Charging Dispatching Strategy for Islanded Microgrid Battery-Swapping Stations

Abstract

To date, few studies have addressed the charging and discharging schedules of electric vehicle battery-swapping stations in China’s isolated microgrids. Given that battery-swapping is expected to become increasingly widespread, this study innovatively considered distributed power sources, such as wind power and photovoltaic power, to analyze battery-swapping station operation and isolated microgrid operation schedules. Considering the needs of potential future communities that might depend on electric vehicle battery-swapping stations, two scenarios were analyzed, and the most effective integer planning method was adopted. The international algorithm software program YALMIP + CPLEX was used to address the problem, and simulation results proved that the proposed model and its solution method could effectively affect the safe operation and scheduling of islanded microgrid battery-swapping stations and reduce the cost of islanded microgrid operation, with significant advantages.

1. Introduction

As of the end of March 2022, the new energy vehicles’ national market share reached 2.90%, and the new energy vehicles’ holding volume reached 8.915 million units, including 7.245 million pure electric vehicles held. The number of pure electric vehicles held has continued to rise in recent years [1]; they are expected to become the new mainstream car for sale by 2035 [2]. Electric vehicles have the advantages of cleanliness and environmental friendliness and are a key supporting technology of national energy structure optimization promotion and realization of the “carbon peak and carbon neutrality” goal [3]. However, a continued increase in the large-scale “plug and charging” behavior associated with electric vehicles is expected to increase the power grid’s peak and valley difference and affect the distribution network’s power quality and reliability [4]. As intelligent charging pile technology matures, electric vehicles, as a flexible resource with the dual attributes of elastic load and mobile energy storage, can increasingly realize interaction with the power grid via V2G (vehicle to grid) [5,6]. As electric car use gradually increases, consumers, through business aggregation, virtual power plant response time, time price, or incentive, can help suppress the grid load and power grid peak load on the one hand; on the other hand, electric cars can also be used as backup power to provide auxiliary frequency modulation and emergency power supply services, thus providing additional economic benefits to their owners [7,8,9].

The best way to achieve V2G is the battery-swapping station because battery-swapping stations are limited only by electric vehicles; battery-changing stations have certain backup batteries, and the State Grid Corporation of China’s policy mainly supports power changing and auxiliary charging.

In the literature [1], a robust two-stage optimization model, min–maximin, has been constructed to take source and load uncertainty into account; its objective is to minimize the average daily total cost of planning and configuring an integrated photovoltaic storage–charging system in a highway service area. Configuring a hybrid supercapacitor–lithium-ion battery energy-storage system on the power side could facilitate applying the system in multiple scenarios, such as smoothing photovoltaic output fluctuation and participating in net load shaving and valley filling. The literature [2] proposes a multi-timescale optimal scheduling model for microgrids considering EV (electric vehicle) resource participation. In the day-ahead scheduling phase, a portion of EV resources is combined with price-based demand response technology to optimize for comprehensive user satisfaction. The scheduling plan, based on price-based EV resources, is optimized with the objective of minimizing economic low-carbon costs with maximum flexibility satisfaction, thus determining the scheduling arrangement of adjustable resources on each side. Another portion of EV resources is combined with incentive-based demand response techniques during the intraday dispatch phase. The microgrid energy management center guides operating cost minimization, and the results show that the proposed model could reduce users’ electricity costs. None of the above literature has considered the battery-swapping model, which has not been adapted to actual future vehicle development prospects.

At present, many articles address coordinating and dispatching battery-swapping stations in China’s main network, but few address future microgrid development with regard to electric vehicles, particularly coordinating and dispatching electric vehicle battery-changing stations on an island microgrid.

On the one hand, island microgrids obtain power from distributed power supplies; on the other hand, they supply power to battery-swapping stations and ordinary users [10,11]. Power distributed through an electronic inverter converts from direct current (DC) to alternating current (AC) to supply microgrids; a power electronic rectifier allows a microgrid to function as an AC rectifier to charge batteries. Through the charging current’s load controller and timing control, a measuring device uploads the terminal voltage via a current measurement microgrid central control (MGCC). The MGCC’s adjustable terminal is an optimal solution that ensures that a microgrid’s in-plant charging current and distributed power supply power regulation operate safely and economically [12,13].

This article discusses the difference between disorderly and orderly charging on an island microgrid, distributed photovoltaic and wind power supplies, and battery-swapping stations and verifies the economy and security of island microgrids’ operation using the orderly charging mode algorithm.

