Charging station planning method based on charging demand and power distribution network development

Abstract

A charging station planning method based on charging requirements and power distribution network development comprises the following steps: on the basis of a target node, determining a to-be-planned network frame model, establishing a double-layer planning model, when a network frame is planned on the upper layer, generating a power distribution network frame planning scheme for the network frame by using a network structure rapid generation method based on a broken circle method, and improving the solving speed; the line bearing capacity of the power distribution network is comprehensively considered during site selection and sizing of the electric vehicle charging station, so that the income of the charging station and the benefit of an EV user are balanced on the premise that the feeder line bearing capacity of the power distribution network is not exceeded, and a win-win situation is realized. And finally, solving the model based on an improved particle swarm algorithm of a single-dimensional search volume probability variation method, and finally obtaining a charging station and net rack planning scheme. According to the method, the reasonability of locating and sizing of the charging station can be ensured in the planning process, the development of a power distribution network frame can be comprehensively considered, and the solving efficiency is improved by using the mutation particle swarm optimization.

Description

Technical Field

[0001] The present invention relates to the field of new energy technology, in particular to the field of electric vehicle charging station technology, and specifically to double-layer planning of electric vehicle charging stations and distribution network frames based on mutant particle swarms. Background Art

[0002] Electric vehicles (EVs) have the advantages of energy saving, environmental protection and low maintenance costs, and a complete and reasonable grid planning scheme is the cornerstone and guarantee for the development of the electric vehicle industry. The planning of distribution grids and electric vehicle charging stations is not only related to the efficient charging of electric vehicles and user satisfaction, but also directly affects the stable operation and economic benefits of the distribution network. Through double-layer planning, the location and capacity of charging stations can be reasonably configured without exceeding the carrying capacity of the distribution network, which can improve the reliability of the power grid. Research on the double-layer planning of distribution grids and electric vehicle charging stations has important theoretical significance and practical value.

[0003] The document "Planning of Electric Vehicle Charging Stations with Comprehensive Consideration of Charging Demand and Distribution Network Acceptance Capacity" uses the entropy weight-corrected hierarchical analysis method to evaluate the distribution network's acceptance capacity for charging station access solutions, and achieves coordinated optimization of the distribution network and charging facilities to ensure that the planning results not only meet the charging needs of electric vehicles, but also meet the operational safety and economic requirements of the distribution network.

[0004] The above existing technologies focus on evaluating the carrying capacity and operating conditions of the existing distribution network under different charging station distribution schemes. However, they fail to consider the interaction between charging stations and distribution networks in the planning process, resulting in the planning of the distribution network topology structure being unable to adapt to the growth of EV users; the planning of charging stations is limited by the fixed distribution network and is difficult to meet the needs of EV users. Summary of the invention

[0005] The purpose of the present invention is to solve the technical problems that the layout of electric vehicle charging stations and the topology of distribution network in the prior art do not fully consider the development of each other in the planning link, resulting in unreasonable charging station planning results and node voltage exceeding the limit at the distribution network node, and propose a charging station planning method based on charging demand and distribution network development.

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

[0007] A charging station planning method based on charging demand and distribution network development includes the following steps:

[0008] Step 1: Establish a road network - distribution network coupling model, which includes a road network model and a distribution network model. In the road network model, the nodes of the distribution network are directly used to represent the nodes of the road network. On the basis of the basic structure of the road network model, a road weight model and a set of time periods are added to generate a mathematical model of the road network topology.

[0009] Step 2: Establish a bilevel programming model, which includes an upper-level programming model and a lower-level programming model. The upper level is the distribution network framework planning layer with the optimal distribution network revenue as the objective function, and the lower level is the charging station location and capacity planning layer with the optimal charging station revenue as the objective function.

[0010] Step 3: According to the bilevel programming model established in Step 2, use the improved particle swarm optimization algorithm with the single-dimensional search volume probability mutation method to solve the bilevel programming model and obtain the charging station and network framework planning scheme.

[0011] In Step 1, when obtaining the road weight model, the following steps are adopted:

[0012] Step 1.1: Obtain the road impedance model:

[0013]

[0014] In the formula: degree of saturation Q is the road traffic flow, C is the road capacity; t0 is the travel time with zero flow; α and β are impedance influence factors, where both the road traffic flow and the road capacity follow a normal distribution, Indicates the road impedance when the degree of saturation S satisfies 0 < S ≤

[0015] 1.0, Indicates the road impedance when the degree of saturation S satisfies 1.0 ≤ S ≤ 2.0;

[0016] Step 1.2: Obtain the node impedance model:

[0017]

[0018] In the formula: c is the traffic signal cycle; γ is the green signal ratio (i.e., the ratio of the green light time to the entire signal cycle); q is the arrival rate of road vehicles per minute, Indicates the node impedance when the degree of saturation S satisfies 0 < S ≤ 1.0, Indicates the node impedance when the degree of saturation S > 0.6;

[0019] Step 1.3: The road weight model can be obtained by combining the models obtained in Step 1.1 and Step 1.2. w ij Represents the road weight of road ij:

[0020]

[0021] In the formula: represents the road impedance when the saturation S satisfies 0 < S ≤ 1.0; represents the road impedance when the saturation S satisfies 1.0 ≤ S ≤ 2.0; represents the node impedance when the saturation S satisfies 0 < S ≤ 1.0; represents the node impedance when the saturation S satisfies S > 0.6.

[0022] In Step2, the upper-level planning model is specifically as follows:

[0023] Objective function:

[0024] maxF = -C inv -C main +C trans

[0025] In the formula: C inv is the equivalent annual value of the distribution network expansion cost, including new lines and tie lines; C main is the operation and maintenance cost of the distribution network; C trans is the revenue from selling electricity of the distribution network;

[0026] Distribution network expansion cost:

[0027]

[0028] In the formula: d is the discount rate for grid construction; m is the service life of the line; i and j are distribution network nodes; x i,j is a 0-1 variable, indicating whether the line between nodes i and j is built. Take 1 when built and 0 when not built (the line construction situation is obtained by the loop-breaking method); L i,j is the length of line ij; C line is the construction cost per unit length of the line;

[0029] L i,j = 10W i,j

[0030] Operation and maintenance cost:

[0031] C main = μC inv

[0032] In the formula: μ is the proportion of operation and maintenance cost;

[0033] Revenue from selling electricity of the distribution network:

[0034]

