Capacity Optimization of Independent Microgrid with Electric Vehicles Based on Improved Pelican Optimization Algorithm

Abstract

In order to reduce the comprehensive power cost of the independent microgrid and to improve environmental protection and power supply reliability, a two-layer power capacity optimization model of a microgrid with electric vehicles (EVs) was established that considered uncertainty and demand response. Based on the load and energy storage characteristics of electric vehicles, the classification of electric vehicles was proposed, and their mathematical models were established. The idea of robust optimization was adopted to construct the uncertain scenario set. Considering the incentive demand response, a two-layer power capacity optimization model of a microgrid was constructed. The improved pelican optimization algorithm (IPOA) was proposed as the two-layer model. In view of the slow convergence rate of the pelican optimization algorithm (POA) and its tendency to fall into the local optimum, methods such as elite reverse learning were proposed to generate the initial population, set disturbance inhibitors, and introduce Lévy flight to improve the initial population of the algorithm and enhance its global search ability. Finally, an independent microgrid was used as an example to verify the effectiveness of the proposed model and the improved algorithm. Considering that the total power capacity optimization cost of the microgrid after addition of electric vehicles was reduced by CNY 139,600, the total power capacity optimization cost of the microgrid after IOPA optimization was reduced by CNY 49,600 compared with that after POA optimization.

1. Introduction

Energy is the cornerstone of social development. With the decrease in coal, oil, and other conventional primary energy sources, it is imperative to develop and utilize renewable energy. Distributed generation can effectively complement traditional centralized generation and fully use renewable energy sources such as wind and solar. However, wind and solar power generation are intermittent and uncertain, which put the stable operation of the grid at risk [1]. Therefore, the concept of the microgrid is proposed to solve these problems. The microgrid refers to the power generation and distribution system composed of distributed energy, a battery storage system (BESS), control device, load, etc. [2]. With the increased load forms and capacity in recent years, more and more electric vehicles are becoming involved in the microgrid. Electric vehicles can be connected to the power grid as mobile loads. They also have the characteristics of energy storage devices [3] and can realize energy regulation together with energy storage devices. Connecting EVs to a microgrid with renewable energy and using clean electricity can promote both sides and improve the overall economic and environmental benefits of microgrids [4].

Current research must pay more attention to the role of EVs in optimizing power capacity allocation in microgrids and their energy storage characteristics, given the contradiction between load fluctuation and economic benefits caused by EV charging. The authors of [5] established a multi-objective programming model for an orderly EV charging microgrid under comprehensive constraints to coordinate the planning of the microgrid. In [6], EVs were regarded as transferable loads and incorporated into the microgrid considering demand side response, which reduced the photovoltaic power abandonment of the microgrid and improved the net present value of the microgrid within the planning cycle. In [7], the EV was regarded as a mobile storage device serving microgrids, and its energy storage characteristics were utilized to optimize wind and EV capacity configuration. Ref. [8] proposed a microgrid optimization management system considering demand response, including distributed generation and electric vehicles, to reduce operating costs and environmental emissions. The authors of [9] considered the charge–discharge characteristics of different types of electric vehicles on the microgrid and considered the uncertainty of source load and the random behavior of EV users. They proposed a two-stage stochastic programming and configuration model of an isolated microgrid composed of wind–wind batteries and EVs, which could minimize the model’s life cycle cost and reliability. The authors of [10] considered the uncertainty of power end, load end, and moving load, taking into account the demand-responsive EV and combined dynamic wireless charging technology, to put forward a two-layer framework of capacity optimization to maximize the utility of the microgrid and minimize the total generalized social cost. The authors of [11] considered the application of a passive enhanced control strategy in the converter and verified its effectiveness in the isolated island microgrid model. A point-to-point layered control method was proposed in [12], and its effectiveness was verified by simulation experiments in a hybrid DC microgrid. The authors of [13] proposed a capacity expansion planning model based on optimal network constraints to improve the topological flexibility of modern distribution networks. The authors of [14] proposed a new method of transmission line expansion and distributed generator distribution that considered load demand and renewable energy uncertainty and reduced economic costs, ensuring network stability and reliability.

Ref. [15] proposed an independent microgrid power capacity optimization allocation model that aimed to achieve a minimum average annual system cost. An improved gray wolf optimization algorithm with a global optimization performance was used to optimize the model. Simulation results indicated that the algorithm had a certain practicability and robustness. In [16], the uncertain output of wind turbines and photovoltaic modules was considered in establishing the capacity optimization allocation model of the microgrid. Based on this, the improved harmonic search algorithm was used to create the model, which improved the reliability of the capacity optimization allocation scheme. In [17], a new stochastic natural heuristic optimization algorithm, the pelican optimization algorithm (POA), was proposed, and its superiority was verified by test function. However, in the actual capacity optimization process of independent microgrids, the algorithm had defects, such as a slow convergence speed, low convergence accuracy, and quickly falling into medium local optimization.

In summary, this paper focuses on the capacity optimization allocation problem of independent microgrids and proposes a two-layer power capacity optimization model of microgrids with electric vehicles under uncertainty and demand response. First, based on the load and energy storage characteristics of electric vehicles, the categories of electric vehicles are given, and their mathematical models are established. Second, robust optimization is used to construct the uncertain scenario set, and the incentive demand response, including dispatchable electric vehicles, is given. Finally, a two-layer model of microgrid power capacity optimization is established. In the outer layer, the power capacity is optimized by taking the minimum annual value of the total cost of the microgrid as the objective function. The inner optimization model fully considers the uncertainty of source load, shiftable load, and dispatchable EV, and optimizes each operating variable to minimize the system’s annual optimization cost. An improved pelican optimization algorithm is proposed to realize the two-layer capacity optimization model of the microgrid proposed in this paper. By introducing a stochastic opposition learning strategy, the initial population is optimized, disturbance inhibition factors are added, and global search and local search capabilities are effectively balanced. Finally, the Lévy flight strategy is applied to the pelican optimization algorithm to conduct a small-range search near the optimal location. This helps the algorithm jump out of the local optimal in the late stage and improves the local search ability of the pelican optimization algorithm. Taking the annual value of the total power cost of the microgrid as the objective function, the improved algorithm is used to solve the problem, and the effectiveness of the improved algorithm in reducing the total cost of microgrid power is verified.

