A method for planning electric vehicle charging stations based on multi-objective optimization

Abstract

The application discloses a kind of electric vehicle charging station planning methods based on multi-objective optimization, including the following contents:1, a maximum charging station annual profit and minimization user waiting time multi-objective optimization electric vehicle charging station planning model is constructed;2, by comprehensively considering cost and distance cost, user selects the charging station of minimum cost to charge by greedy strategy;3, by designing efficient local search strategy, the optimal charging pile quantity and service fee price of each site are optimized;4, the model is optimized using improved MOEA / D-M2M algorithm.This method can optimize charging station annual profit and user charging waiting time in electric vehicle charging station location problem simultaneously, quickly and efficiently obtain Pareto optimal solution, provide charging station location and determine the multiple optimal schemes that can be selected for electric vehicle charging station construction enterprise charging pile quantity and service fee price.

Description

TECHNICAL FIELD

[0001] The application relates to a multi-objective optimization-based electric vehicle charging station planning method, in particular to a MOEA / D-M2M-based electric vehicle charging station planning method which can determine the positions of charging stations, the numbers of charging piles and the service prices of the electric vehicle charging stations. BACKGROUND

[0002] With the increasing number of electric vehicles, the demand for electric vehicle charging stations is increasing, and therefore, more charging stations need to be built. Existing electric vehicle charging station site planning models mainly include:

[0003] 1. A single-objective planning model in which investment cost, user time cost, environmental cost, power grid capacity and one or more targets are aggregated into one target;

[0004] 2. A multi-objective planning model considering station construction cost and user waiting time, or station construction cost and user distance from the charging station;

[0005] However, the above two methods have the following disadvantages: 1. No multi-objective optimization of charging station annual profit and user average waiting time; 2. No consideration of price influencing factors in station site selection; 3. No consideration of the upper limit of the number of buildable charging piles at each candidate point and the full use of charging station sites; therefore, no Pareto optimal station construction scheme close to reality and feasible is provided for electric vehicle charging station constructors. SUMMARY

[0006] The application aims to solve the above problems and provide an electric vehicle charging station site selection method which simultaneously optimizes charging station annual profit and user average waiting time, maximizes the utilization rate of charging station construction sites and minimizes user cost.

[0007] The application can be achieved by the following technical solutions:

[0008] A multi-objective optimization-based electric vehicle charging station planning method, comprising the following steps:

[0009] (1) A multi-objective optimization electric vehicle charging station planning model which maximizes charging station annual profit and minimizes user waiting time is constructed.

[0010] (2) By comprehensively considering cost and distance costs, users select the charging station with the minimum cost for charging through a greedy strategy.

[0011] (3) By designing an efficient local search strategy, the optimal number of charging piles and service prices of each station are optimized.

[0012] (4) The model is optimized by using an improved MOEA / D-M2M algorithm.

[0013] The content (1) specifically refers to:

[0014] Construct a multi-objective optimization model for electric vehicle charging station planning that maximizes annual charging station profits and minimizes user waiting times:

[0015] Assuming there are n candidate charging stations, for each candidate station i∈I={1,2,…,n}, at most one charging station can be built. Due to objective limitations such as site availability, there is an upper limit to the number of charging piles that can be built at each charging station. Through research, the number of charging piles in a typical charging station is between 5 and 50. This patent sets the upper limit for the number of charging piles in the charging station planning to be between 5 and 50 (inclusive). Assuming there are m users, each user j∈J={1,2,…,m} can choose at most one charging station, and the vector Q=(q1,q2,…,q…) n Records the maximum number of charging piles that can be built at each candidate site. Vector X = (x1, x2, ..., x...) n The ) indicates the construction status of each candidate site, where

[0016]

[0017] Vector P = (p1, p2, ..., p) n This indicates the number of charging piles at each candidate site that has been built.

[0018]

[0019] Matrix Z = [Z ij ] n×m This represents the relationship between user j and charging station i.

[0020]

[0021] Based on this, this patent models the electric vehicle charging station planning problem as a multi-objective optimization model that maximizes the annual profit of the charging station and minimizes the user waiting time.

[0022] Maximize the annual profit of charging stations:

[0023]

[0024] Where w is the power of a single charging pile; t is the average charging time, i.e., service time; θ is the percentage of annual operation and maintenance costs for each charging station to the total investment; c is the cost of a single charging pile; and p i c represents the total investment cost of each charging station i; vector u i This represents the number of users served by charging station i per day, i.e., the total number of users who choose charging station i.