2. Microgrid Battery-Swapping Station Scheduling Model

2.1. Microgrid Grading Scheduling Framework

Based on the operating characteristics of isolated island microgrids, the concept of hierarchical dispatch is significant. This study aimed to construct a scientific and efficient hierarchical scheduling framework for renewable energy sources (such as photovoltaic and wind power), as well as energy loads (such as battery-swapping stations in isolated island microgrids). A microgrid needs to achieve self-sufficiency within the islanded operation mode that emphasizes local optimal energy allocation, as well as management to support its stable operation and power supply reliability. The microgrid scheduling model used in this study is shown in Figure 1.

When distributed power sources in an islanded microgrid cannot meet the entire load demand, the system needs to obtain power from the main grid to supplement the deficiency. To accomplish this, the island operation mode is converted to a master–slave operation mode, in which the main grid becomes the main power supply, although distributed power sources still play a certain role. When this occurs, control and management of the entire microgrid are primarily handled by the MGCC to ensure the grid’s operating costs are minimized and a certain level of economy is met.

An islanded microgrid’s performance and efficiency can be improved by taking advantage of the economics and environmental friendliness of distributed power sources. Distributed power sources are able to more flexibly adapt to load changes, reduce dependence on traditional energy sources, and lower energy costs. When distributed power sources cannot meet load demand, access to the main grid ensures the microgrid’s operational stability and reliability.

This paradigm shift also guarantees the grid’s reliability and security in the event of insufficient energy supply or faults. Access to the main grid ensures that the microgrid is able to operate normally and avoids power supply interruptions or instability. In addition, the MGCC’s overall control enables coordination and optimization between distributed power sources and the main grid to maximize the load demand while reducing operating costs [14].

This article presents a steady-state analysis; the electronic model in the scheduling model was not considered. If the electronic model had been considered, a temporary analysis would have been performed, and fast optimization calculation could not have been performed [15,16]. Similarly, the measurement device model was not included in this study’s model. This study primarily considered the distributed power supply as a PV node, the power station as a PQ node, and the traditional main network as a balanced contact. The trendy non-linear equation was not considered. Non-linear models were simply processed using only the power balance equation. The specific mathematical model is shown below.

2.2. Scheduling Model

The superior MGCC controller aimed to minimize the total consumption of each distributed generation (DG) economy:

$$\min \sum _{i=1}^{{\mathrm{N}}_i}{\mathrm{P}}_{\mathrm{G}}^i(t){\mathrm{C}}_{\mathrm{G}}^i+\sum _{C=1}^{{\mathrm{N}}_C}\sum _{k=1}^{{\mathrm{N}}_k}\left|{\mathrm{P}}_C^k\left(t\right)\right|{\mathrm{C}}_C^k$$

(1)

In Formula (1), ${\mathrm{C}}_{\mathrm{G}}^i$ indicates the DGi operating cost; ${\mathrm{N}}_i$ indicates the amount of DG (distributed generation) in the number of DG; ${\mathrm{P}}_{\mathrm{G}}^i\left(t\right)$ is the power output of DG output o in time period t node i; ${\mathrm{P}}_C^k\left(t\right)$ is the energy storage device of energy storage station k in the charging power in time period t; ${\mathrm{C}}_C^k$ is the storage of energy storage station k storage and the charging and discharge operation cost of the capable C; and ${\mathrm{N}}_C$ and ${\mathrm{N}}_k$ are the number of energy storage stations C and energy storage devices k, respectively.

Model Constraints

In this study, the selection of constraints was crucial to the dispatch model’s accuracy and rationality. The mathematical model’s constraints regarding the charging and discharging states, charging and discharging power, the battery state of charge, battery replacement capability, and battery-swapping demand were set as follows.

• Charging and Discharging States Constraints

Consideration of the battery’s charging and discharging states was an important factor in the scheduling model. Batteries generally have only two states: charging and discharging.

$$X_{z,k,t} + Y_{z,k,t}\le 1$$

(2)

$$X_{z,k,t}=\left\{\begin{array}{cc}1& t\in \left[t_{z,k}^s,t_{z,k}^e\right]\\ 0& t\in \left[t_{z,k}^s,t_{z,k}^e\right]\end{array}\right.$$

(3)

$$Y_{z,k,t}=\left\{\begin{array}{c}1 t\in [t_{z,k}^s,t_{z,k}^e]\\ 0 t\in [t_{z,k}^s,t_{z,k}^e]\end{array}\right.$$

(4)

In the formula, $k\in H_z^{\left(i\right)}$; $X_{z,k,t}$ and $Y_{z,k,t}$ are 0–1 variables, and $[t_{z,k}^s,t_{z,k}^e]$ is the battery access time on battery-swapping station z, the kth charging and discharging device. This is the source of mixed integer models. Moore’s variables appeared. An increase in the number of battery changes causes a large-scale mixed integer model if a non-linear trend model is considered at the same time. There was no mature algorithm to solve; therefore, this study did not address the corresponding consideration of the trend model and only considered the power balance analysis.