The rest of the paper Is organize” as follows. Section 2 presents the model of the microgrid, electric vehicle, uncertainty, and demand response. Section 3 focuses on the capacity optimization model of the microgrid considering the uncertainty and demand response. Section 4 introduces the basic POA and proposes the IPOA. Section 5 analyzes the example and discusses the influence of different factors on the capacity optimization results of the microgrid, and the Conclusion synthesizes our findings on the effectiveness of the model and the economy of the IPOA.

2. Basic Model

2.1. Microgrid Model

The independent microgrid studied in this paper is composed of wind power, photovoltaic, diesel generator, energy storage equipment, load, and electric vehicles. The load is divided into rigid and transferable loads, and the analog load of electric vehicles is divided into non-dispatchable and dispatchable EVs. Both transferable load and dispatchable EVs are included in the incentive demand response. The uncertain scenario is constructed by robust optimization. The architecture of the independent microgrid is shown in Figure 1.

2.1.1. Wind Power Output

The power P_wt(t) of a wind turbine (WT) varies with wind speed [18], and its value can be expressed in the following piecewise function relation:

$$P_{wt}\left(t\right)=\left\{\begin{array}{ll}0,\hfill & v<v_l,or,v\ge v_u\hfill \\ \frac{v^3-v_l^3}{v_o^3-v_l^3}{P}_o,\hfill & v_l\le v<v_o\hfill \\ P_o,\hfill & v_o\le v<v_u\hfill \end{array}\right.$$

(1)

where P_o is the rated power of the WT; v, v_l, v_o, v_u are the wind speeds at time period t of the WT, including the input-wind speed, rated wind speed, and cut-out wind speed.

2.1.2. Solar Power Output

Photovoltaic (PV) converts light into electric energy through photovoltaic modules and power generation devices, and its power output is mainly affected by the ambient temperature and illumination intensity [19]. The output power P_pv of PV can be calculated according to Equation (2):

$$\left\{\begin{array}{l}{P}_{pv}=\frac{P_{STC}{G}_{AC}[1+k\left(T_c-T_r\right)]}{G_{STC}}\hfill \\ T_c=T_{amt}+\frac{30G_{AC}}{1000}\hfill \end{array}\right.$$

(2)

where P_STC is the maximum test power under standard test conditions; G_AC is light intensity; k is the power temperature coefficient; T_r is the reference temperature, 25 °C; G_STC is 1 kW/m²; T_c is the operating temperature of the panel; T_amt is ambient temperature.

2.1.3. BESS Power Output

The battery energy storage system (BESS) consists of multiple batteries. State of charge (SOC) is an important parameter of the battery’s remaining power, and the SOC is determined by the power remaining from the previous moment and the charging and discharging power of the adjacent moment [20].

$$\left\{\begin{array}{c}{S}_{SOC}\left(t+1\right)=S_{SOC}\left(t\right)-{\eta }_d{P}_d\left(t\right)\Delta t\\ S_{SOC}\left(t+1\right)=S_{SOC}\left(t\right)+{\eta }_c{P}_c\left(t\right)\Delta t\end{array}\right.$$

(3)

where Δt is the unit of an hour; P_d(t) is the discharge quantity in unit time Δt; η_d is discharge efficiency; P_c(t) is the charging amount in unit time Δt; η_c is the charging efficiency.

2.1.4. DIE Power Output

Diesel generators (DIEs) are small and manageable, and are powered by diesel fuel; for [21], the fuel consumption F_oil (L/kW·h) is related to its output power:

$$F_{oil}=k_{die1}{P}_{die}^r+k_{die2}{P}_{die}$$

(4)

where k_die1 is the intercept coefficient of the fuel curve; $P_{die}^r$ is the rated power of the diesel generator; k_die2 is the slope of the fuel curve; P_die is the actual output power of the diesel generator.

2.2. Load Characteristic

In this paper, the total load per unit of time is considered to be composed of rigid load and shifting load:

$$P_{zfh}\left(t\right)=P_{gfh}\left(t\right)+P_{kfh}\left(t\right)$$

(5)

where P_zfh(t), P_gfh(t), P_kfh(t) are, respectively, the total load power, rigid load power, and translational load power within a unit of time.

2.3. Electric Vehicle Charging and Discharging Model

In this paper, the “analogical reasoning” method [22] was adopted to classify electric vehicles through the classification form of analog load. There are two main reasons for this analogy: In terms of form, the classification of electric vehicles corresponds to the type of load one by one. In the planning research, in order to simplify the calculation, the number of variables involved in the calculation should be reduced as much as possible. Therefore, by analogy with the load classification, this paper focused on mining the typical characteristics of load devices and giving a simplified mathematical model of electric vehicles without considering their complex operation process.

For simple calculation, this paper did not consider the distribution of charging piles and the driving path of electric vehicles. In terms of classification, by analogy with load classification, this paper divided electric vehicles into vehicles that accept dispatch and vehicles that do not accept dispatch. Electric vehicles that accept dispatch accept the agreement of incentive electricity price. Electric vehicles that do not accept dispatch will be charged without order. In summary, the simplified mathematical model of the electric vehicle is given as follows:

$$P_{EV}\left(t\right)=P_{gEV}\left(t\right)+P_{kEV}\left(t\right)$$

(6)

where P_EV(t), P_gEV(t), and P_kEV(t) are the power of the electric vehicle, power of the undispatchable electric vehicle, and power of the dispatchable electric vehicle, respectively.

2.3.1. Non-Dispatchable Electric Vehicles

The EVs that do not participate in the contract regulation of the electricity market are non-schedulable EVs, which mainly meet the load demand of the microgrid. In this paper, P_gEV(t) is used to represent the non-schedulable EV power.

2.3.2. Dispatchable Electric Vehicles

Electric vehicles perform charge and discharge, and dispatchable electric vehicles can be used as energy storage devices [7] and as loads; that is, they can consume electric energy as loads, and the unconsumed electric energy stored in them can be used as a distributed power supply to the microgrid. The charging amount of dispatchable EVs will not entirely be used as a power supply to the microgrid, which still consumes energy for the regular running of EVs. At the same time, when supplying power to the microgrid, an EV can only be supplied within its own dispatching period, which will not affect the normal energy consumption of the EV afterwards.