[0025]

[0026] is the annual income of charging station i; θx i p i c is the annual operation and maintenance cost of charging station i; is the annual profit of charging station i. s i is the service price of charging station i (yuan / person / hour), which is assumed in this patent that the charging service price is related to the load of the charging station. The higher the load, the greater the demand for charging, and according to the supply and demand relationship, the higher the service price; otherwise, the lower the charging service price. Thus

[0027]

[0028] where a is the lower limit of the service price of the charging station, and b is the upper limit of the service price of the charging station. Through investigation, this patent sets a = 0.5 yuan / person / hour, and b = 2.0 yuan / person / hour. Wherein represents the maximum value of daily service users corresponding to the case that charging station i has p i piles and the average waiting time of users is 15 minutes. This patent uses the average waiting time calculation formula of the M / G / K queuing model

[0029]

[0030] The minimum average waiting time of user charging, the user charging queuing problem conforms to the M / G / K queuing model of queuing theory, so according to the average waiting time calculation formula of the M / G / K queuing model, the total user charging average waiting time is calculated:

[0031]

[0032] where V T is the variance of the user charging time obeying Gaussian distribution; ρ′ i = u i t;

[0033] The following constraints are met:

[0034]

[0035]

[0036] Where constraint (6) ensures that each user j selects at most one charging station; constraint (7) ensures that user j can only select a built charging station; constraint (9) ensures that each charging station in the built station scheme has an annual profit of no less than 0. Since the electric vehicle charging station planning model established in this project is a high-dimensional multi-objective optimization problem, and it is very difficult to optimize all variables at the same time, this project divides the multi-objective optimization problem into three sub-problems: user autonomous decision-making layer (lower layer), charging station charging pile quantity decision-making layer (middle layer), and charging station location decision-making layer (upper layer) through step-by-step alternating optimization. In these sub-problems, the lower layer decision is limited by the higher layer decision, i.e. users can only choose built charging stations; the number of charging piles and service prices can only be selected at the candidate location of the charging station to be built. In turn, the lower layer decision will affect the higher layer decision, for example, the user's decision will affect the charging station's revenue and load, thereby affecting the decision of the upper layer problem. The overall process is as follows Figure 1 , specifically:

[0037] The specific content of the lower layer sub-problem is:

[0038] Under the given middle and upper layer decisions, i.e. built station location X, built charging station charging pile quantity P, and charging station service price S, users make autonomous decisions by comprehensively considering the cost and distance cost.

[0039]

[0040] Where λ is the balance factor, used to balance the distance cost and the cost, D ij is the distance from user j to charging station i. This patent uses a greedy strategy to solve problem (10), and assigns each user j to the charging station i* that has the minimum cost for user j:

[0041] Thus, z i*j = 1, and other z ij = 0, to obtain the user allocation matrix Z.

[0042] The specific content of the middle layer sub-problem is:

[0043] Under the built station location X and the user allocation matrix Z, use a local search strategy to optimize the optimal charging pile quantity P of each station, maximizing the charging station revenue.

[0044]

[0045] Where is the charging station load rate.

[0046] (a) Update the number of charging piles: solve problem (12) under the condition of station site selection X and user allocation matrix Z to find an optimal such that

[0047]

[0048] where is the upper limit of the number of users that can be accommodated per day when the number of charging piles is and the waiting time of users is not more than 15 minutes.

[0049] (b) Update the service price: calculate the load μ of each charging station and set the service price S = a + (b-a) x min(μ, 1); obtain the number of charging piles P and the service price S corresponding to station site selection X.

[0050] The specific content of the upper sub-problem is:

[0051] Based on the improved MOEA / D-M2M, the following steps and modules are used to optimize the station site selection X. The process is shown in Figure 2 :

[0052] S1, initialize using M1 initialization module; where the specific content of M1 initialization module is:

[0053] The main idea of MOEA / D-M2M is to decompose the multi-objective optimization problem into multiple multi-objective optimization sub-problems, and solve these sub-problems simultaneously in one run; select K uniformly distributed unit vectors v 1 ,..., v K as center vectors from the first quadrant limit of the solution space , and divide into K disjoint sub-regions Ω1,..., Ω K according to the K center vectors; where Ω k , k = 1,..., K

[0054]