2.

Charging and Discharging Power Constraint

A battery’s charging and discharging power is a measure of the amount of electrical energy that it can provide or absorb in a given time period. In the scheduling model, the battery’s power constraints were considered to ensure that the battery would not overload in a short time period.

$$P_{z,k,t}^{CH}-P_{z,k,t}^{DCH}={\tilde{P}}_{z,k,t}$$

(5)

$$\left\{\begin{array}{l}{P}_{z,k,t}^{CH}\le P_{z,k}^{NC}{X}_{z,k,t}\\ P_{z,k,t}^{DCH}\le P_{z,k}^{NDC}{Y}_{z,k,t}\end{array}\right.$$

(6)

In the formula, $k\in H_z^{\left(i\right)}$; $P_{z,k,t}^{CH}$ is the charging power of the kth charging and discharging device of the battery-swapping station z in time period t, and $P_{z,k,t}^{DCH}$ is the discharge power [17] of the kth charging and discharging device of the battery-swapping station z in time period t.

3.

Battery State of Charge Constraint

A battery’s state of charge (SOC) is the ratio of its available power to its total power. In scheduling models, a battery’s SOC constraints need to be considered to avoid overcharging or undercharging the battery.

$$\left\{\begin{array}{l}{S}_{z,k,t_{z,k}^s}=S_{z,k}^0\\ S_{z,k,t}=S_{z,k,t-1}+\frac{{\eta }^{CH}{P}_{z,k,t}^{CH}\Delta t}{B_C}-\frac{P_{z,k,t}^{DCH}\Delta t}{{\eta }^{DCH}{B}_C}\end{array}\right.$$

(7)

In the formula, $k\in H_z^{\left(i\right)}$; $S_{z,k}^0$ and $S_{z,k,t}$ are the initial state of charge of the battery on the kth charging and discharging device of the lower battery-swapping station z and the state of charge at the starting moment in time period t, respectively; ${\eta }^{CH}$ and ${\eta }^{DCH}$ are charging and discharging efficiencies, respectively; and $B_C$ is the battery’s capacity.

4.

Battery Replacement Capability Constraint

Battery replacement capacity is a battery’s ability to be replaced within a certain time frame. In the scheduling model, the battery replacement capacity’s limitations were considered to ensure that batteries could be replaced in a timely manner when needed.

$$\left\{\begin{array}{l}{A}_{z,t}=N_{Bz}\\ A_{z,t}=A_{z,t-1}+C_{z,t-1}-D_{z,t-1}\end{array}\right.$$

(8)

$$C_{z,t}=\sum _{k\in H_z^{\left(i\right)}}(⌊\frac{S_{z,k,t}}{S_U}⌋-⌊\frac{S_{z,k,t-1}}{S_U}⌋)$$

(9)

$$S_{z,k,t}\ge S_L$$

(10)

In the formula, $k\in H_z^{\left(i\right)}$; $N_{Bz}$ is the number of spare batteries at the battery-swapping station z, $C_{z,t}$ is the number of filled batteries in time period t, $D_{z,t}$ is the swapping demand at the battery-swapping station z in time period t, and $A_{z,t}$ is the number of batteries that can be used for swapping in time period t. $S_U$ and $S_L$ are the upper and lower SOC limits, respectively, and ⌊ ⌋ denotes the downward-rounding function [18,19].

5.

Battery-Swapping Demand Constraint

Battery-swapping demand is a user’s battery replacement requirement at a specific point in time. The constraints of battery-swapping demand were considered in the scheduling model to ensure that batteries could be replaced in a timely manner when needed by the user.

$$(1+\beta )D_{z,t}\le A_{z,t}$$

(11)

$$N_{Bz}\le A_{z,t^{ }}$$

(12)

In the formula, $\beta$ is the margin of the number of batteries available to be switched, primarily to ensure that the swapping demand under the prediction error can be met. Considering the need to maintain the battery-swapping station’s normal operation in the next dispatch cycle, the number of available swapping batteries at the end of the current dispatch cycle should be at least equal to the number of standby batteries.