The charging and discharging power model of electric vehicles is similar to that of the battery charging and discharging power model. When electric vehicles are connected to the microgrid as a power source and supply power to the microgrid, their remaining capacity at time t is:

$$S_{EV}\left(t\right)=S_{EV}\left(t-1\right)-P_{KEV,D}\left(t\right){\eta }_{EV,D}\Delta t$$

(7)

When the electric vehicle needs to be charged as a load, the remaining capacity at time t is:

$$S_{EV}\left(t\right)=S_{EV}\left(t-1\right)+P_{KEV,C}\left(t\right){\eta }_{EV,C}\Delta t$$

(8)

Considering the energy consumption stage of the dispatchable EV, it is obvious that the energy consumption period of the dispatchable EV and the dispatchable charging and discharging period will not overlap, and the remaining capacity at time t is:

$$\left\{\begin{array}{l}{S}_{EV}\left(t\right)=S_{EV}\left(t-1\right)-P_{KEV,h}\left(t\right)\Delta t\hfill \\ S_{EV}\left(t_1\right)-S_{EV}\left(t_2\right)={\displaystyle \sum _{t=t_1}^{t_2}{P}_{KEV,h}\left(t\right)\Delta t}\hfill \\ s.t.\hfill \\ S_{EV\min }\le S_{EV}\left(t\right)\le S_{EV\max }\hfill \\ S_{EV\min }=n_{EV\min }{S}_{kEV}\hfill \\ S_{EV\max }=n_{EV\max }{S}_{kEV}\hfill \end{array}\right.$$

(9)

where S_EV(t) is the capacity of the electric vehicle at time t, S_EVmin and S_EVmax are its minimum and maximum, respectively; n_EVmin and n_EVmax are the minimum and maximum coefficients; P_KEV,D(t), P_KEV,C(t), and P_KEV,h(t) are the discharge, charging, and energy consumption power of electric vehicles; η_EV,D and η_EV,C are the charging and discharging coefficients of electric vehicles.

2.3.3. Constraints and Power Balance Constraints

The total capacity of an EV is the sum of the capacities of a non-dispatchable EV and a dispatchable EV, and its power is also the sum of the capacities of the two. The charging and discharging capacity per unit time of an EV cannot exceed the maximum capacity.

$$\left\{\begin{array}{l}{S}_{EV}=S_{gEV}+S_{kEV}\hfill \\ S_{gEV}=S_{EVO}{\eta }_{gEV}\hfill \\ S_{kEV}=S_{EVO}{\eta }_{kEV}\hfill \\ {\eta }_{EV}={\eta }_{gEV}+{\eta }_{kEV}\hfill \\ P_{EV}\left(t\right)=P_{gEV}\left(t\right)+P_{kEV}\left(t\right)\hfill \\ P_{EV}\left(t\right)\Delta t\le S_{EV}\hfill \\ P_{gEV}\left(t\right)\Delta t\le S_{gEV}\hfill \\ P_{kEV}\left(t\right)\Delta t\le S_{kEV}\hfill \end{array}\right.$$

(10)

where S_EV, S_gEV, and S_kEV are the total capacity of electric vehicles, the total capacity of non-schedulable electric vehicles, and the total capacity of schedulable electric vehicles, respectively. S_EVO is the capacity of a single electric vehicle; P_EV(t), P_gEV(t), and P_kEV(t) are, respectively, the power of electric vehicles, the power of undispatchable electric vehicles, and the power of dispatchable electric vehicles. η_EV, η_gEV, and η_kEV are the total number of electric vehicles, the number of undispatchable electric vehicles, and the number of dispatchable electric vehicles, respectively.

2.4. Incentive Demand Response

The incentive DR considered in this paper is realized through an agreement signed between the independent microgrid operator and the user. When needed, the independent microgrid operator can adjust the working period of the load’s time-shifting load and give specific compensation to the user according to the transfer amount. At the same time, since this paper considered the power supply of electric vehicles to the microgrid, the microgrid should pay the corresponding power purchase cost.

$$P_{zy}\left(t\right)=P_{zyo}\left(t\right)+P_{zyEV}\left(t\right)$$

(11)

$$S_{zyEV}\left(t\right)=S\left(t\right)-S\left(t-1\right)$$

(12)

where P_zy(t), P_zyo(t), and P_zyEV(t) are, respectively, the total load transfer power, the transfer power of the shifting load, and the transfer power of the dispatchable EV. S_zyEV(t) refers to the power supply of the electric vehicle to the microgrid within a unit of time. At this time, time t refers to the power supply stage of the electric vehicle, and the unit power supply at other times is 0.

2.5. Uncertainty

Both the power end and the load end have significant uncertainty. The uncertainty of the power end comes mainly from wind speed and solar radiation intensity. The load end comes from the ordinary load and electric vehicle load. Meanwhile, the shiftable and dispatchable electric vehicle loads will also have uncertain effects owing to user comfort.

This paper adopted a robust optimization method in processing source load uncertainty. The key to robust optimization is constructing an uncertain set [23]. The magnitude of the source charge can be expressed as its predicted value plus the predicted error. For an independent microgrid, when wind speed and solar radiation reach the lower limit and load reaches the upper limit, the configuration and operation cost of the microgrid is higher, which is more in line with the idea of robust optimization. Therefore, the uncertain set can be expressed as:

$$\left\{\begin{array}{l}{u}_{i,t}=u_{for,i,t}+r_{i,t}{x}_{i,t}\hfill \\ {\displaystyle \sum _{t=1}^{24}{r}_{i,t}\le N_i}\hfill \end{array}\right.$$

(13)

where u_i,t is the value of time period t of six uncertain variables, namely, wind speed, solar radiation intensity, load, shiftable load, electric vehicle load, and dispatchable electric vehicle load; u_for,i,t are the predicted values of uncertain variables; x_i,t is the maximum prediction error of the uncertain variable, the maximum prediction error of scenery is 10%, the maximum prediction error of the load is 10%, and the maximum prediction error of the translation load and dispatchable electric vehicle load is 10%. r_i,t is 0 or 1. When it is 1, the uncertain variable is taken to the upper or lower limit value, and when it is 0, the predicted value is taken. N_i,t is the degree of uncertainty, indicating the number of upper or lower limits of uncertainty variables in a day, which can be set as an integer ranging from 0 to 24. The larger the value, the more conservative the configuration scheme obtained using the uncertainty set, and this paper’s uncertainty was set at 6.

3. Capacity Optimization Model

This paper proposed a two-layer optimization model for capacity optimization of the independent microgrid. In the outer planning level, considering the equipment investment cost as well as the operation and maintenance costs, planning decisions were made with the minimum annual value of total cost as the goal. At the inner operation level, considering the uncertainty of the source load and combined with the incentive demand response, an optimal operation analysis of different types of power supply, energy storage, and electric vehicle output was carried out.