[0055] where f = (F1, F2) is the objective value vector of the individual, <f, v d > is the acute angle of f to v d , i.e. f ∈ Ω k if and only if v k has the smallest acute angle with f among the K center vectors;

[0056] Select L uniformly distributed weight vectors w 1 ,..., w L in each sub-region; binary encode the station site selection scheme X as Figure 3The sub-population O of each sub-region maintains the population size L k The population G of the population size K x L is composed of {O1, …, O k Each station site X is randomly initialized, and the elite external set is initialized

[0057] S2, using the M2 target function value calculation module, calculating the target function value of each station site X;

[0058] Wherein the M2 target function value calculation module: according to formula (1), the total annual profit F1(X, P, Z) of each station site X is calculated; according to formula (5), the user average waiting time F2(X, P, Z) of each station site X is calculated.

[0059] S3, traversing the sub-population of each sub-region, for each individual X,

[0060] (a) using M3 crossover mutation module for crossover mutation;

[0061] (b) using M4 local search module again, deleting the charging station with low load;

[0062] (c) using M2 module to calculate the target function value again; the new individual is added to the offspring population H. After traversing K regions, the offspring population H of the population size K x L is obtained;

[0063] (d) merging the offspring population H and the original population G, obtaining a population R of the population size 2 x K x L.

[0064] (e) according to formula (14), the individuals in the population R are distributed to K sub-regions, and the sub-population size of each sub-region is ensured to be L, and K new sub-populations constitute a new population G1; the rule of ensuring the sub-population size of each sub-region to be L is as shown in Figure 4 , and specifically:

[0065] If the number of individuals N assigned to a sub-region is L, the N individuals constitute a new sub-population; if N < L, L-N individuals are randomly selected from the population R to make up L, and the L individuals constitute a new sub-population; if N > L, the N individuals are sequentially calculated the Chebyshev distance in the weight vector w 1 , …, w L of the sub-region, and the individual with the smallest distance is selected as the new individual of the sub-region.

[0066]

[0067] Wherein, z *= (maxF1, minF2), select the individual with the minimum Chebyshev distance under each weight to join the new sub-population, and finally obtain a sub-population containing L individuals; thus, the size of each regional sub-population is ensured to be L.

[0068] (f) Traverse the individuals in the new population G1, and update the elite external set E using the non-dominated sorting method. The rules are as follows:

[0069] (4) If or or , then X i dominates X j .

[0070] (5) If the individual X i in the population G1 is not dominated by other individuals in the population G1 and the elite external set E, then X i is added to the elite external set E.

[0071] (6) If the individual X e in the population G1 dominates the individual X e in the elite external set E, then X e is deleted.

[0072] S4, if the specified number of generations is reached, terminate the program, and output the elite external set E and the corresponding charging pile number P and service price S of each charging station of the elite individual; otherwise, go to step S3.

[0073] The program flow of the M3 crossover and mutation module is shown in Figure 5 : traverse the individuals in the region, randomly select an individual X i from the regional sub-population, and perform crossover on X j :

[0074] (a) Determine the number of stations built by the new individual:

[0075]

[0076] where rand is a random number between 0 and 1; |X i | and |X j | are the numbers of stations built by the individuals X i and X j .

[0077] (b) Determine the station address, record the candidate points of the two individuals, and randomly select |X c | candidate points to build stations;

[0078] (c) Return the individual X c after crossover;

[0079] Then perform mutation:

[0080] (a) if rand>0.5 if rand>0.5, randomly select a candidate point without built station to build station;

[0081] (b) if rand>0.5 if rand>0.5, randomly select a built station to delete;

[0082] (c) return the individual X after mutation cm ;

[0083] After crossover and mutation, the crossover and mutation individual X cm .

[0084] Wherein the specific content of the M4 local search module is: in order to make full use of the charging station space, delete the charging station with low space utilization rate. The space utilization rate of each charging station is calculated If Then delete the charging station i

[0085] The present application has the following beneficial effects:

[0086] The present application can simultaneously optimize the annual profit of the charging station and the user charging waiting time in the electric vehicle charging station site selection problem, quickly and efficiently obtains the Pareto optimal solution, and provides multiple optimal schemes for the electric vehicle charging station construction enterprise to select the charging station site and determine the number of charging piles and service fee price of each charging station. DETAILED DESCRIPTION

[0087] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0088] Figure 1 It is a three-layer optimization method schematic diagram of the electric vehicle charging station site selection planning method based on MOEA / D-M2M of the present application.

