [0,1]; Through the White Shark optimization algorithm, the speed update and global optimal update are executed cyclically until the stopping condition is met. The stopping condition is the preset number of iterations. The optimal fitness value and site selection and capacity determination scheme for different number of sites n are obtained. is the current position vector of the i-th population sample in the k-th iteration.

2. A charging station site selection method according to claim 1, characterized in that: For each charging station x i First, collect the charging demand, and then collect the quantity based on M POI features; POI features are set into 11 categories, including: catering services, retail stores, entertainment facilities, business services, scientific and educational cultural facilities, commercial residences, leisure and entertainment, companies and enterprises, transportation hubs, medical services, and public facilities; After collecting the relevant data of k charging stations, the weight information corresponding to different features is obtained through the hierarchical-coefficient of variation method combined coefficient method.

3. A charging station site selection method according to claim 1, characterized in that: The combined coefficient method of AHP and coefficient of variation method includes using AHP and coefficient of variation method to obtain the final feature weight w; The process of using AHP includes: Construct judgment matrix A n×n , calculate the judgment matrix A n×n The eigenvalues ​​and eigenvectors of the maximum eigenvalue λmax and the eigenvector T corresponding to the maximum eigenvalue are found, and the first weight vector W is obtained by calculation i ; The process of using the coefficient of variation method includes: Constructing the data matrix R k×n , calculate the mean A of feature i j and standard deviation S j , and obtain the coefficient of variation V j , the second weight vector W is obtained by calculation j ; The final feature weight is obtained by calculating two weight vectors.

4. A charging station site selection method according to claim 1, characterized in that: Through the clustering algorithm, the clusters of the target area are obtained as charging demand points, including: Using K-means clustering algorithm to plan charging stations for electric vehicles, first collect the point of interest data in the target area; When processing POI data, the K-means clustering algorithm groups them according to the similarity of their geographical locations. During the clustering process, the algorithm minimizes the distance between each cluster point and the cluster center, forming a POI group centered on the cluster center in a multidimensional space. The cluster center is the geometric center of the cluster. POI is the POI data. Each cluster corresponds to a dense area of ​​POIs in a region, corresponding to a higher charging demand.

5. A charging station site selection method according to claim 4, characterized in that: The mathematical representation of the K-means clustering algorithm is as follows: For a given data set X = {x1, x2, ..., x n }, where each xi is a 2D vector, which is divided into k clusters; Define k center points c j , j = 1, 2, ..., k, minimize the sum of the distances from each point to the center point of the cluster to which it belongs; this is achieved by minimizing the following objective function: In the iterative process, by moving the center point c j To optimize the objective function until the convergence condition is reached.

6. A charging station site selection method according to claim 4, characterized in that: The clustering algorithm is used to obtain the clusters of the target area, and the specific steps to use them as charging demand points are as follows: S301: Determine the number of clusters K, including: selecting K cluster centers as initial center points, where K is the required number of clusters; S302: Initializing cluster centers, including: randomly selecting K POIs as initial cluster centers C = {c1, c2, ..., c k }; S303: assigning data points to the nearest cluster center, including: for each POI point p i , assign it to the nearest cluster center; the purpose is to minimize the distance from the POI point to its cluster center; specifically expressed as: Assign p i to cluster k such that k=argmin j ||p i -c j || 2 Among them, ||p i -c j || 2 It is POI point p i To cluster center c j The square of the Euclidean distance; S304: recalculating the cluster center by weighted centroid summation, including: updating the center of each cluster to make it the weighted centroid of all POI points belonging to the cluster, ensuring that the regenerated cluster center tends to be in the category with large weight; for each cluster k, the new center c k The calculation is as follows: Among them, w i Indicates the point of interest p i The weight value corresponding to the category to which it belongs; S305: Repeating iterations, including: repeating S303 and S304 until k clusters are clustered; S306: Perform weighted calculation on the relevant features of the k groups to obtain the corresponding demand level, which is calculated as follows: Among them, N i represents the number of feature categories i in the cluster to which the current cluster center q belongs, W i Represents the weight value of the current feature i, which is the weight value after combining the hierarchical analysis method and the coefficient of variation method.

7. A charging station site selection method according to claim 6, characterized in that: The White Shark optimization algorithm aims to optimize the objective function representing the total cost, that is, the function defined by equation (1). During the iterative search process, the White Shark optimization algorithm guides the search direction by calculating the speed of each iteration, gradually approaching the solution that minimizes the total cost.

Images