3.1. Upper-Level Model

The upper layer focuses on the lowest annual C_wai of the total cost and the equal value of the operator’s micro-network construction under independent micro-network construction and operation requirements.

$$\left\{\begin{array}{l}{C}_{wai}=\underset{x}{\min }\left(C_{inv}+C_{nei}\right)\hfill \\ C_{inv}=C_{ini}+C_{main}+C_{cha}\hfill \end{array}\right.$$

(14)

where C_inv is the annual planning cost of micro-network construction by operators; C_nei is the annual operating cost of the system; C_inv mainly includes average annual investment cost C_ini, annual maintenance cost C_main, and replacement cost C_cha.

(1)

The average annual investment cost is related to the equipment’s first investment cost and the system’s operating life.

$$\left\{\begin{array}{l}{C}_{ini}=\left({\displaystyle \sum _{i=1}^N{k}_i^p{n}_i{P}_i^r}\right){\eta }_{crf}\hfill \\ {\eta }_{crf}=\frac{a{\left(1+a\right)}^Y}{{\left(1+a\right)}^Y-1}\hfill \\ a=\frac{a\prime -y}{1+y}\hfill \end{array}\right.$$

(15)

where $k_i^p$ is the power cost coefficient of each distributed power supply; n_i is the number of installed power supplies; $P_i^r$ is the rated power of each power supply; η_crf is the fund recovery coefficient. a is the effective interest rate; Y is the whole life of the system, and 20 years was taken in this paper. a′ is the rated interest rate, 8% in this paper; y is the annual inflation rate, 5% in this paper.

(2)

Average annual equipment operation and maintenance expenses are related to operation and maintenance costs in the first year.

$$C_{main}={\eta }_{crf}\left(\frac{1+y}{a-y}\right)\left[1-\left(\frac{1+y}{1+a}\right)\right]{\displaystyle \sum _{i=1}^N{k}_i^{fix}{n}_i}$$

(16)

where $k_i^{fix}$ is each power supply’s operation and maintenance cost coefficient.

(3)

The average annual replacement cost is related to the service life of each distributed power source.

$$C_{cha}=\frac{a}{{\left(1+a\right)}^Y-1}{\displaystyle \sum _{i=1}^N{k}_i^p{n}_i{P}_i^r{L}_i}$$

(17)

where L_i is the replacement times of each power supply within the system life.

3.2. Lower-Level Model

The lower-level model is optimized with the goal of the lowest C_nei of the system’s annual operating cost under the operation scenario with the uncertainty of source load.

$$C_{nei}=\underset{u}{\max }\left\{\underset{y\in \mathsf{\Omega }\left(x,u\right)}{\min }\left(C_{fuel}+C_P+C_W+C_{DR}\right)\right\}$$

(18)

where C_fuel is the fuel cost of the diesel generator; C_P is the pollution control cost of the system; C_W is the renewable energy waste penalty cost; and C_DR is the compensation cost for the incentive DR.

(1)

The independent microgrid considered in this paper consumes fuel only when the diesel generator runs, and the fuel cost is generated only by the diesel engine.

$$C_{fuel}={\displaystyle {\int }_0^T{k}_{oil}{P}_{die}\left(t\right)dt}$$

(19)

where k_oil is the fuel cost coefficient of the diesel generator, and P_die(t) is the actual operating power of the diesel generator.

(2)

A diesel engine will emit CO₂, SO₂, NO_x, and other polluting gases during operation. To unify the dimension, this must be converted into pollution control costs for treatment. Pollution control cost C_P is related to the output power of the diesel generator:

$$C_P={\displaystyle {\int }_0^T{\displaystyle \sum _{j=1}^3{k}_j^{emi}{k}_j^{pol}{P}_{die}\left(t\right)dt}}$$

(20)

where $k_j^{emi}$ is the emission coefficient of different pollutants, and $k_j^{pol}$ is the treatment cost coefficient of different pollutants.

(3)

Due to the randomness of scenery resources, there will be some problems, such as wind and light abandonment during the operation of the system; that is, the power output of the microgrid system will be greater than the power demand, which leads to energy waste and a decrease in the utilization rate of renewable energy. Energy waste penalty cost C_W is introduced here to measure the utilization rate of renewable energy in the system:

$$C_W={\displaystyle {\int }_0^T{k}_{was}{P}_{was}\left(t\right)dt}$$

(21)

where k_was is the penalty cost coefficient, and P_was(t) is the power of energy surplus.

(4)

The total cost C_DR of the incentive type includes the compensation cost C_DR,1 and C_DR,2 of the shiftable load and dispatchable EV to users, and the cost C_DR,3 of the microgrid to purchase electricity from dispatchable EV users, which is expressed as follows:

$$\left\{\begin{array}{l}{C}_{DR}={\displaystyle \sum _{t=1}^T\left(C_{DR,1}\left(t\right)+C_{DR,2}\left(t\right)+C_{DR,3}\left(t\right)\right)}\hfill \\ C_{DR,1}\left(t\right)=a_1{P}_{zyo}\left(t\right)\Delta t\hfill \\ C_{DR,2}\left(t\right)=a_2{P}_{zyEV}\left(t\right)\Delta t\hfill \\ C_{DR,3}\left(t\right)=a_3{P}_{KEV,D}\left(t\right)\Delta t\hfill \end{array}\right.$$

(22)

where a₁, a₂, and a₃ are the compensation costs of unit electricity transferred by shiftable load, dispatchable EV, and the cost of unit electricity purchased by the microgrid, respectively.

3.3. Constraint Condition

(1)

Because the area available for the installation of distributed power devices in the microgrid is limited, the number of power supplies to be installed should meet the constraints

$$n_i^{\min }\le n_i\le n_i^{\max }$$

(23)

(2)

The battery’s life is related to whether it is deeply discharged. Overcharging or overdischarging will reduce the battery life. In order to extend the service life of the battery, the charging and discharging power and the state of charge of the battery should meet the constraints

$$\left\{\begin{array}{c}{P}_d^t\le 0.2S_{SOC}^r/\Delta t\\ P_c^t\le 0.2S_{SOC}^r/\Delta t\end{array}\right.$$

(24)

$$0.1*S_{SOC}^r\le S_{SOC}\left(t\right)\le 0.9*S_{SOC}^r$$

(25)

(3)

The power deficit constraint and the reliability of the system are represented by the size of the system power deficit; a low power deficit will not cause a more significant impact on production and life. In this paper, the maximum power deficit rate of the system was set at 0.1%; that is, the power deficit cannot exceed 0.1% of the total load of the system. The power deficit rate k_sho can be calculated as follows:

$$k_{sho}={\displaystyle {\int }_0^T\left({\displaystyle \sum _{i=1}^4{P}_i\left(t\right)-P_{fh}\left(t\right)}\right)dt/P_{FH}}$$

(26)

where P_i(t) is the power of each power supply at time t; P_fh(t) is the power at time t of the load; P_FH is the total annual load.