[0089] Figure 2 It is a schematic diagram of the upper decision-making process of the electric vehicle charging station site selection planning method based on MOEA / D-M2M of the present application.

[0090] Figure 3 It is a schematic diagram of the individual coding of the electric vehicle charging station site selection planning method based on MOEA / D-M2M of the present application.

[0091] Figure 4 It is a schematic diagram of the crossover and mutation module program of the electric vehicle charging station site selection planning method based on MOEA / D-M2M of the present application.

[0092] Figure 5 It is the method flow diagram for ensuring the size of the sub-population in the region based on the MOEA / D-M2M electric vehicle charging station site planning method of the application.

[0093] Figure 6 It is the Pareto front interface diagram obtained by the MOEA / D-M2M electric vehicle charging station site planning method of the application.

[0094] Figure 7 It is the 1st station building scheme effect diagram in the Pareto optimal solution set obtained by the MOEA / D-M2M electric vehicle charging station site planning method of the application.

[0095] Figure 8 It is the 2nd station building scheme effect diagram in the Pareto optimal solution set obtained by the MOEA / D-M2M electric vehicle charging station site planning method of the application.

[0096] Figure 9 It is the 3rd station building scheme effect diagram in the Pareto optimal solution set obtained by the MOEA / D-M2M electric vehicle charging station site planning method of the application. DETAILED DESCRIPTION

[0097] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0098] Embodiment:

[0099] Referring to Figures 1 to 6 The embodiment relates to an electric vehicle charging station planning method based on multi-objective optimization, and the method comprises the following contents: 1, a multi-objective optimization electric vehicle charging station planning model maximizing annual profit of a charging station and minimizing waiting time is constructed:

[0100] Supposing that there are n candidate station sites, at most one charging station is built at each candidate station site i E I = {1, 2,..., n}, due to objective restrictions such as site, each charging station has an upper limit of buildable charging piles, through investigation, the number of charging piles of a charging station is generally between 5 and 50, and the number of charging piles is set to be between 5 and 50 (including 5 and 50) in the patent when the charging station is planned; supposing that there are m users, each user j E J = {1, 2,..., m} selects at most one charging station, and vector Q = (q1, q2,..., q n) record the upper limit of the number of charging piles that can be built at each candidate site. Vector X = (x1, x2,..., x n ) represents the station building status of each candidate site, where

[0101]

[0102] Vector P = (p1, p2,..., p n ) represents the number of charging piles at each built station candidate point,

[0103]

[0104] Matrix Z = [z ij ] n×m represents the relationship between user j and charging station i,

[0105]

[0106] Based on this, the patent models the electric vehicle charging station planning problem as a multi-objective optimization model that maximizes the annual profit of the charging station and minimizes the user waiting time.

[0107] Maximize the annual profit of the charging station:

[0108]

[0109] where w is the power of a single charging pile; t is the average charging time, i.e. service time; θ is the annual operation and maintenance cost of each charging station as a percentage of total investment; c is the cost of a single charging pile; where p i c is the total cost of investment of each charging station i; vector u i represents the number of daily service users of charging station i, i.e. the total number of users who choose charging station i

[0110]

[0111] is the annual income of charging station i; θx i p i c is the annual operation and maintenance cost of charging station i; is the annual profit of charging station i. s i is the service price of charging station i (yuan / person / hour), which is assumed in this patent to be related to the load of the charging station. The higher the load, the greater the demand for charging, and according to the supply and demand relationship, the higher the service price; conversely, the lower the charging service price. Thus we get

[0112]

[0113] where a is the lower limit of the service price of the charging station, and b is the upper limit of the service price of the charging station. Through investigation, the patent sets a = 0.5 yuan / person / hour, and b = 2.0 yuan / person / hour. Wherein represents the maximum value of the daily service user number under the condition that there are p i charging piles in the charging station i, and the average waiting time of the user is 15 minutes. The patent uses the average waiting time calculation formula of the M / G / K queuing model

[0114]

[0115] The user charging queuing problem minimizes the average waiting time of user charging, and conforms to the M / G / K queuing model of queuing theory. Therefore, according to the average waiting time calculation formula of the M / G / K queuing model, the total user charging average waiting time is calculated as follows:

[0116]

[0117] where V T is the variance of the user charging time obeying the Gaussian distribution; p′ i = u i t;

[0118] The following constraints are met:

[0119]