[0035] In the formula: N node is the number of nodes; C sold,tP is the electricity price at time t; i,t is the charging power of the charging station at node i at time t, corresponding to the power of the lower-level charging station, which is determined by the number of charging piles built at the charging station at node i and the number of EV users going to the node to charge. The specific construction node of the charging station is obtained by the particle swarm algorithm calculation;

[0036] Node voltage constraints:

[0037] U bmin ≤U b ≤U bmax

[0038] U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

[0039] In Step 2, the lower-level planning model aims to maximize the total revenue of the charging station consisting of construction cost, operation and maintenance cost, user cost, network loss cost and electricity sales revenue, and optimizes the construction location and capacity of the charging station with the road distance between the charging stations and the number of charging piles and the location of the charging station obtained in Step 1.3 as constraints; the lower-level planning model is specifically as follows:

[0040] Objective function:

[0041] maxF=-C I -C EV -C L +C B

[0042] Where: C I is the average annual investment cost of the kth charging station; C L is the annual network loss cost of the charging station, C EV and C B They are annual user costs and revenue from electricity sales at charging stations;

[0043] Annual investment cost:

[0044]

[0045] C M,i,k =β k C J,i,k

[0046] N C is the total number of charging stations to be built, which is determined by the number of nodes with charging demand; C is the node set of the alternative site for the charging station. It is assumed that except for node 1, all other nodes in the distribution network are alternative sites. J,i,k , C M,i,k and C o,i,kare the construction cost, operation and maintenance cost and other costs of the kth charging station at the i-th node; r0 is the discount rate; n0 is the operating life of the charging station; c ch is the unit price of the charging pile; N ch,i,k , A j,k , C f,i,k and β k are the number of charging piles at the kth charging station at the i-th node, the area occupied by a single charging pile, the land price per unit area, and the proportion of operation and maintenance cost to construction cost;

[0047] Annual user cost:

[0048] C EV =365C j

[0049] Where: C j is the user cost per day;

[0050] The daily user cost is composed of the travel energy consumption, travel time and queue time costs:

[0051]

[0052] Where: c ev,t,k 、N i,c are the user cost of charging at the kth charging station at time t and the number of nodes with charging demand; t n,k,d ,t n,k,w are the average travel time and average queue waiting time (t n,k,w According to the subsequent formula T w (get); c time and c n,k,e are the user's unit time cost and the average power consumption from the nth charging demand node to the kth charging station (obtained by combining the road weight with the electric vehicle energy consumption model. Users choose charging stations according to the principle of proximity. When going to the charging station, the shortest path, i.e. the path with the smallest sum of road weights, is chosen by default); P b,t is the unit electricity price of the charging station at time t;

[0053] Electric vehicle energy consumption model:

[0054]

[0055] c n,k,e =w ij X

[0056] Where: w ij The road v where the vehicle is located generated in Step 1.3 ij The road weight at , Z is the average vehicle speed;

[0057] Annual network loss cost of charging station:

[0058]

[0059] Where: P L,t is the network loss at time t, P c is the cost coefficient.

[0060]

[0061] Where: I ij is the absolute value of the branch current between nodes i and j; Z ij is the branch impedance between nodes i and j;

[0062] Revenue from electricity sales at charging stations:

[0063]

[0064] P b,t is the peak / flat / valley electricity price of the charging station at time t, P i,t is the charging power at node i at time t; (Note: P i,t The number of charging piles built at the charging station at the i node and the number of EV users going to the node to charge are determined. The specific construction nodes and capacity of the charging station are obtained by particle swarm algorithm calculation)

[0065] If the process of vehicles arriving at each charging station follows Poisson distribution, according to the queuing theory system average waiting time T w for:

[0066]

[0067] N ch,k is the number of charging piles in the charging station; ε is the average charging time of a single charging pile; E bt The charging demand for a single EV is obtained by combining Table 3 with Table 4; P ch is the charging power of a single charging pile (see Table 5 for details); g is the average number of EV users arriving per unit time, which is the reciprocal of ε; ρ is the utilization rate of the charging pile (ρ<1);

[0068] Distance Constraint:

[0069] u s ≥u min

[0070] A distance that is too small will cause traffic congestion during peak charging periods, so the distance between two charging stations needs to be limited; where the distance between any two charging stations is greater than or equal to the minimum distance limit;

[0071] Restrictions on the number of charging piles:

[0072]

[0073] P ch The charging power of the charging pile; E max is the maximum value among all charging demand nodes at the same time; this constraint is the lower limit of the number of charging piles to be built, which can reduce the solution set of charging station capacity;

[0074] Charging station construction node constraints:

[0075]

[0076] a i is a 0-1 variable, which takes 1 when a charging station is built at node i and takes 0 when no charging station is built at node i; A alt 、N a are the charging station candidate point set and the number of candidate points, N C is the total number of charging stations to be built;

[0077] Feeder load factor constraints:

[0078]

[0079] S L and S L,max are the actual load and rated capacity of feeder i respectively;

[0080] Line load factor constraints:

[0081]

[0082] S L,i and S L,i,max are the actual load and rated capacity of line i respectively;

[0083] Node voltage constraints:

[0084] U bmin ≤U b ≤U bmax

[0085] U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

[0086] In Step 3, the following steps are included:

[0087] Step 3.1 Encode the location of the EV charging station

[0088] The number of particles in the particle swarm algorithm is set to T (proposed to be set to 100), the number of candidate points for EV charging stations is set to D; the initial position of the particle is X n ={X n1 ,X n2 ,…,X ni}, n = 1, 2, ..., T; corresponding to the charging station planning, X ni represents the construction status of the EV candidate distribution network node i of particle n (X ni =1 means to build an EV charging station at this candidate point, X ni =0 means not to build an EV charging station at this candidate point, X n The number of elements in the set equal to 1 must be equal to D), X n Each quantity in will change with the iteration until convergence; the initial velocity of the particle is V n = {V n1 ,V n2 ,…,V ni},V ni It represents the change of the objective function of the planning model in a single iteration during the iterative calculation of particle n, V n Each quantity in will change with the iteration until convergence; best,n For a single particle X n The historical optimal point after iteration, g best is the global optimal point iterated by all particles (corresponding to the global optimal configuration of charging station location and capacity); the particle update speed and position update formula are:

[0089] V n,r+1 =ωV n，r +c1r1(P best,n -X n,r )+c2r2(g best -X n，r )

[0090] X n，r+1 =X n,r +V n,r+1

[0091] Where: ω is the inertia weight; c1 and c2 are the self-learning coefficient and social learning coefficient respectively; r1 and r2 are mean random perturbations, which are used to simulate small perturbations caused by nature to increase the diversity of particles (Note: r is the number of iterations);

[0092] An improved particle swarm algorithm based on the probabilistic mutation method of single-dimensional search volume is used to divide the dimensions of a single particle into multiple intervals. The historical search volume of the particle in each interval is calculated as the probability of selection mutation. At the same time as the algorithm iteration, the inertia weight and particle learning coefficient are updated. This improved algorithm has been proven to improve particle diversity, avoid premature particle maturation and accelerate convergence.