(4)

The environmental protection index of the system is represented by the emission of pollutants at the power generation end of the system. The only pollution source of the microgrid system considered in this paper was the pollutant emission from diesel generators. In order to ensure the environmental protection requirements of the microgrid, pollutant emissions generated by the operation of diesel generators need to be controlled. In this paper, pollution control costs were used for restraint, and the maximum pollution control cost in this paper was set at CNY 1 million.

$$C_P\le C_P^{\max }$$

(27)

where $C_P^{max}$ is the maximum pollution control cost of the system.

3.4. Microgrid Operation Control Strategy

In the unit scheduling time, the net generation is positive, and the scheduling of the shifting load is the priority, followed by the scheduling of the electric vehicle load and, finally, the charging of the energy storage equipment. Within the unit scheduling time, when the net energy generation is harmful, in order to ensure a sound environmental protection index and the comfort level of EV users, the discharge of the energy storage device is given priority, followed by generation of the diesel generator, and finally, the EV can be dispatched to supply power to the microgrid. The model-solving flow chart is shown in Figure 2.

4. Model-Solving Method

The optimal allocation problem of the microgrid is a complex integer optimization problem that is multi-objective, multi-constraint, and strongly nonlinear. The traditional mathematical optimization algorithm is often ineffective, and the swarm intelligence algorithm is the most commonly used method to solve this problem. In the past few years, many swarm intelligence algorithms have been applied to the optimization of microgrid configuration, but most studies have directly applied the continuous optimization algorithm to the microgrid configuration problem in the way of integration, causing the algorithm to fall into the local optimal so that there is no satisfactory configuration scheme. Therefore, this is a significant problem that must be solved to determine the optimization algorithm that can solve the optimization allocation problem of the microgrid.

4.1. Pelican Optimization Algorithm

The pelican optimization algorithm (POA) is a population-based algorithm in which pelicans are population members. In a population-based algorithm, each population member represents a candidate solution. Each group member proposes a value for the optimization problem variable based on its location in the search space. Initially, an equation is used to randomly initialize the population members according to the lower and upper limits of the problem.

$$\begin{array}{l}{x}_{i,j}=l_j+rand\cdot \left(u_j-l_j\right),\\ i=1,2,\dots ,N,j=1,2,\dots ,m,\end{array}$$

(28)

where x_i,j is the value of the j-th variable specified by the i-th candidate solution; N is the number of population members; m is the number of problem variables; rand is the random number in the interval [0, 1]; l_j is the j-th lower bound; and u_j is the j-th upper bound of the problem variable.

A matrix called the population matrix was used in Equation (29) to identify the population members of pelicans in the proposed POA. Each row of the matrix represents the candidate solution, and the columns of the matrix represent the suggested value of the problem variable.

$$X={\left[\begin{array}{c}{X}_1\\ ⋮\\ X_i\\ ⋮\\ X_N\end{array}\right]}_{N\times m}={\left[\begin{array}{ccccc}{x}_{1,1}& \cdots & x_{1,j}& \cdots & x_{1,m}\\ ⋮& \ddots & ⋮& ⋰& ⋮\\ x_{i,1}& \cdots & x_{i,j}& \cdots & x_{i,m}\\ ⋮& ⋰& ⋮& \ddots & ⋮\\ x_{N,1}& \cdots & x_{N,j}& \cdots & x_{N,m}\end{array}\right]}_{N\times m}$$

(29)

where X is the pelican population matrix, X_i is the i-th pelican.

In the proposed POA, each group member is a pelican, representing a candidate solution to a given problem. Therefore, the objective function of a given problem can be evaluated based on each candidate’s solution. The vector called the objective function vector in Equation (30) is used to determine the value obtained for the objective function.

$$F={\left[\begin{array}{c}{F}_1\\ ⋮\\ F_i\\ ⋮\\ F_N\end{array}\right]}_{N\times 1}={\left[\begin{array}{c}F\left(X_1\right)\\ ⋮\\ F\left(X_i\right)\\ ⋮\\ F\left(X_N\right)\end{array}\right]}_{N\times 1}$$

(30)

where F is the objective function vector, and F_i is the objective function value of the i-th candidate solution.

The proposed POA simulates pelican behavior and strategies when attacking and hunting prey to update candidate solutions. This hunting strategy is simulated in two stages:

(1)

Move toward prey (exploration phase).

(2)

Wingspan above water (development stage).

4.1.1. Phase 1: Move to Prey (Exploration Phase)

In the first stage, the pelican identifies the location of its prey and then moves toward that identified area. Modeling this pelican strategy enables scanning of the search space and improves the exploration ability of the proposed POA in discovering different areas of the search space. The important thing about the POA is that the location of the prey is randomly generated in the search space. This increases the POA’s ability to explore the space for precise search problem solving. The equation mathematically simulates the concept and the pelican’s strategy of moving to the location of prey.

$$x_{i,j}^{P_1}=\left\{\begin{array}{ll}{x}_{i,j}+rand\cdot \left(p_j-I\cdot x_{i,j}\right),\hfill & F_p<F_i\hfill \\ x_{i,j}+rand\cdot \left(x_{i,j}-p_j\right),\hfill & else\hfill \end{array}\right.$$

(31)

where $x_{i,j}^{P_1}$ is the new state of the i-th pelican in the j-th dimension based on stage 1, I is a random number equal to 1 or 2, p_j is the position of the prey in the j-th dimension, and F_p is its objective function value. The parameter I is a number that could randomly be equal to 1 or 2. This parameter is randomly selected for each iteration and each member. When the value of this parameter is equal to 2, it creates more displacement for the member, which may cause the member to enter a newer part of the search space. Therefore, parameter I affects the POA probe’s ability to accurately scan the search space.

In the proposed POA, the new location of the pelican is acceptable if the value of the objective function is improved at that location. In this type of update, called a valid update, the algorithm is prevented from moving to a non-optimal region. The process is modeled using Equation (32):

$$X_i=\left\{\begin{array}{ll}{X}_i^{P_1},\hfill & F_i^{P_1}<F_i\hfill \\ X_i,\hfill & else\hfill \end{array}\right.$$

(32)

where $X_i^{P_1}$ is the new state of the i-th pelican, and $F_i^{P_1}$ is its objective function value based on stage 1.