[0120] Constraint (6) ensures that each user j selects at most one charging station; constraint (7) ensures that user j can only select a built charging station; constraint (9) ensures that the annual profit of each charging station in the station scheme is not less than 0. Since the electric vehicle charging station planning model established in the project is a high-dimensional multi-objective optimization problem, and it is a very difficult problem to optimize all variables, the project divides the multi-objective optimization problem into three sub-problems through step-by-step alternating optimization: user self-determination layer (lower layer), charging pile quantity decision layer of charging station (middle layer), and charging station location decision layer (upper layer). In these sub-problems, the lower layer decision is limited by the higher layer decision, that is, the user can only select a built charging station; the number of charging piles and the service price can only be selected in the candidate location of the charging station to be built. In turn, the lower layer decision will affect the higher layer decision, for example, the user's decision will affect the charging station's revenue and load, thereby affecting the decision of the higher layer problem. The total process is as follows Figure 1 , specifically:

[0121] The upper decision determines the site selection X, and initializes the charging station charging station number P0 as the upper limit of the charging pile that can be built at each candidate point, and the charging station service fee price S0 as 1.25 yuan per person per hour. Through the lower decision, the corresponding Z is obtained; X, P0 and Z are input into the middle decision to obtain the new charging station charging pile number P and the service fee price S; X, P, S and Z are input into the lower decision to obtain Z'; the middle and lower sub-problems are alternately optimized for 30 times to obtain P * , S * and Z * . Return to the upper layer; the upper layer optimizes X according to P * , S * and Z * . The optimization of the model is completed through such repeated cycles.

[0122] The specific content of the lower sub-problem is as follows:

[0123] Under the given middle and upper decisions, i.e., the station site selection X, the charging pile number P of the charging station to be built, and the charging station service fee price S, the user makes a comprehensive consideration of the cost and distance cost through autonomous decision.

[0124]

[0125] Where λ is a balance factor for balancing the distance cost and the cost, and D ij is the distance from the user j to the charging station i. The problem (10) is solved by using the greedy strategy, and each user j is allocated to the charging station i* with the minimum cost for the user j:

[0126]

[0127] Thus, z i*j = 1, and other z ij = 0, to obtain the user allocation matrix Z. The M2 user allocation module is specifically used to solve.

[0128] The specific content of the middle sub-problem is as follows:

[0129] Under the condition of the station site selection X and the user allocation matrix Z, the local search strategy is used to optimize the optimal charging pile number P of each station to maximize the charging station revenue.

[0130]

[0131] Where is the charging station load rate.

[0132] (a) Update the charging pile number: under the condition of the station site selection X and the user allocation matrix Z, solve the problem (12) to find a best such that

[0133]

[0134] wherein is the upper limit of the number of users that can be accommodated per day, when the number of charging piles is and the waiting time of users is not more than 15 minutes.

[0135] (b) updating the service price: calculating the load μ of each charging station, and letting the service price S = a + (b - a x min(μ, 1); obtaining the number of charging piles P corresponding to the station location X and the service price S.

[0136] The specific content of the upper sub-problem is:

[0137] Based on the improved MOEA / D-M2M, the following steps and modules are used to optimize the station location X, and the process is as shown in Figure 2 :

[0138] S1, using M1 initialization module for initialization;

[0139] The specific content of the M1 initialization module is:

[0140] The main idea of MOEA / D-M2M is to decompose the multi-objective optimization problem into multiple multi-objective optimization sub-problems, and solve these sub-problems simultaneously in one run; from the first quadrant limit of the solution space , select K uniformly distributed unit vectors v 1 , …, v K as the center vectors, and divide into K disjoint sub-regions Ω1, …, Ω K ; wherein Ω k , k = 1, …, K

[0141]

[0142] wherein f = (F1, F2) is the objective value vector of the individual, and <f, v d > is the acute angle of f to v d , that is, f ∈ Ω k if and only if the acute angle of v k to f is the smallest among the K center vectors;

[0143] In each sub-region, select L uniformly distributed weight vectors w 1 , …, w L ;

[0144] The station location scheme X is binary coded, as shown in Figure 3 ; each sub-region maintains a sub-population O k; the population G = {P1,..., O k} of size K x L is formed; each station site X is randomly selected. The elite external set

[0145] S2, using the M2 objective function value calculation module, calculate the objective function value of each station site X;

[0146] Wherein the M2 objective function value calculation module: according to formula (1) to calculate the total annual profit of each station site X F1 (X, P, Z) ; according to formula (5) to calculate the user average waiting time of each station site X F2 (X, P, Z).