[0093] Step 3.2 Update the inertia weight, self-learning coefficient and social learning coefficient in the particle update speed and position update formula

[0094] The inertia weight is updated in a linear decreasing form as the number of iterations increases. The update formula is as follows:

[0095]

[0096] Where: max The initial weight of the particle is set to 0.9; ω min Set the final weight of the particle to 0.4; I iter.max is the maximum number of iterations (to be set to 1000); i iter is the current iteration number, and its value is equal to r;

[0097] For the learning coefficient, in the initial iteration, a larger self-learning coefficient and a smaller social learning coefficient are set so that the particles can search throughout the entire space; and in the later iteration, in order to make the search target of the particles tend to the global optimal solution, a smaller self-learning coefficient and a larger social learning coefficient are set. The update formula is as follows:

[0098]

[0099] Where: c1 and c2 are the particle's self-learning coefficient and particle's social learning coefficient respectively, and the rest are constants. After research, it is found that when c1e, c1f, c2e, and c2f are set to 2.5, 0.5, 0.5, and 2.5 respectively, the optimization efficiency is better.

[0100] Step 3.3 Use single-dimensional mutation to prevent the particle swarm algorithm from falling into the local optimal solution too early

[0101] In the particle swarm algorithm, each particle represents a charging station location and capacity configuration scheme, and the dimension of the particle corresponds to the decision variable of the problem; in this specific application, the dimension of the particle refers to the different candidate locations of the charging station and its capacity configuration; the key to this algorithm is that when a particle falls into a local optimal solution too early, the position of this particle in this dimension is mutated so that the particle can continue to find the optimal solution; the steps are as follows:

[0102] 1) Initialization parameters: First, initialize the number of particles N, the number of dimensions D (i.e. the number of candidate points), and the initial position X of each particle. n and speed V n ; Initialize the inertia weight ω and learning coefficients c1 and c2;

[0103] 2) Update the position and velocity of the particle: Use the formula to update the velocity and position of the particle;

[0104] 3) Calculate the fitness value: For each particle, calculate its fitness value and update the local optimal value and the global optimal value;

[0105] 4) Divide the intervals: Divide each dimension of the particle (i.e., candidate points) into several intervals; assume that the search limit of each interval is A max , that is, each interval can have at most A max Particles to search;

[0106] 5) Detect search volume: When particle X n When the fth dimension of is about to be updated to a certain interval, check the search volume of the interval; if there is already A in the interval max If there are particles, the position of the particle in this dimension will not be updated to this interval, but will mutate to other intervals according to probability;

[0107] 6) Calculate the probability of mutation: Determine the probability of mutation based on the historical search volume of each interval. Specifically, the larger the historical search volume of an interval, the smaller the probability of mutation to that interval. The probability of mutation to each interval can be expressed as:

[0108]

[0109] Among them, V fe is the search volume of the e-th interval in dimension f, Aims to calculate the total search volume for all intervals of dimension f;

[0110] 7) Update the position of the mutated particle: According to the mutation probability, randomly select a new interval and update the position of the particle in the fth dimension to this interval;

[0111] 8) Iteration loop: Repeat steps 2 to 7 until the maximum number of iterations is reached or the fitness value converges to obtain the optimal charging station and grid planning scheme.

[0112] Compared with the prior art, the present invention has the following technical effects:

[0113] 1) The present invention can generate a road network structure diagram that can use weights to reflect real-time traffic conditions;

[0114] 2) The present invention proposes a charging station planning method based on charging demand and distribution network development, which can optimize the distribution network structure while considering the charging demand, ensuring efficient charging of electric vehicles and stable operation of the distribution network;

[0115] 3) In the upper-level planning, the present invention uses the circle-breaking method to simplify and optimize the distribution network frame, thereby minimizing the construction cost and operation and maintenance cost of the distribution network, while maximizing the electricity sales revenue of the distribution network, thereby improving the economic benefits of the distribution network;

[0116] 4) The present invention uses an improved variant particle swarm algorithm to optimize the location and capacity of charging stations. The algorithm can effectively avoid falling into local optimality, improve the solution efficiency and adaptability of the planning scheme, and ensure the rationality of the location of the charging station and the optimal configuration of the number of charging piles;

[0117] 5) During the planning process, the breaking circle method is used to quickly generate grid planning schemes, and electric vehicle charging loads are introduced into the distribution network grid planning link; when selecting the site and determining the capacity of EV charging stations, the voltage offset of the distribution network nodes is comprehensively considered to balance the benefits of charging stations and the interests of EV users to achieve a win-win situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0119] Figure 1 This is a topological diagram of the development-type distribution network to be planned, where the dotted lines are lines that can be put into construction;

[0120] Figure 2 Basic transportation network topology map;

[0121] Figure 3 Traffic network diagram for calibrating road network weights;

[0122] Figure 4 Flow chart of the present invention;

[0123] Figure 5 This is the operation steps of the circle breaking method;

[0124] Figure 6 This is a schematic diagram of the double-layer planning results of electric vehicle charging stations and distribution network;

[0125] Figure 7 Schematic diagram of voltage deviation of distribution network nodes corresponding to the double-layer planning results;

[0126] Figure 8 A schematic diagram of voltage offset at distribution network nodes when only charging stations are planned using a fixed distribution network topology;

[0127] Fig. 9 Comparison chart of the operation of the improved PSO algorithm.

[0128] Fig.10 Fixed distribution network topology structure diagram, in which there are no lines that can be built and put into operation; Detailed implementation manners

[0129] As Figure 4 shown, a double-layer planning method for an electric vehicle charging station and a distribution network framework based on a mutated particle swarm includes the following steps:

[0130] Step1. Establish a road network - distribution network coupling model, which includes a road network model and a distribution network model; the distribution network model is the IEEE33-node standard distribution network model, and its structure is as Figure 1 shown; the road network model directly uses the distribution network nodes to represent the road network nodes, and the basic structure of the road network model is as Figure 2 shown. On this basis structure, a road weight model and a time period set are added to generate a road network topology mathematical model, and the road distance is directly represented by the road weight.