4.1.2. Phase 2: Flying over Water (Development Phase)

In the second stage, after the pelicans reach the surface, they spread their wings above the water to move the fish upward before collecting their prey in throat bags. This strategy results in more fish being caught by pelicans in areas under attack. Modeling this behavior in pelicans allows the proposed POA to converge to a better point in the hunting area. This process improves the local search and development capabilities of the POA. From a mathematical point of view, the algorithm must examine points near the pelican position to converge on a better solution. Equation (6) mathematically simulates this pelican behavior during hunting.

$$x_{i,j}^{P_2}=x_{i,j}+R\cdot \left(1-\frac{t}{T}\right)\cdot \left(2\cdot rand-1\right)\cdot x_{i,j}$$

(33)

where $x_{i,j}^{P_2}$ is the i-th pelican based on phase 2 of the new state of the j-th dimension, R is constant and equal to 0.2, R·(1 − t/T) is x_i,j neighborhood radius, t is the iteration counter, T is the largest number of iterations. Coefficient “R·(1 − t/T)” represents the neighborhood radius of the group members, each member in the nearby local search converging to a better solution. This coefficient effectively brings the POA development capability closer to the optimal global solution. In the initial iteration, the value of this coefficient is more significant; therefore, the area around each component is more significant. With the increase in the algorithm of repetition, the “R·(1 − t/T)” coefficient decreases, along with the neighborhood radius of each member. This allows for us to scan the area around each member of the population at a smaller and more precise step size so that the POA can converge closer to a global (or even fully global) optimal solution based on the usage concept.

At this stage, valid updates are also used to accept or reject the new pelican location modeled in Equation (34).

$$X_i=\left\{\begin{array}{ll}{X}_i^{P_2},\hfill & F_i^{P_2}<F_i\hfill \\ X_i,\hfill & else\hfill \end{array}\right.$$

(34)

where $X_i^{P_2}$ is the new state of the i-th pelican, and $F_i^{P_2}$ is its objective function value based on the second stage.

4.2. Improved Pelican Optimization Algorithm

The basic POA still has some problems, such as premature convergence and poor local search ability, which need to be improved. Therefore, an improved pelican optimization algorithm was proposed to solve the above optimization model.

4.2.1. Population Initialization Based on Stochastic Opposition Learning

The random method was used to initialize the POA population, and the generated population was not uniform, which affected the convergence speed and accuracy. To obtain a better initial population [24], this paper used a stochastic contrastive learning strategy [25] to obtain the latest position X_i,new, denoted as:

$$X_{i,new}=\left(l+u\right)-k{X}_i$$

(35)

where X_i is the current individual’s location information, and k is a random number between (0, 1).

After the random opposition learning optimization, the corresponding individual’s fitness function value is calculated. By comparing the fitness function value of the current individual and the optimized individual, the better fitness value is selected.

$$X_i=\left\{\begin{array}{ll}{X}_{i,new},\hfill & f\left(X_{i,new}\right)<f\left(X_i\right)\hfill \\ X_i,\hfill & else\hfill \end{array}\right.$$

(36)

4.2.2. Disturbance Suppressor

Perturbation inhibitors can effectively balance the global and local search capabilities of the algorithm [26,27]. The pelican optimization algorithm introduces disturbance inhibition factor θ, and the formula is as follows:

$$\theta =\sin \left(\frac{\pi \cdot t}{2\cdot T_{\max }}+\pi \right)+1$$

(37)

where t is the number of current iterations, and T_max is the maximum number of iterations.

After adding disturbance inhibitors, the optimization models of the prey moving stage and the flight stage contained in the pelican optimization algorithm are expressed as:

$$x_{i,j}^{P_1}=\left\{\begin{array}{ll}{x}_{i,j}+\theta \cdot rand\cdot \left(p_j-I\cdot x_{i,j}\right),\hfill & F_p<F_i;\hfill \\ x_{i,j}+\theta \cdot rand\cdot \left(x_{i,j}-p_j\right),\hfill & else\hfill \end{array}\right.$$

(38)

$$x_{i,j}^{P_2}=x_{i,j}+\theta \cdot R\cdot \left(1-\frac{t}{T}\right)\cdot \left(2\cdot rand-1\right)\cdot x_{i,j}$$

(39)

4.2.3. Lévy Flight Strategy

Lévy flight strategy was introduced to solve the problem of losing the POA diversity in the later period and easily falling into the local optimal. Lévy flight is a random strategy widely used in optimization algorithms. It has a high step probability in a random walk, which can effectively improve the randomness of the algorithm [28,29]. To further enhance the method’s exploration abilities, Lévy flights were integrated into Phase 2 of the POA formula, as follows:

$$X_{i,new}=X_{best}+\omega \cdot R\cdot \left(1-\frac{t}{T}\right)\cdot \left(2\cdot rand-1\right)\cdot X_{best}\cdot Levy\left(D\right)$$

(40)

$$Levy\left(D\right)=0.01\times \frac{c\times \sigma }{{\left|d\right|}^{\frac{1}{\beta }}}$$

(41)

$$\sigma ={\left(\frac{\mathsf{\Gamma }\left(1+\beta \right)\times \sin \left(\frac{\pi \beta }{2}\right)}{\mathsf{\Gamma }\left(\frac{1+\beta }{2}\right)\times \beta \times 2^{\left(\frac{\beta -1}{2}\right)}}\right)}^{\frac{1}{\beta }}$$

(42)

where Levy represents Lévy flight function, and D is the dimension of the problem. c and d are random values between 0 and 1, and β is a constant equal to 1.5.

4.3. IPOA Algorithm Flow Chart and Pseudo-Code

Capacity optimization process of the microgrid based on the IPOA is shown in Figure 3. Pseudo-code of IPOA (Algorithm 1) is as follows.