[0147] S3, traverse each sub-region sub-population, for each individual X,

[0148] (a) using M3 crossover mutation module for crossover mutation;

[0149] (b) again using M4 local search module, delete the low load charging station;

[0150] (c) again using M2 module to calculate the objective function value; the new individual is added to the offspring population H. After traversing K regions, the offspring population H of size K x L is obtained;

[0151] (d) merge the offspring population H and the original population G, to obtain a population R of size 2 x K x L.

[0152] (e) according to formula (14) to distribute the individuals in the population R to K sub-regions, and ensure that the size of each sub-region sub-population is L, and the K new sub-populations constitute a new population G1; the rule for ensuring that the size of each sub-region sub-population is L is as shown in Figure 4 , specifically:

[0153] If the number of individuals assigned to a sub-region N = L, then the N individuals constitute a new sub-population; if N < L, then L-N individuals are randomly selected from the population R to make up L individuals, and the L individuals constitute a new sub-population; if N > L, then the N individuals are sequentially calculated in the weight vector w 1 ,..., w L of the sub-region, and the Chebyshev distance is calculated according to

[0154]

[0155] Wherein, z * = (maxF1, FminF2), the individual with the minimum Chebyshev distance under each weight is selected to join the new sub-population, and finally a sub-population containing L individuals is obtained; in this way, the size of each regional sub-population is ensured to be L.

[0156] (f) Iterate through the individuals of the new population G1, and update the elite outer set E using the non-dominated sorting method. The rules are as follows:

[0157] (7) If or or Then X i DominateX j .

[0158] (8) If individual X in population G1 i If an individual is not dominated by other individuals in population G1 and the elite outer set E, then it joins the elite outer set E.

[0159] (9) If individual X in population G1 i Individual X that dominates the elite external set E e Then delete X. e .

[0160] S4. If the specified algebra is reached, terminate the program and output the elite external set e and the number of charging piles P and service fee price S of each charging station corresponding to the elite individual; otherwise, go to step S3.

[0161] The program flow for the M3 crossover and mutation module is as follows: Figure 5 As shown: Traversing the individuals in the region, for individual X i An individual X is randomly selected from the regional subpopulation. j First, perform a crossover:

[0162] (a) Determine the number of new sites to be created for each individual:

[0163]

[0164] Where rand is a random number between 0 and 1; |X i |,|X j |For individual X i X j The number of websites built.

[0165] (b) Determine the site location, record the candidate sites for the two individuals, and randomly select |X| from them. c | Build a website at each candidate site;

[0166] (c) Return the crossover result for individual X c ;

[0167] Further mutation:

[0168] (a) If rand > 0.5, randomly select a candidate point that has not yet been built and build the site;

[0169] (b) if rand> 0.5 if rand> 0.5, randomly select one built site to delete;

[0170] (c) return the mutated individual X cm ;

[0171] After crossover mutation, the crossover mutation individual X cm .

[0172] Wherein the M4 local search module specific content is: in order to make full use of the charging station space, delete the space utilization rate of charging station. Calculate the space utilization rate of each charging station If Then delete the charging station i

[0173] When reaching the specified algebra, the program terminates, and the elite population external set E is output, and the multi-objective optimization Pareto optimal solution set is obtained. As shown in Figure 6 , it is the Pareto frontier interface of the two objectives of simultaneously optimizing the internal rate of return of investment and the average waiting time of users after running the program for 150 generations; the final output scheme includes selecting the candidate point X of building station, the number of charging piles P and the service fee price S of each charging station, and the total average waiting time of users and the annual profit of charging station.

[0174] Wherein the scheme 1 effect diagram in the Pareto optimal solution set is as shown in Figure 7 ; the scheme 2 effect diagram is as shown in Figure 8 ; the scheme 9 effect diagram is as shown in Figure 9 ; the data on the right side of the charging station icon is as follows:

[0175] 1. The service fee price of the station (yuan / person / hour)

[0176] 2. The built charging pile / charging pile upper limit (piece)

[0177] 3. The average waiting time of the station (min)

[0178] 4. The annual profit of the station (ten thousand yuan)

[0179] The blue dot represents the demand point, and the size of the dot represents the number of users contained in the demand. The charging station icon represents the building site.

[0180] The above disclosure is only one preferred embodiment of the present application, and of course cannot limit the scope of the right of the present application, so the equivalent changes made according to the claims of the present application still belong to the scope covered by the present application.