[0131] The generation steps of the road network topology mathematical model are as follows:

[0132]

[0133] In the formula: G is the road network topology mathematical model; V represents the set of all nodes in the road network; E represents the set of all roads in the road network; W is the set of road weights;

[0134] The road weight w ij is the road weight at road v ij and represents the road travel cost, which is composed of the road impedance and the node impedance. The electric vehicles studied in this paper mainly target the urban internal roads, and in the urban road network, traffic lights are generally set for control. Therefore, the concepts of road impedance and node impedance will be introduced when calculating the road weight, and the road weight model is composed of the road impedance model and the node impedance model.

[0135] In this paper, the saturation S index is used to describe different urban traffic conditions: unobstructed (0 < S ≤ 0.6), slow (0.6 < S ≤ 0.8), congested (0.8 < S ≤ 1.0), and severely congested (1.0 < S ≤ 2.0). Different saturations can be obtained according to different road intersections and road capacities, and then different road impedances and node impedances can be obtained.

[0136] Step1.1 Road impedance model

[0137]

[0138] In the formula: saturation Q is the road traffic flow, C is the road capacity; t0 is the zero-flow travel time; α and β are impedance influence factors. Among them, both the road traffic flow and the road capacity follow a normal distribution. It represents the road impedance when the saturation S satisfies 0 < S ≤ 1.0. The relevant data of the road impedance when the saturation S satisfies 1.0 ≤ S ≤ 2.0 are shown in Table 1 below.

[0139] Step1.2 Node Impedance Model

[0140]

[0141] In the formula: c is the traffic signal cycle; γ is the green ratio (i.e., the proportion of the green light time in the whole signal cycle); q is the arrival rate of road vehicles per minute. It represents the node impedance when the saturation S satisfies 0 < S ≤ 1.0. The relevant data of the node impedance when the saturation S > 0.6 are shown in Table 1 below.

[0142] Combining Equation (3) and Equation (4) can obtain the complete road impedance model.

[0143] Road traffic flow (Q) <![CDATA[Normal distribution N(1800, 150 2 )]]> Road capacity (C) <![CDATA[Normal distribution N(2100, 200 2 )]]> <![CDATA[Zero flow travel time (t0)]]> <![CDATA[t0 = 10 (minutes)]]> Impedance influence factors (α, β) α＝1.3、β＝1.2 Traffic light signal cycle (c) c = 30 (seconds) Green letter ratio (γ) 0.7 Road vehicle arrival rate per minute (q) 2(vehicles / minute)

[0144] Table 1 Road Information Table Step1.3 Combining Step1.1 and Step1.2 can obtain the road weight model

[0145]

[0146] In the formula: It represents the road impedance when the saturation S satisfies 0 < S ≤ 1.0. It represents the road impedance when the saturation S satisfies 1.0 ≤ S ≤ 2.0. It represents the node impedance when the saturation S satisfies 0 < S ≤ 1.0. It represents the node impedance when the saturation S > 0.6. The model calculation results are shown in Figure 3 .

[0147] Step2. The basic technical framework is as Figure 4 shown. Based on the improved IEEE33 nodes, the grid model to be planned is confirmed. A two-layer planning model of the distribution network - charging station is established. The upper layer is the distribution network grid planning layer, which optimizes the grid construction plan with the goal of maximizing the revenue of the distribution company; the lower layer aims to minimize the comprehensive cost of the charging station construction and operation and maintenance costs, network loss costs, power purchase costs, and user costs, and optimizes the construction location and capacity of the charging station with the road distance between charging stations (obtained using the road weight in Step1.3), the number of charging piles, and the charging station location as constraints.

[0148] Step 2.1 Upper level planning model:

[0149] 2.1.1 Objective Function

[0150] When constructing the grid, the total revenue of the distribution company, which is composed of operation and maintenance costs, expansion costs, and user transaction revenue, is considered, and the maximum total revenue is calculated as follows:

[0151] maxF=-C inv -C main +C trans

[0152] Where: C inv is the equivalent annual value of the cost of expanding the distribution network, including new lines and interconnection lines; C main is the operation and maintenance cost of the distribution network; C trans Revenue from electricity sales to distribution networks

[0153] Expansion costs of distribution network:

[0154]

[0155] Where: d is the grid construction discount rate; m is the service life of the line; i and j are distribution network nodes; x i,j is a 0-1 variable, indicating whether the line between nodes i and j is under construction. It takes 1 if it is under construction and 0 if it is not (the line construction status is obtained by breaking the circle method); L i,j is the length of line ij (here the weight of the road between nodes i and j is used to calculate it, unit: meter); C line is the construction cost per unit length of line. The specific parameters are shown in Table 2.

[0156] L i，j =10W i，j

[0157] Operation and maintenance costs:

[0158] C main =μC inv

[0159] Where: μ is the proportion of operation and maintenance costs.

[0160] Revenue from electricity sales in the distribution network:

[0161]

[0162] Where: N node is the number of nodes; C sold,t P is the electricity price at time t (the electricity price is usually divided into peak / flat / valley time according to the change of electricity load), and the specific parameters are shown in Table 2; i,tis the charging power of the charging station at node i at time t, corresponding to the power of the lower-level charging station, which is determined by the number of charging piles built at the charging station at node i and the number of EV users going to the node to charge. The specific construction node of the charging station is obtained by the particle swarm algorithm calculation;

[0163] 2.1.2 Constraints

[0164] Node voltage offset constraints:

[0165] U bmin ≤U b ≤U bmax

[0166] U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

[0167] 2.1.3 Breaking the Circle Method

[0168] 1. Simplify the distribution network structure (topology structure shown in Table 6), remove invalid branches (i.e., lines that do not form a ring network), merge equivalent branches (i.e., if there is no bifurcated line between two nodes, the nodes between the two nodes can be omitted and merged into one line), and obtain a simplified network equivalent to the original distribution network topology structure.

[0169] Second: Use the breaking circle method to simplify the network. According to the number of feeders N, calculate the number of special loops that connect two different feeders and do not intersect each other at the end of breaking the circle. The formula is: The next steps are as follows Figure 5 Shown

[0170] Grid Construction Discount Rate 0.12 Line service life 20 years Construction cost per unit length of line 3,000 yuan / meter Electricity sales price (peak / flat / valley time) 1.2 / 0.8 / 0.4 (yuan / (kW·h))

[0171] Table 2 Parameters related to grid planning model

[0172] Step 2.2 Lower layer model

[0173] 2.2.1 Objective Function

[0174] The spatiotemporal distribution of the charging demand of electric vehicles in the distribution network is shown in Table 3. Assuming that all vehicles that need to be charged are located at the node where the charging demand is generated, and start from the node to the charging station for charging, the total number of EVs is 1,000. The number of electric vehicles that need to be charged in each node at each time is shown in Table 4. Combining Table 3 and Table 4, the charging demand of each electric vehicle can be obtained.