Algorithm 1. Pseudo-code of IPOA

Start IPOA 1. Enter the optimization problem information. 2. Determine the overall size of IPOA (N) and the number of iterations (T). 3. Initialize the pelican position and calculate the objective function. 4. Update the pelican position and calculate the objective function using Equation (35). 5. Locate the pelican using Equation (36). 6. For t = 1:T 7.   Randomly generate prey positions. 8.   For i = 1:N 9.   Phase 1: Moving toward prey (exploration stage). 10.    For j = 1:m 11.     Calculate the new state of the j-th dimension using Equation (38). 12.    End 13.    Update the ith population member using Equation (32). 14.   Phase 2: Flying over water (development phase). 15.    For j = 1:m. 16.     Calculate the new state of dimension j using Equation (39). 17.    End 18.    Update the i-th population member using Equation (34). 19.    Calculate the new state of dimension j using Equation (40). 20.    Update the i-th population member using Equation (34). 21.   End 22.   Update the best candidate solutions. 23. End. 24. Output the best candidate solution obtained through IPOA.    End IPOA.

5. Example Analysis

In order to verify the rationality and effectiveness of the model and method proposed in this paper, we used a microgrid system in a region as the example, taking the scenery data of the region and the load data of local users for one year, as shown in Figure 4. The two-layer optimization model and solving algorithm proposed in this paper were adopted to optimize the capacity of the microgrid.

5.1. Basic Data and Parameter Settings

Parameters included economic index, reliability index (power shortage rate), environmental protection index (pollution control cost), wind and light abandonment rate (WLAR). Technical and economic indicators of microgrid components are shown in

Table 1. Technical and economic indicators of microgrid components. (CNY 1 = USD 0.1450, CNY 1 = EUR 0.1368).

Parameter	Numerical Value	Parameter	Numerical Value
[vl,vo,vu]	[4,15,25] (m/s)	[ηEV,ηgEV,ηkEV]	[120,20,100]
k	−0.34%/K	$k_p$I	[0.5,0.6,1,0.0625] (104 CNY/kW)
Tr	25 °C	$P_i^r$	[10,1,10,25] (kW)
GSTC	1 kW/m2	$k_i^{fix}$	[0.1,0.023,0.23,0.2] (104 CNY/per·year)
[ηd,ηc]	[1.1,0.9]	Li	[0,0,0,3]
kdie1	0.246	koil	1.76 (CNY/kW·h)
kdie2	0.084	$k_j^{emi}$	[232.04,0.46,4.33] (g/kW·h)
[nEVmin,nEVmax]	[0.2,0.8]	$k_j^{pol}$	[0.0076,0.77,1.53] (CNY/kg)
[ηEV,D,ηEV,C]	[1.05,0.95]	kwas	0.5 (CNY/kW·h)
SEVO	20 kW	[a1,a2,a3]	[0.5,0.2,0.6] (CNY/kW·h)

.

5.2. Analysis and Comparison of Different Influencing Factors

In order to fully consider the economic improvement in the power capacity optimization results of the microgrid with electric vehicles included and to compare the impact of the uncertainty and demand response on the results, the following four schemes were set for comparative analysis.

Scheme 1: A model that does not consider the uncertainty and incentive demand response and does not include the load of electric vehicles;

Scheme 2: A model that considers the incentive demand response without considering the uncertainty and electric vehicle load;

Scheme 3: A model that considers the incentive demand response and adds the electric vehicle without considering the uncertainty. In order to make the comparison results more credible, this paper assumed that the total load and total schedulable load of this scheme with the addition of electric vehicles were the same as those of the former scheme;

Scheme 4: A model that considers the uncertainty and incentive demand response, adding the electric vehicle model.

Figure 5 shows the curves of wind speed, solar radiation intensity, and load in a typical day after considering uncertainty that was set at 6. Wind speed and solar radiation intensity decreased to a certain extent under the influence of uncertainty, while the load increased slightly under the influence of uncertainty.

Figure 6 shows the result of the power balance of the source load with or without EVs. It can be concluded from the figure that the power balance result of the microgrid is better when the operation of EVs considers the scheduling optimization of the microgrid.

Influence of DR excitation: In Scheme 1, the number of energy storage devices reached the upper limit because of system reliability and environmental protection constraints. Compared with Scheme 1, after the addition of the incentive DR in Scheme 2 because of the environmental indicator constraints, the installed capacity of the diesel generator did not change, the capacity of the fan and photovoltaic was reduced, and the capacity of the energy storage equipment was greatly reduced. The total cost was reduced by CNY 764,406, the investment cost was reduced by CNY 557,000, and the total cost of the incentive DR was CNY 39,107. Wind and light discard rates decreased by 6.97%. Therefore, it was concluded that more optimized configuration results can be obtained by considering the influence of the excitation DR in the optimization stage of the microgrid power supply; in particular, the capacity of the energy storage equipment will be greatly reduced.

Influence of the electric vehicle: The power balance result of the source load with or without electric vehicles is shown in Figure 3. Since the total load and total dispatchable load of the scheme with electric vehicles added were assumed in this paper to be the same as that of the former scheme, the modified load curves of both showed little change. After the electric vehicle was added, the output of the original energy storage equipment was greatly reduced, considering its energy storage characteristics. As seen in

Table 2. Capacity optimization cost of the microgrid for different schemes.

Scheme	WT	PV	DIE	BESS	Cwai/104 CNY	Cini/104 CNY	CDR/104 CNY	CDR,2&3/104 CNY	WLAT/%
1	388	3043	107	200	895.7545	442.03	/	/	28.15
2	323	3205	107	128	819.3139	386.33	3.9107	/	21.18
3	314	3212	109	98	805.3506	368.10	5.8230	1.8631	21.18
4	356	3360	110	103	840.1928	391.92	5.7566	1.8475	25.04

, after the electric vehicles were added to the microgrid, the number of energy storage devices was larger and smaller, and the capacity of the scenic wood was almost unchanged. The total cost of the microgrid was reduced by CNY 139,633, the investment cost was reduced by CNY 182,300, the total incentive DR cost was increased by CNY 19,123, and the incentive DR cost of the trolley was CNY 18,631. It was determined from the table that the number of energy storage equipment configurations was greatly reduced, which benefited from the energy storage characteristics of the dispatchable electric vehicles. It was concluded that the inclusion of electric vehicles in the incentive DR can reduce the investment cost of the microgrid, especially the investment cost of the energy storage equipment.

Effects of different proportions of dispatchable electric vehicles: According to the analysis of

Table 3. Capacity optimization cost of the microgrid with different proportions of dispatchable EVs.

KEV/%	Cwai/104 CNY	Cini/104 CNY	CDR/104 CNY	CDR,2&3/104 CNY	WLAT/%
50%	805.35	368.10	5.82	1.86	21.18
75%	798.65	358.63	6.01	2.11	20.63
100%	791.95	350.14	6.17	2.29	19.72

, as the proportion of dispatchable electric vehicles increased, the required incentive DR cost slightly increased, but the total system cost, investment cost, and wind and light abandonment rate all decreased.