Claims

1. A method for planning electric vehicle charging stations based on multi-objective optimization, characterized in that, The content includes: (1) A multi-objective optimization electric vehicle charging station planning model is constructed, which maximizes the annual profit of charging stations and minimizes the user waiting time; (2) Through comprehensive consideration of cost and distance cost, users select the charging station with the minimum cost for charging through the greedy strategy; (3) Through the design of an efficient local search strategy, the optimal number of charging piles and service prices of each station are optimized; (4) The improved MOEA / D-M2M algorithm is used to optimize the model; The content (1) is specifically: Assume that there are n candidate station sites, and each candidate station site i∈I={1, 2, …, n} can build at most one charging station, and each charging station has an upper limit of the number of charging piles, and the number of charging piles of the charging station is between 5 and 50, and the upper limit of the number of charging piles is set to be between 5 and 50 when the charging station is planned; assume that there are m users, and each user j∈J={1, 2, …, m} selects at most one charging station, and the vector Q=(q1, q2, …, q n ) records the upper limit of the number of charging piles that can be built for each candidate station site; Vector X = (x1, x2,..., xN) represents the station building state of each candidate station site, wherein n x1 = 1 if the first candidate station site is built, otherwise x1 = 0 Vector P = (p1, p2,..., pn) represents the number of charging piles of each candidate station n} represents the number of charging piles of each candidate station, Matrix Z = [z ij ] n×m represents the relationship of user j with charging station i, Maximize the annual profit of charging stations: where w is the power of a single charging pile; t is the average charging time, i.e., service time; θ is the annual operation and maintenance cost of each charging station as a percentage of total investment; c is the cost of a single charging pile; where p i c is the total cost of investment of each charging station i; vector u i represents the number of daily service users of charging station i; is the annual revenue of charging station i; θx i p i c is the annual operation and maintenance cost of charging station i; is the annual profit of charging station i; s i is the service price of charging station i, assuming that the charging service price is related to the load of the charging station, the higher the load means the greater the charging demand, and according to the supply and demand relationship, the service price is also higher; On the contrary, the lower the charging service price is; thereby obtaining wherein a is the lower limit of the service fee price of the charging station, and b is the upper limit of the service fee price of the charging station; it is assumed that a = 0.5 yuan / person / hour, and b = 2.0 yuan / person / hour; wherein represents the maximum value of daily service users under the condition that there are p i charging piles at the charging station i, and the average waiting time of users is the longest for 15 minutes. Using the average waiting time calculation formula of the M / G / K queuing model Minimize the average waiting time of user charging, and the user charging queuing problem conforms to the M / G / K queuing model of queuing theory. According to the average waiting time calculation formula of the M / G / K queuing model, the total user charging average waiting time is calculated: where V T is the variance of the user charging times that are Gaussian distributed; p' i = u i t; Satisfy the following constraints: Among them, constraint (6) ensures that each user j selects at most one charging station; constraint (7) ensures that user j can only select the built charging station; constraint (9) ensures that the annual profit of each charging station in the station building scheme is not less than 0; Through the step-by-step alternating optimization method, the multi-objective optimization problem is divided into three layers and three sub-problems: user autonomous decision-making layer, charging pile number decision-making layer of charging station, and charging station location decision-making layer.

2. The multi-objective optimization based electric vehicle charging station planning method of claim 1, wherein, The specific content of the user autonomous decision-making layer sub-problem is: Under the given middle and upper layer decision-making, that is, the station building site X, the charging pile number P of the charging station, and the charging station service price S, the user comprehensively considers the cost and distance cost through autonomous decision-making; where λ is a balancing factor to balance the distance cost and the cost of electricity, D ij is the distance of user j to charging station i; the problem (10) is solved using a greedy strategy, assigning each user j to the charging station i* that has the minimum cost for user j: Thus, z i*j = 0, other z ij = 0, obtaining the allocation matrix Z of the users.

3. The method of claim 1, wherein, The specific content of the charging pile number decision-making layer sub-problem is: Under the condition of station building site X and user distribution matrix Z, the optimal charging pile number P of each station is optimized using the local search strategy to maximize the charging station revenue; wherein is the charging station load rate; (a) Update the number of charging piles: under the condition of the station site selection X and the user's allocation matrix Z, solve the problem (12) to find a best such that Wherein The upper limit of the number of users that can be accommodated when the number of charging piles is 15 minutes, the upper limit of the number of users that can be accommodated. (b) Update the service price: calculate the load μ of each charging station, and let the service price S=a+(b-a)×min(μ,1); obtain the charging pile number P and service price S corresponding to the station building site X.

