Optimal sizing of electric vehicle charging station considering charging scheduling strategy Authors: Wenbin Shao, Dingding Hu Published: 2026 ABSTRACT AbstractThe widespread use of electric vehicles (EVs) strongly depends on the construction of EV charging stations (EVCSs) that should fulfill charging requests while considering various economic issues. This paper investigates the optimal sizing of an EVCS integrated with renewable energy generators, an energy storage system (ESS), and multiple types of chargers. The optimal sizing problem is formulated as a mixed-integer linear program that takes seasonal variations of renewable energy generation into account, while aiming to minimize the equivalent annual total cost of the EVCS. Then, a novel charging scheduling strategy for EVs is developed by considering charger utilization efficiency, thereby reducing unnecessary expenses. Finally, different case studies are analyzed to verify the effectiveness and economic performance of the proposed strategy. The results indicate that incorporating EV charging scheduling into the optimal sizing problem can improve the economic performance of the EVCS.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-53285-x.Keywords: Electric vehicle charging station, Optimal sizing, Renewable energy, Charging behavior, Scheduling strategySubject terms: Energy infrastructure, Energy storage Introduction The excessive utilization of fuel-powered vehicles is an important factor in climate change and air pollution1, causing a series of problems such as climate change, air pollution, and energy shortages2. Since the Netherlands proposed a plan to ban the sale of gasoline vehicles, France, the United Kingdom, Sweden and other countries have announced plans to ban the sale of traditional fuel vehicles3. EVs, as a highly-regarded green mode of transportation, can not only reduce the consumption of fossil fuels and air pollution4, but also serve as an important means to achieve the global net-zero emissions goal5. According to the Global Electric Vehicle Outlook 2024, nearly 14 million new EV registrations will be made globally in 2023, with global EV sales accounting for about 18% of total vehicle sales, an increase of 3.5 million units from 2022 and a year-on-year increase of 35%6. With the continuous increase in the number of EVs, the shortage of charging infrastructure and the problem of power grid overload have become increasingly prominent7. Therefore, building EVCSs to meet the main charging needs of EVs is of great significance for promoting EVs on a larger scale. Compared with traditional vehicles, EVs have problems such as range anxiety, charging anxiety, inadequate charging infrastructure, and grid overload8,9, while the simultaneous charging of large-scale EVs will put pressure on the grid and jeopardize the safe operation of the power system10. With the promotion and implementation of multiple renewable energy policies, the utilization and development of renewable energy have been improved11. Energy storage technology can regulate grid loads, balance power supply and demand, alleviate grid pressure, and help address range anxiety through vehicle-to-grid (V2G) and grid-to-vehicle (G2V) technologies12. Although renewable energy generation is widely used in power systems to support supply-demand balance, its inherent uncertainty may hinder effective supply-demand coordination and lead to mismatches with energy demand13. The ESS can not only store electricity and improve energy utilization, but also enhance grid stability and support the development of EVs and other electrification technologies. The integration of these components in EVCSs helps reduce electricity purchase cost and environmental pollution. Although the planning of EVCSs has attracted widespread attention and research, there is a shortage of charging infrastructure for EVs, and the research on the incorporation of EV charging scheduling in the scale optimization model of EVCSs has not been clearly explored in the literature. At present, the research on the EV charging scheduling mainly focuses on its impact on the economy and safety of the power grid14. For instance, Chaudhari et al.15 proposed an intelligent charge-discharge scheduling method to reduce the overload of power grid transformers. Zhang et al.16 proposed a two-stage optimization strategy for EVs charging and discharging according to the user’s charging and discharging demand response, so as to reduce the charging cost and reduce the load on the power grid. Izci et al.17 designed a new load frequency control (LFC) method to achieve real-time response of power systems with minimal overshoot. Furthermore, various EV scheduling schemes have been proposed to reduce the power losses18. In practical applications, the charging time and power of EVs can be adjusted according to the electricity price and EV charging information, which is beneficial to the planning and operating of EVCSs19. Although some studies have addressed the importance of incorporating diverse charger types and distributed energy resources (DERs) into EVCS design, the complexity of EV users’ charger-selection behavior is often simplified in operational modeling20,21. Specifically, many operational models rely on the assumption that EV users are assigned or uniformly select a charger based on the minimal rated power sufficient to complete their charging task within the required timeframe22,23. In practice, EV users’ decisions are likely to be influenced by individual preferences and dynamic charging-station conditions24,25. This simplification may not accurately reflect the impact of charger diversity on key operational metrics such as wait time, equipment utilization rate, and operating cost. Furthermore, the integration of multiple charger types is crucial for meeting diverse charging demands. Chargers with