[0175]

[0176]

[0177] Table 3. Spatiotemporal distribution of charging demand over 24 hours

[0178]

[0179]

[0180] Table 4 Spatiotemporal distribution of the number of vehicles with charging demand

[0181] When planning the site selection and capacity of EV charging stations, the planning objectives of EV charging stations are constructed with the goal of maximizing the total revenue of the charging station, which is composed of construction cost, operation and maintenance cost, user cost, network loss cost and electricity sales revenue:

[0182] maxF=-C I -C EV -C L +C B

[0183] Where: C I is the average annual investment cost of the kth charging station; C L is the annual network loss cost of the charging station, C EV and C B They are annual user cost and charging station electricity sales revenue respectively.

[0184] Annual investment cost:

[0185]

[0186]

[0187] C M,i,K =β k C J,i,k

[0188] N C is the total number of charging stations to be built, which is determined by the number of nodes with charging demand. N is set according to Table 3. C =8; C is the node set of the alternative site for the charging station. It is assumed that except for node 1, all other nodes in the distribution network are alternative sites. J,i,k , C M,i,k and C o,i,k are the construction cost, operation and maintenance cost and other costs of the kth charging station at the i-th node respectively); r0 is the discount rate; n0 is the operating life of the charging station; c ch is the unit price of the charging pile; N ch,i,k , A j,k , C f,i,k and β kThey are the number of charging piles at the kth charging station at the i-th node, the floor area of ​​a single charging pile, the land price per unit area and the operation and maintenance cost conversion factor. The specific parameters are shown in Table 5.

[0189]

[0190]

[0191] Table 5 Charging station construction related parameters

[0192] Annual user cost:

[0193] C EV =365C j

[0194] Where: C j The cost per user per day.

[0195] The daily user cost is composed of the travel energy consumption, travel time and queue time costs:

[0196]

[0197] Where: c ev,t,k 、N i,c are the user cost of charging at the kth charging station at time t and the number of nodes with charging demand (the number is 8 at this time as shown in Table 3); t n,k,d ,t n,k,w are the average travel time and average queue waiting time (t n,k,w According to the subsequent formula T w (get); c time and c n,k,e are the user unit time cost and the average road power consumption from charging demand node y to charging station node k (c n,k,e It is obtained by combining the road weight with the electric vehicle energy consumption model. Users choose charging stations according to the principle of proximity. When going to the charging station, the shortest path is selected by default, that is, the path with the smallest sum of road weights); P b,t is the unit electricity price of the charging station at time t (the electricity price is often divided into peak / flat / valley times in a day according to the changes in electricity load). For specific parameters, see Table 5.

[0198] Electric vehicle energy consumption model:

[0199]

[0200] C n,k,e =w ij X

[0201] Where: w ijThe road v where the vehicle is located generated in Step 1.3 ij The road weight at , Z is the average vehicle speed, and a fixed speed of 40KM / hour is used here.

[0202] Annual network loss cost of charging station:

[0203]

[0204] Among them: P L,t is the network loss at time t, P c is the cost coefficient.

[0205]

[0206] Where: I ij is the absolute value of the branch current between nodes i and j; Z ij is the branch impedance between nodes i and j. The grid parameters are detailed in Table 6. The grid reference voltage is 10KV and the reference power is 10MVA.

[0207]

[0208]

[0209] Table 6 Distribution network grid structure parameters

[0210] Revenue from electricity sales at charging stations:

[0211]

[0212] P b,t is the peak / flat / valley electricity price of the charging station at time t, see Table 5 for details, P i,t is the charging power at node i at time t (determined by the number of charging piles built at node i and the number of EV users going to this node to charge. The specific construction node of the charging station is obtained by particle swarm algorithm calculation).

[0213] If the process of vehicles arriving at each charging station follows Poisson distribution, according to the queuing theory system average waiting time T w for:

[0214]

[0215] N ch,k is the number of charging piles in the charging station; ε is the average charging time of a single charging pile; E bt The charging demand for a single EV is obtained by combining Table 3 with Table 4; P ch is the charging power of a single charging pile (see Table 5 for details); g is the average number of EV users arriving per unit time, which is the reciprocal of ε; ρ is the utilization rate of the charging pile (ρ<1).

[0216] 2.2.2 Constraints

[0217] Distance Constraint:

[0218] u s ≥u min

[0219] A distance that is too small will cause traffic congestion during peak charging periods, so the distance between two charging stations needs to be limited; where u is set to min =1 km.

[0220] Restrictions on the number of charging piles:

[0221]

[0222] P ch The charging power of the charging pile; E max is the maximum value among all charging demand nodes at all times. This constraint is the lower limit of the number of charging piles to be built, which can reduce the solution set of charging station capacity.

[0223] Charging station construction node constraints:

[0224]

[0225] a i is a 0-1 variable, which takes 1 when a charging station is built at node i and takes 0 when no charging station is built at node i; A alt 、N a are the charging station candidate point set and the number of candidate points, N C is the total number of charging stations to be built (a i calculated by particle swarm algorithm).

[0226] Node voltage offset constraints:

[0227] U bmin ≤U b ≤U bmax

[0228] U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

[0229] Step 3: An improved particle swarm algorithm solution model based on the single-dimensional search volume probability mutation method is used to finally obtain the charging station and grid planning scheme.

[0230] Step 3.1 Encode the location of the EV charging station

[0231] The number of particles in the particle swarm algorithm is set to T, and the number of candidate points for EV charging stations is set to D (the number of candidate points is set to 8 according to Table 3). The initial position of the particle is X n ={X n1 ,X n2 ,…,X ni}, n = 1, 2, ..., T (Note: corresponding to the charging station planning, X ni represents the construction status of the EV candidate point of particle n at the distribution network node i, X ni =1 means to build an EV charging station at this candidate point, X ni =0 means not to build an EV charging station at this candidate point, X n The number of elements in the set equal to 1 must be equal to D), X n Each quantity in will change with the iteration until convergence; the initial velocity of the particle is V n = {V n1 ,V n2 ,…,V ni},V ni It represents the change of the objective function of the planning model in a single iteration during the iterative calculation of particle n, V n Each quantity in will change with the iteration until convergence; best,n For a single particle X n The historical optimal point after iteration, g best is the global optimal point iterated by all particles. The update speed and position update formula of particles are:

[0232] V n，r+1 =ωV n,r +c1r1(P best,n -X n，r )+c2r2(g best -X n，r )

[0233] X n,r+1 =X n，r +V n,r+1

[0234] Where: ω is the inertia weight; c1 and c2 are the self-learning coefficients and social learning coefficients respectively; r1 and r2 are mean random perturbations, which are used to simulate small perturbations caused by nature to increase the diversity of particles (Note: r is the number of iterations).