The impact of uncertainty: After considering the uncertainty, the capacity of the fan and photovoltaic significantly increased, the capacity of the diesel generator was almost unchanged, the capacity of the energy storage equipment slightly increased. The total cost increased by CNY 348,422, the investment cost increased by CNY 238,200, the wind and light abandonment rate increased by 3.86%, and the total incentive DR cost decreased by CNY 0.0664. The trolley incentive DR cost was reduced by RMB 0.0156 million. To meet the constraints of system reliability and environmental protection indicators, the capacity of fan and photovoltaic, the total cost, investment cost, and the wind and light abandonment rate of the microgrid significantly increased after considering the adverse impact of source load uncertainty; the schedulable load decreased when considering the uncertain impact of user comfort. The total cost of the excitation DR and the cost of the trolley excitation DR were slightly reduced, and the capacity of the energy storage equipment was slightly increased.

5.3. Comparative Analysis of IPOA and Other Algorithms

In order to verify the optimization effect of the IPOA proposed in this paper on the two-layer model of the microgrid, the improved IPOA was compared with the original POA, the whale optimization algorithm (WOA), and the gray wolf optimization algorithm (GWO); then, the total cost objective function of the power supply allocation of the microgrid was optimized. To ensure the fairness of the comparison, the population number of the four algorithms was set to 30, and the maximum number of iterations was set to 100. The comparison results are shown in Figure 7.

It can be seen in Figure 7 that the optimal initial population of the WOA was CNY 8,876,457, the optimal initial population of the GWO was CNY 9,148,053, the optimal initial population of the POA was CNY 8,611,484, and the optimal initial population of the IPOA was CNY 8,552,837. By comparison, the initial population objective function value of the IPOA was optimal, so at the initial optimization stage of the IPOA, the random opposition learning strategy was adopted to obtain a better initial population, making the algorithm converge faster in the early stage. It can be seen in

Table 4. Capacity optimization cost of the microgrid with different algorithms.

Algorithm	Starting Population Optimums/104 CNY	Cwai/104 CNY	Iteration Time/s
WOA	887.6457	867.4193	25.36
GWO	914.8053	861.4213	26.07
POA	861.1484	845.1529	50.51
IPOA	855.2837	840.1928	74.47

that the cost of the IPOA was lower than that of the WOA, GWO, and POA by CNY 272,265, CNY 212,285, and CNY 49,601, respectively, indicating that the improved IPOA obtained better objective function values in the proposed two-layer optimization model of the microgrid.

6. Summary of the Main Results

In this section, the main results are synthesized, and all factors are considered, including uncertainty, the incentivized demand response, the impact of electric vehicles, and the IPOA in the power capacity optimization for the independent microgrids, as shown in

Table 5. Influence of different factors on the capacity optimization cost of the microgrid.

Factors	Influence
Incentive DR	Capacity optimization cost of microgrid is reduced by CNY 764,406, wind and light abandonment rate is reduced by CNY 69,700, distributed power capacity is reduced, and the capacity of energy storage equipment is significantly reduced.
EV	The capacity optimization cost of the microgrid is reduced by CNY 139,633 compared with the cost when only the incentive demand response is considered, and the capacity of energy storage equipment is greatly reduced. Considering the influence of different proportions of dispatchable electric vehicles on the capacity optimization of microgrid, and considering a higher proportion of dispatchable electric vehicles, the cost of the capacity optimization of the microgrid is lower.
Uncertainty	The capacity optimization cost of the microgrid increases by RMB 348,422 compared with the cost without considering the uncertainty, and the power supply capacity also increases, but the reliability of the independent microgrid is greatly improved.
IPOA	The capacity optimization cost of the microgrid is reduced by CNY 49,601 compared with the cost of using the POA as the solution algorithm. The IPOA is effective in solving the capacity optimization model of the microgrid.

.

It can be concluded from the table that the two-layer model of power capacity optimization of the microgrid proposed in this paper performed well in terms of economy, environmental protection, and reliability. The IPOA proposed for this model also reduced the capacity optimization cost of the microgrid.

7. Conclusions

Considering the capacity optimization allocation problem of the independent microgrid, this paper proposed taking the comprehensive cost of the power capacity optimization allocation of the microgrid as the objective function, then established and analyzed the capacity optimization model of the microgrid by considering the uncertainty of the source load and the excitation DR. The main conclusions are as follows:

(1)

This paper constructed an independent microgrid capacity optimization configuration model that included electric vehicles. By introducing electric vehicles and including dispatchable electric vehicles in the incentive DR, the total cost and investment cost of the power capacity optimization of the microgrid were reduced compared with the traditional model. Because of the energy storage characteristics of dispatchable electric vehicles, the number of energy storage equipment construction configurations was significantly reduced. The example analysis found that the capacity optimization cost of the microgrid was reduced by CNY 139,633 after considering electric vehicles.

(2)

The capacity optimization allocation method of the independent microgrid proposed in this paper considered uncertainty and the DR. When the uncertainty of the source load was considered, the user comfort level was taken into account, and the uncertain influence of the shifting load and dispatchable EV load was fully considered. The optimal configuration model of the independent microgrid was further improved, and the calculated configuration results were more reasonable.

(3)

An improved pelican optimization algorithm was proposed. In view of the problems with the traditional pelican optimization algorithm, such as uneven initial population distribution and easily falling into the local optimum, improved methods such as the stochastic opposition learning strategy, disturbance inhibition factor, and Lévy flight strategy were proposed to improve the initial population and increase the convergence rate, the global search ability, and the solution stability. After the POA improved, the capacity optimization cost of the microgrid was reduced by CNY 49,601.

The independent microgrid optimization model with electric vehicles proposed in this paper optimized the power capacity of the microgrid to become more economical and reliable. However, there are still some defects. Only the independent microgrid was considered. The model construction of electric cars is not detailed enough; the optimization index is too simple. In the follow-up research, real microgrid source load data should be considered for simulation analysis. More microgrid distributed power supply operators should be added to carry out the fine management of price elasticity, etc. A weight coefficient should be added to different indexes, and its influence on the optimal value should be analyzed. The optimal allocation of power capacity, including electric vehicles under grid-connected and multi-microgrid conditions, must be considered to give full play to the potential of dispatchable electric vehicles regarding energy storage characteristics and to provide guidance for the planning and operation of the microgrid.