4. The method of claim 1, wherein, The specific content of the charging station location decision-making layer sub-problem is: Based on the improved MOEA / D-M2M, the following steps and modules are used to optimize the station building site X, and the process is as follows: S1, initialize using M1 initialization module; wherein the specific content of M1 initialization module is: MOEA / D-M2M is to decompose a multi-objective optimization problem into multiple multi-objective optimization sub-problems, and solve these sub-problems simultaneously in one run; K evenly distributed unit vectors v are selected from the first quadrant of the solution space 1 , …, v K as the center vectors, and the solution space is divided into K disjoint sub-regions Ω1, …, Ω according to the K center vectors; wherein Ω k , k = 1, …, K k ; where f = (F1, F2) is the objective value vector of the individual, <f, v d is the acute angle of f to v d , i.e. f ∈ Ω k iff among the K center vectors, v k has the smallest acute angle with f. Select L uniformly distributed weight vectors w in each sub-region 1 , …, w L Binary encode the site selection scheme X; each sub-region maintains a sub-population O of size L k ; the population G = {O1, …, O k} of size K × L is composed; each site of the site selection X is randomly initialized; and the elite external set is initialized S2, calculate the objective function value of each station building site X using M2 objective function value calculation module; Wherein M2 objective function value calculation module: calculate the total annual profit F1(X,P,Z) of each station building site X according to formula (1); calculate the user average waiting time F2(X,P,Z) of each station building site X according to formula (5); S3, traverse each sub-population of each sub-region, and perform the following steps on each individual X: (a) use M3 crossover mutation module for crossover mutation; (b) use M4 local search module to delete the charging stations with low load again; (c) use M2 module to calculate the objective function value again; add the new individual to the offspring population H; after traversing K regions, the offspring population H with a population size of K×L is obtained; (d) Merge the sub-population H and the original population G to obtain a population R with a population size of 2 x K x L; (e) Distribute the individuals in the population R to K sub-regions according to the formula (14), and ensure that the size of each sub-region sub-population is L, and the K new sub-populations constitute a new population G1; the rule for ensuring that the size of each sub-region sub-population is L is as follows: If the number of individuals assigned to a sub-region is N = L, then the N individuals form a new sub-population; if N < L, then L - N individuals are randomly selected from the population R to make up L individuals, and the L individuals form a new sub-population; if N > L, then the N individuals are sequentially multiplied by the weight vector w 1 , …, w L The Chebyshev distance is calculated below, and the individual with the minimum distance is selected according to where z * = (maxF1, minF2)), select the individual with the minimum Chebyshev distance under each weight to join the new subpopulation, and finally obtain a subpopulation containing L individuals; (f) Traverse the individuals in the new population G1, and update the elite external set E using the non-dominated sorting method; the rule is as follows: (1) if and or and or and then X i dominates X j ; (2) If individual X in population G1 i is not dominated by other individuals in population G1 and elite external set E, then join elite external set E; (3) if individual X in population G1 i dominates individual X in elite external set E e then remove X e ; S4, if the specified number of generations is reached, terminate the program, and output the elite external set E and the number P of charging piles and the service price S of each charging station corresponding to the elite individuals; otherwise, go to step S3; The M3 crossover variation module program flow is: traversing the individuals in the region, selecting an individual X i from the region sub-population at random j First, crossover: (a) Determine the number of stations to be built for the new individual: where rand is a random number between 0 and 1 ; |X i | is the number of websites of individual X j | is the number of websites of individual X i | is the number of websites of individual X j | is the number of websites of individual X (b) determining the site address, recording the two individual's built site candidate points, randomly selecting one of the |X c | candidate points to build the site; (c) returning the individual X after crossing c ; Then, perform mutation: (a) If rand>0.5, randomly select an un-built candidate point to build a station; (b) If rand>0.5, randomly select a built station to delete; (c) returning the mutated individual X cm ; After the crossover mutation, the crossover mutation individual X is obtained cm .

5. The method of claim 4, wherein, The specific content of the M4 local search module is: in order to make full use of the space of the charging station, delete the charging station with low space utilization; calculate the space utilization of each charging station If Then delete the charging station i.

Images