[0235] However, like other global optimization algorithms, the PSO algorithm also has disadvantages in practical applications, such as being prone to falling into local optimal points and slow convergence in the late stage of evolution. Therefore, this paper adopts an improved particle swarm algorithm based on the probabilistic mutation method of single-dimensional search volume, divides each dimension of a single particle into multiple intervals, calculates the historical search volume of the particle in each interval as the probability of selection mutation, and updates the inertia weight and particle learning coefficient while the algorithm is iterating. This improved algorithm has been proven to improve particle diversity, avoid premature particle maturation and accelerate convergence.

[0236] Step 3.2 Update the inertia weight, self-learning coefficient and social learning coefficient in the particle update speed and position update formula

[0237] The inertia weight is updated in a linear decreasing form as the number of iterations increases. The update formula is as follows:

[0238]

[0239] Where: max The initial weight of the particle is set to 0.9; ω min Set the final weight of the particle to 0.4; I iter.max is the maximum number of iterations (to be set to 1000); i iter is the current iteration number, and its value is equal to r.

[0240] For the learning coefficient, in the initial iteration, in order for the particles to search throughout the entire space, a larger self-learning coefficient and a smaller social learning coefficient can be set; in the later iteration, in order to make the search target of the particles tend to the global optimal solution, a smaller self-learning coefficient and a larger social learning coefficient should be set. The update formula is as follows:

[0241]

[0242] In the formula: c1 and c2 are the particle's self-learning coefficient and particle's social learning coefficient respectively, and the rest are constants. 1e , c 1f , c 2e , c 2f When set to 2.5, 0.5, 0.5, and 2.5 respectively, the optimization efficiency is better.

[0243] Step 3.3 Use single-dimensional mutation to prevent the particle swarm algorithm from falling into the local optimal solution too early

[0244] In the particle swarm algorithm, each particle represents a charging station location and capacity configuration scheme, and the dimension of the particle corresponds to the decision variable of the problem. In this specific application, the dimension of the particle refers to the different candidate locations of the charging station and its capacity configuration. The key to this algorithm is that when a particle falls into a local optimal solution too early, the position of this particle in this dimension is mutated so that the particle can continue to find the optimal solution. The steps are as follows:

[0245] 1) Initialization parameters: First, initialize the number of particles N, the number of dimensions D (i.e. the number of candidate points), and the initial position X of each particle. n and speed V n . Initialize the inertia weight ω and the learning coefficients c1 and c2.

[0246] 2) Update the particle's position and velocity: Use the formula to update the particle's velocity and position.

[0247] 3) Calculate the fitness value: For each particle, calculate its fitness value and update the local optimal value and the global optimal value.

[0248] 4) Divide the intervals: Divide each dimension of the particle (i.e., candidate points) into several intervals. Assume that the search limit of each interval is A max , that is, each interval can have at most A max Search for particles.

[0249] 5) Detect search volume: When particle X n When the fth dimension of is about to be updated to a certain interval, check the search volume of the interval. max If there are particles, the position of the particle in this dimension will not be updated to this interval, but will mutate to other intervals according to probability.

[0250] 6) Calculate the probability of mutation: Determine the probability of mutation based on the historical search volume of each interval. Specifically, the larger the historical search volume of the interval, the smaller the probability of mutation to that interval. The probability of mutation to each interval can be expressed as:

[0251]

[0252] Among them, V fe is the search volume of the e-th interval in dimension f, Aims to calculate the total search volume for all intervals in dimension f.

[0253] 7) Update the position of the mutated particle: According to the mutation probability, randomly select a new interval and update the f-th dimension position of the particle to this interval.

[0254] 8) Iteration loop: Repeat steps 2 to 7 until the maximum number of iterations is reached or the fitness value converges to obtain the optimal charging station and grid planning scheme.

[0255] The planning model provided by the present invention is characterized by establishing a network-station double-layer planning model to plan the distribution network grid structure and EV charging piles. In the planning process, the breaking circle method is used to quickly generate a grid planning scheme, and the electric vehicle charging load is introduced into the distribution network grid planning link; when selecting the site and determining the capacity of the EV charging station, the distribution network line carrying capacity is comprehensively considered, so as to balance the charging station revenue and EV user interests without exceeding the distribution network feeder carrying capacity, and achieve a win-win situation.

[0256] like Figure 6 As shown in the figure, the grid planning result obtained by the distribution network planning layer, corresponding to the upper planning model objective function result of 2.43532 million yuan, the black asterisk is the charging station investment location, corresponding to the lower planning model objective function of 1.94314 million yuan, at this time the location and capacity of the charging station can meet the charging demand in Table 3, the investment capacity is shown in Table 7:

[0257] Invest in Node Number of charging piles 3 4 11 5 13 4 18 3 21 4 24 5 28 6 32 5

[0258] Table 7 Charging station capacity

[0259] <![CDATA[Reference voltage U B > 10KV <![CDATA[Reference power S B > 10MVA

[0260] Table 8 IEEE33 voltage level and reference power

[0261] There are also conventional loads in the distribution network that do not change over time, as shown in Table 9:

[0262]

[0263]

[0264] Table 9 Conventional load

[0265] Take the reference voltage as 10KV, the reference power as 10MVA, and superimpose the conventional load and the load generated by the electric vehicle charging demand to calculate the node voltage offset of the distribution network. The node voltage offset comparison is obtained from the reference voltage and power in Table 8. Figure 7 Figure 8 .

[0266] Figure 7 Adopt fixed grid structure (see Fig.10 ),because Figure 6 The location and capacity of the charging station in the table are the best choice (that is, they can meet the charging demand in Table 3 and the charging station benefits are the best). Fig.10 Fixed grid input and Figure 6Eight charging stations with the same location and capacity are installed, and the node voltage deviation is observed.

[0267] Depend on Figure 7 It can be seen that even if the distribution and capacity of charging stations are Figure 8 However, due to the lack of planning for the grid structure, the fixed distribution network structure is unable to cope with the additional load of charging stations. The voltage deviation of some nodes is -5.3%, which exceeds the -5.0% required by the GB / T12325-2008 "Power Quality Supply Voltage Deviation" standard.