3. Example and Simulation Results Analysis

3.1. Solution Method

This model’s control variable was the number of replacement battery-swapping stations (a mixed integer model). Its objective function was the system’s operating cost; therefore, it was linear. The whole mathematical model used was a mixed integer linear model, which was solved using IBM’s world-leading CPLEX software (version 12.10.01) [20] to increase the speed of the solution. CPLEX has a strong solving ability for mixed integer linear models; its core algorithm is the Benders algorithm, first proposed by J.F. Benders in 1962. The Benders decomposition algorithm decomposes a planning problem with complex variables into linear and integer programming, decomposes the main problem and subproblems by cutting the plane, and then solves the optimal value using iterative methods. Benders’ decomposition algorithm is commonly used to compute difficult problems such as least integer nonlinear programming problems and stochastic programming problems. Theoretically, the Benders algorithm is able to solve most integer programming problems. In practice, it is mainly used to optimize a problem in terms of the number of operations and computation time required to solve it.

This study’s algorithm flowchart is shown below as Figure 2.

3.2. Test Platform and Algorithmic System

In this study, the microgrid scheduling model was developed using the globally renowned YALMIP modeling platform [16]; the model was solved by directly invoking the CPLEX solver through the internationally used optimization box OPTIToolbox [21]. The program development background was MATLAB R2021b, OPTIToolbox version v5.11, YALMIP version 20220204, and IPOPT version 3.11.8; the test system’s computer environment was the Win10 64-bit operating system and a 2.7 GHz Intel quad-core i7 CPU.

The above modeling platforms and tools have been widely used in studies published by internationally renowned journals and are known for their programming simplicity, high efficiency, fast running speed, strong solving capability for mixed integer linear models, and convenient statistical analysis.

IEEE 14 node arithmetic is commonly used to evaluate power systems’ performance and to study various power systems problems. Figure 3 shows a visual topology of IEEE 14-node calculus, which consists of 14 nodes and 20 branches. Each node represents a generator or a load and is connected through branch circuits. It is a complex power system model used to simulate the real operation of power networks [22].

The algorithm is commonly used in research addressing stability analysis, power flow analysis, and current calculation of power systems. Parameters such as voltage stability, line loading, power balance, and a grid’s current distribution can be assessed by simulating and analyzing the algorithm.

In this paper, node 1 represents the main grid, node 4 represents the DG1 as photovoltaic generator, node 8 represents the DG2 as wind turbine generator, node 9 and 13 represents battery-swapping stations.

Table 1. Parameters of DG units.

Name	Type	Position	Smax (pu)	Cost (CYN/kWh)
DG1	Photovoltaic generator	4	1	0.158
PCC	Main grid	1	\	1.120
DG2	Wind turbine generator	8	1	0
BSS1/BSS2	Battery-swapping station	9,13	\	0.005

shows this study’s distribution and output of each DG in the IEEE 14 node of this study’s model.

3.3. DG Output Analysis

Figure 4 presents curves for wind turbine generation and photovoltaic generation [23], which show that photovoltaic and wind turbine generators were highly volatile within a day and did not match the base load’s load demand curve (Figure 5). The high base load hours were not necessarily the most sufficient hours for photovoltaic and wind energy generation. For example, during daytime, when photovoltaic generation is highest, loads were not the highest, which led to user overvoltage on the distribution network. At night, when photovoltaic generation is lowest, user load demand was highest, thus resulting in user undervoltage. Therefore, islanded microgrids often need to be converted to operate in master–slave mode, supplementing and transferring energy from the main grid.

3.4. Simulation Results and Analysis

We analyzed two scenarios to verify the model’s correctness through example simulations. Scenario 1 comprised disordered scheduling of a battery-swapping station under island operation, and Scenario 2 comprised orderly scheduling of a battery-swapping station under island operation. Battery-swapping station loads, as well as battery-swapping demand, are shown in Figure 6.

In Scenario 1, battery-swapping station loads were charging during the peak hours of the base load from 14:00 to 23:00, which would lead to problems such as voltage reduction, network loss increase, and power imbalance. Users switch to electricity according to their own needs, and the swapping station arranges charging as soon as possible if it does not have spare batteries, which would lead to a concentration of charging and a corresponding concentration of load. However, in Scenario 2, battery-swapping loads were charged in an orderly manner, and optimizations were introduced; battery-swapping stations were charging during the load valley—that is, 24:00–05:00. It is obvious that, although charging demand occurs at peak hours, the existence of spare batteries to meet power exchange demands would have the effect of shaving demand peaks and filling load valleys, thus improving the voltage level and reducing line network loss. Ultimately, Scenario 2 met users’ battery-swapping demands, facilitated the grid’s economy and security, and reduced the microgrid’s operating cost.