[0268] In contrast, Figure 8 The distribution network structure generated by the two-layer optimization model ( Figure 6 ) after 8 charging stations were put into use. From the comparison, it can be seen that the distribution network structure after double-layer optimization can better distribute the charging load and ensure that the voltage deviation of each node meets the GB / T 12325-2008 standard, which further verifies the effectiveness of the double-layer optimization model in maintaining the voltage stability of the distribution network.

[0269] like Figure 8 As shown in the figure, due to the large number of EVs connected for charging, the voltage of the distribution network nodes has a drop deviation. The deviation of each node is the largest at 18:00, and the deviation of traffic node 18 is higher than that of other traffic nodes, at -4.6%.

[0270] Depend on Fig. 9 It can be seen that compared with the traditional PSO algorithm, genetic algorithm and differential algorithm, the improved PSO algorithm proposed in this patent has a fast convergence speed and the best optimization effect.

Claims

1. A charging station planning method based on charging demand and distribution network development, characterized in that: The following steps are involved: Step 1: Establish a road network-distribution network coupling model, which includes a road network model and a distribution network model. In the road network model, the distribution network nodes are directly used to represent the road network nodes. The road weight model and time period set are added to the basic structure of the road network model to generate a road network topology mathematical model. Step 2: Establish a two-layer planning model. The two-layer planning model includes an upper-layer planning model and a lower-layer planning model. The upper layer is the distribution network planning layer with the optimal distribution network revenue as the objective function, and the lower layer is the charging station site selection and capacity planning layer with the optimal charging station revenue as the objective function. Step 3: According to the two-level planning model established in Step 2, the improved particle swarm algorithm with the single-dimensional search volume probability mutation method is used to solve the two-level planning model and obtain the charging station and grid planning scheme.

2. The method according to claim 1, characterized in that In Step 1, when obtaining the road weight model, the following steps are taken: Step 1.1: Get the road impedance model: Where: Saturation Q is the road traffic flow, C is the road capacity; t0 is the travel time at zero flow; α and β are impedance influence factors, where both the road traffic flow and the road capacity follow a normal distribution, represents the road impedance when the saturation S is in the range of 0 < S ≤ 1.0, represents the road impedance when the saturation S is in the range of 1.0 ≤ S ≤ 2.0; Step 1.2: Get the node impedance model: Where: c is the traffic signal cycle; γ is the green ratio, that is, the proportion of the green light time in the whole signal cycle; q is the arrival rate of road vehicles per minute, represents the node impedance when the saturation S satisfies 0 < S ≤ 1.0, represents the node impedance when the saturation S satisfies S > 0.6; Step 1.3: The road weight model can be obtained by combining the models obtained in Step 1.1 and Step 1.2, w ij Represents the road weight of road ij: In the formula: represents the road impedance when the saturation S satisfies 0 < S ≤ 1.0, represents the road impedance when the saturation S satisfies 1.0 ≤ S ≤ 2.0, represents the node impedance when the saturation S satisfies 0 < S ≤ 1.0, represents the node impedance when the saturation S satisfies S > 0.

6.

3. The method according to claim 1, characterized in that In Step 2, the upper-level planning model is as follows: Objective function: maxF=-C inv -C main +C trans Where: C inv is the equivalent annual value of the cost of expanding the distribution network, including new lines and interconnection lines; C main is the operation and maintenance cost of the distribution network; C trans Profits from electricity sales to the distribution network; Expansion costs of distribution network: Where: d is the grid construction discount rate; m is the service life of the line; i and j are distribution network nodes; x i,j L is a 0-1 variable, indicating whether the line between nodes i and j is under construction. If it is under construction, it takes 1, and if it is not under construction, it takes 0; i,j is the length of line ij; C line is the construction cost per unit length of line; L i,j =10W i,j Operation and maintenance costs: C main =μC inv Where: μ is the operation and maintenance cost ratio; Revenue from electricity sales in the distribution network: Where: N node is the number of distribution network nodes; C sold,t P is the electricity price at time t; i,t is the charging power of the charging station at node i at time t, corresponding to the power of the lower-level charging station, which is determined by the number of charging piles built at the charging station at node i and the number of EV users going to the node to charge. The specific construction node of the charging station is obtained by the particle swarm algorithm calculation; Node voltage constraints: IN bmin ≤U b ≤U bmax U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

4. The method according to claim 2, characterized in that: In Step 2, the lower-level planning model aims to maximize the total revenue of the charging station consisting of construction cost, operation and maintenance cost, user cost, network loss cost and electricity sales revenue, and optimizes the construction location and capacity of the charging station with the road distance between the charging stations and the number of charging piles and the location of the charging station obtained in Step 1.3 as constraints; the lower-level planning model is specifically as follows: Objective function: maxF=-C I -C EV -C L +C B Where: C I is the average annual investment cost of the kth charging station; C L is the annual network loss cost of the charging station, C EV and C B They are annual user costs and revenue from electricity sales at charging stations; Annual investment cost: V M,i,k =b k C J,i,k N C is the total number of charging stations to be built, which is determined by the number of nodes with charging demand; C is the node set of the alternative charging station site. It is assumed that except for node 1, all other nodes in the distribution network are alternative points. J,i,k , C M,i,k and C o,i,k are the construction cost, operation and maintenance cost and other costs of the kth charging station at the i-th node; r0 is the discount rate; n0 is the operating life of the charging station; c ch is the unit price of the charging pile; N ch,i,k , A j,k , C f,i,k and β k are the number of charging piles at the kth charging station at the i-th node, the area occupied by a single charging pile, the land price per unit area, and the proportion of operation and maintenance cost to construction cost; Annual user cost: C EV =365C j Where: C j is the cost per user per day; The daily user cost is composed of the travel energy consumption, travel time and queue time costs: Where: c ev,t,k 、N i,c are the user cost of charging at the kth charging station at time t and the number of nodes with charging demand; t n,k,d ,t n,k,w They are the average journey time and average queue waiting time of EV from charging demand node n to charging station k; c time and c n,k,e are the user's unit time cost and the average power consumption from the nth charging demand node to the kth charging station; P b,t is the unit electricity price of the charging station at time t; Electric vehicle energy consumption model: c n,k,e =w ij X Where: w ij The road v where the vehicle is located generated in Step 1.3 ij The road weight at , Z is the average vehicle speed; Annual network loss cost of charging station: Where: P L,t is the network loss at time t, P c is the cost coefficient. Where: I ij is the absolute value of the branch current between nodes i and j; Z ij is the branch impedance between nodes i and j, N node is the number of distribution network nodes; Revenue from electricity sales at charging stations: P b,t is the peak / flat / valley electricity price of the charging station at time t, P i,t is the charging power at node i at time t; If the process of vehicles arriving at each charging station follows Poisson distribution, according to the queuing theory system average waiting time T w for: N ch,k is the number of charging piles in the charging station; ε is the average charging time of a single charging pile; E bt The charging demand for a single EV is obtained by combining Table 3 with Table 4; P ch is the charging power of a single charging pile; g is the average number of EV users arriving per unit time, which is the reciprocal of ε; ρ is the utilization rate of the charging pile; Distance Constraint: in s ≥u min A distance that is too small will cause traffic congestion during peak charging periods, so the distance between two charging stations needs to be limited; where the distance between any two charging stations is greater than or equal to the minimum distance limit; Restrictions on the number of charging piles: P ch The charging power of the charging pile; E max is the maximum value among all charging demand nodes at the same time; this constraint is the lower limit of the number of charging piles to be built, which can reduce the solution set of charging station capacity; Charging station construction node constraints: a i is a 0-1 variable, which takes 1 when a charging station is built at node i and takes 0 when no charging station is built at node i; A alt 、N a are the charging station candidate point set and the number of candidate points, N C is the total number of charging stations to be built; Feeder load factor constraints: S L and S L,max are the actual load and rated capacity of feeder i respectively; Line load factor constraints: S L,i and S L,i,max are the actual load and rated capacity of line i respectively; Node voltage constraints: IN bmin ≤U b ≤U bmax U b is the node voltage per unit value; U bmin is the lower limit of the node voltage per unit value; U bmax It is the upper limit of node voltage per unit value.

5. The method according to any one of claims 1 to 4, characterized in that: In Step 3, the following steps are included: Step 3.1 Encode the location of the EV charging station The number of particles in the particle swarm algorithm is set to T, the number of candidate points for EV charging stations is set to D; the initial position of the particle is X n ={X n1 ,X n2 ,…,X ni }, n = 1, 2, ..., T; corresponding to the charging station planning, X ni represents the construction status of the EV candidate distribution network node i of particle n; X ni =1 means to build an EV charging station at this candidate point, X ni =0 means not to build an EV charging station at this candidate point, X n The number of elements in the set equal to 1 must be equal to D, X n Each quantity in will change with the iteration until convergence; the initial velocity of the particle is V n = {V n1 ,V n2 ,…,V ni },V ni It represents the change of the objective function of the planning model in a single iteration during the iterative calculation of particle n, V n Each quantity in will change with the iteration until convergence; best,n For a single particle X n The historical optimal point after iteration, g best It is the global optimal point iterated by all particles; the update speed and position update formula of particles are: V n,r+1 =ωV n,r +c1r1(P best,n -X n,r )+c2r2(g best -X n,r ) X n,r+1 =X n,r +V n,r+1 Where: ω is the inertia weight; c1 and c2 are the self-learning coefficient and social learning coefficient respectively; r1 and r2 are mean random perturbations, which are used to simulate small perturbations caused by nature to increase the diversity of particles; An improved particle swarm algorithm based on the probabilistic mutation method of single-dimensional search volume is used to divide the dimensions of a single particle into multiple intervals. The historical search volume of the particle in each interval is calculated as the probability of selection mutation. At the same time as the algorithm iteration, the inertia weight and particle learning coefficient are updated. This improved algorithm has been proven to improve particle diversity, avoid premature particle maturation and accelerate convergence. Step 3.2 Update the inertia weight, self-learning coefficient and social learning coefficient in the particle update speed and position update formula The inertia weight is updated in a linear decreasing form as the number of iterations increases. The update formula is as follows: Where: max is the initial weight of the particle; ω min is the final weight of the particle; I iter.max is the maximum number of iterations; i iter is the current iteration number, and its value is equal to r; For the learning coefficient, in the initial iteration, a larger self-learning coefficient and a smaller social learning coefficient are set so that the particles can search throughout the entire space; and in the later iteration, in order to make the search target of the particles tend to the global optimal solution, a smaller self-learning coefficient and a larger social learning coefficient are set. The update formula is as follows: Where: c1 and c2 are the particle's self-learning coefficient and particle's social learning coefficient respectively, and the rest are constants. After research, it is found that when c1e, c1f, c2e, and c2f are set to 2.5, 0.5, 0.5, and 2.5 respectively, the optimization efficiency is better. Step 3.3 Use single-dimensional mutation to prevent the particle swarm algorithm from falling into the local optimal solution too early In the particle swarm algorithm, each particle represents a charging station location and capacity configuration scheme, and the dimension of the particle corresponds to the decision variable of the problem; in this specific application, the dimension of the particle refers to the different candidate locations of the charging station and its capacity configuration; the key to this algorithm is that when a particle falls into a local optimal solution too early, the position of this particle in this dimension is mutated so that the particle can continue to find the optimal solution; the steps are as follows: 1) Initialization parameters: First, initialize the number of particles N, the number of dimensions D, and the initial position X of each particle n and speed V n ; Initialize the inertia weight ω and learning coefficients c1 and c2; 2) Update the position and velocity of the particle: Use the formula to update the velocity and position of the particle; 3) Calculate the fitness value: For each particle, calculate its fitness value and update the local optimal value and the global optimal value; 4) Divide the intervals: Divide each dimension of the particle into several intervals; assume that the search limit of each interval is A max , that is, each interval can have at most A max Particles to search; 5) Detect search volume: When particle X n When the fth dimension of is about to be updated to a certain interval, check the search volume of the interval; if there is already A in the interval max If there are particles, the position of the particle in this dimension will not be updated to this interval, but will mutate to other intervals according to probability; 6) Calculate the probability of mutation: Determine the probability of mutation based on the historical search volume of each interval. Specifically, the larger the historical search volume of an interval, the smaller the probability of mutation to that interval. The probability of mutation to each interval can be expressed as: Among them, V fe is the search volume of the e-th interval in dimension f, V fb Aims to calculate the total search volume for all intervals of dimension f; 7) Update the position of the mutated particle: According to the mutation probability, randomly select a new interval and update the position of the particle in the fth dimension to this interval; 8) Iteration loop: Repeat steps 2 to 7 until the maximum number of iterations is reached or the fitness value converges to obtain the optimal charging station and grid planning scheme.

Images