Under the premise that the battery-swapping station could meet user demands, Scenario 2 reduced expenses associated with obtaining high-cost power from the main grid while shaving demand peaks and filling load valleys; this would ultimately achieve a win–win situation for the user and the power grid and achieve a better balance between security and economy.

In the planning stage, if the node power meets the grid’s constraints, the simple treatment of power balance is more or less the same as the result of considering the tidal nonlinear equation. The result of peak shaving is also applicable to the constraints of the tidal nonlinear equation and has the same positive effect on voltage stabilization; the reduction in network loss is also different. Therefore, this study’s proposed model, as well as its problem-solving method, are, objectively speaking, in line with production; that is to say, the model and solution method proposed in this paper are objectively in line with production and operation realities.

If trending nonlinear equations were incorporated, the model would become a large-scale mixed integer nonlinear model that has not yet been solved using a mature solution algorithm; the hierarchical algorithm is the only engineering algorithm that can be solved, but its speed of operation is so slow that it would be impossible to practicality compute. This study’s algorithm can be computed in seconds because it uses the world’s leading CPLEX mathematical software and efficient YALMIP modeling software and because it simplifies the load equation. Compared to hierarchical algorithms that require an hour or more to solve, this study’s algorithm is more applicable to practical engineering use. The robustness of this study’s proposed algorithm is, in fact, also the robustness of the CPLEX mathematical software because this study’s mathematical model is in line with the basic logic of planning algorithms. Therefore, as there is no problem in terms of the algorithm’s robustness, comprehensive judgment indicates that the algorithms and models proposed in this study are realistic and support the validity of the algorithm’s high efficiency.

Table 2. Operating costs of DG units.

Scenario		1	2
DG Operating cost (CYN)	DG1	26.9	23.8
	DG2	79.3	78.7
	Main grid	985.3	0
Total cost (CYN)		1091.5	102.5

shows the cost of each generator output; under Scenario 2’s orderly charging, there was no need to purchase power from the main grid, so no cost was incurred; the microgrid could be operated in an islanded manner without incurring any additional cost. At the same time, photovoltaic and wind turbine generator outputs decreased; thus, Scenario 2 verified the economy of this study’s proposed method.

4. Conclusions

Under the islanded microgrid, the battery-swapping station’s charging demand, left uncontrolled, resulted in peaks on peaks, and the DG output of the islanded microgrid was far from meeting the load demand, which increased the cost of purchasing power from the main grid, resulting in poor economy. At the same time, due to the load increase, the line loss, the network loss, and the voltage level were poorly affected, which was not conducive to the power grid’s security. However, under the orderly battery-swapping station, charging was concentrated in the grid’s low period, taking advantage of its backup batteries. Demand was concentrated in the grid’s trough period; when the load was low, the grid could fully meet the battery-swapping station’s charging demand without causing the grid’s performance to decrease due to high load. This not only solved the problem of grid economy and reduced the cost of purchasing power from the main grid but also effectively improved the grid’s security. Charging during the low load valley period helped avoid the load peak and reduced the grid’s load, thus reducing the problems of line loss, network loss, and voltage fluctuation.

The optimization method and optimization model proposed in this study were simulated using the YALMIP + CPLEX solution platform with 14 nodes of arithmetic cases to prove the feasibility and effectiveness of an ordered power exchange in an islanded microgrid. This method introduces new ideas to address the charging demands of battery-swapping stations and provides a reliable solution to ensure the grid’s stability and economy. In addition, the method provides a useful reference for future research and practice regarding isolated island microgrids.

The potential for further research in this field remains untapped. Subsequent research could explore how to further reduce the number of backup batteries while ensuring safety and stability. This might involve optimizing and adjusting the capacity of standby batteries, the efficiency of the energy storage, etc., to maximize the use of available standby battery resources. Reasonably arranging the discharging time and volume of battery-swapping stations can better meet load demand, reduce power purchases from the main grid, and improve grid economics; smarter and more efficient grid operation and management can be achieved using advanced algorithms and technologies in combination with rational grid planning and scheduling strategies. This will be important for the future development of islanded microgrids and sustainable energy utilization.