higher power output naturally entail greater investment costs but offer faster charging capabilities. Consequently, determining the optimal mix and quantity of different charger types within an EVCS presents a significant planning challenge. At present, considerable research works on the optimal sizing of EVCSs have been conducted from different perspectives. Firstly, various optimization objectives have been proposed according to different research scenarios. Wu et al.26 proposed a vehicle-road-network collaborative operation framework and a spatiotemporal scheduling method based on user-personalized navigation to optimize the charging load of EVs and reduce the travel cost of EV users. Chaudhari et al.27 introduced time of use (TOU) wholesale electricity pricing into the planning of EVCSs with PV and ESS, which aims to minimize the operating cost. Xie et al.28 proposed a two-layer optimal scheduling strategy for the “distribution network-charging station” system, which optimizes the peak shaving cooperation between EVCSs. Li et al.29 proposed an improved road model and a bi-level planning method for EVCS, aiming to optimize the travel cost of EVs and the operation of the coupling network, and reduce the investment cost of charging stations. In addition to economic indicators, many studies have developed multi-objective optimization models that incorporate environmental-benefit and reliability indicators to investigate capacity configuration, thereby compensating for the limitations of single-objective optimization. Xu et al.30 considered power supply reliability maximization and investment cost minimization as objective functions to size a hybrid energy system with wind, PV, and water. Zhang et al.31 investigated the planning problem of multi-load demand response and its uncertainty by maximizing annual expected profits, minimizing carbon emissions, and minimizing power fluctuations in connecting lines as multi-objective functions. Secondly, in terms of optimization algorithms, heuristic algorithms and mathematical methods are widely utilized in solving the sizing problem for EVCSs. The biggest advantage of heuristic algorithms is that they are relatively simple when dealing with complex models with a large number of parameters, but sacrifice optimality for speed32. Genetic algorithm (GA)33, simulated annealing (SA)34, particle swarm optimization (PSO)35, ant colony optimization (ACO)36 are utilized frequently. Compared with heuristic algorithms, mathematical methods are more capable of obtaining globally optimal solutions, but they are often less practical for highly complex problems37. Furthermore, commercial optimization solvers such as CPLEX, MOSEK, and GUROBI are also widely used in the literature38–40. Third, from the perspective of EV charging behavior, previous studies have used various EV charging demand models under different scenarios. To simplify the modeling process, EV charging demand is often treated as a constant, which does not consider charging start and end times. Although this method reduces the computational complexities, the characteristics of charging behavior of EV users are neglected. According to Ref. 40, EV users tend to adjust their charging patterns and charging levels in response to the price signals provided by the EVCS so as to minimize energy cost. Thus, the charging demand of EV users varies with the price tariff. Jiang et al.41 predicted public charging demand from travel trajectories based on the heterogeneity of EV users’ charging decisions. Huang et al.42 optimized the location of specific charging stations based on travel congestion, charging station congestion, and charging price elasticity, so as to reduce the operation and maintenance costs of charging stations and increase reasonable profits. In addition, Monte Carlo simulation (MCS) and congestion analysis based on queuing theory are common methods for dealing with the potential uncertainty of EV charging behavior43,44. While these studies have advanced the understanding of EVCS sizing, we compare and summarize this study with previous work to guide future research, as shown in Table 1. Table 1A comparative summary of this study and previous works.PaperMulti-types of chargersDifferent seasonsDERs allocationEVCS costsUser costsChaudhari et al. (2017)×× √ √ ×Luo et al. (2018)×× √ √ ×Huang et al. (2020)×××× √ Li et al. (2020)××× √ √ Xu et al. (2020)×× √ √ ×Dong et al. (2021)××× √ ×Sun et al. (2021)×× √ √ ×Yang et al. (2021)×× √ √ ×Xie et al. (2024)××× √ ×Zhang et al. (2024)×× √ √ ×Wu et al. (2025)××× √ √ This study √ √ √ √ √ √: Yes. ×: No. A comparative summary of this study and previous works. √: Yes. ×: No. Based on the aforementioned discussion, it is seen that size optimization methodologies of EVCSs still have to be improved, and the research on EV charging scheduling needs to be further deepened. In this study, a novel EV charging scheduling strategy is proposed to improve the utilization rate of chargers and promote the station economic operation, which works together with the optimal sizing of EVCSs. Based on the above consideration, this work addresses the optimal sizing problems of EVCSs with the objective of minimizing annual total cost. The non-dispatchable renewable energy resources are considered in the design of EVCSs to meet the desired energy. Unlike existing studies, the EVCS acts as a self-interested entity in our design, taking account of power transactions and the EV charging behavior during operation. In this case, the EVCS is connected to the power grid, allowing surplus energy to be sold back to the grid for profit. Most studies assume that EV load is fixed in charging-station design, whereas few studies fully exploit the technical characteristics of EV charging. In this study, different types of chargers are considered to satisfy diverse charging demands, and EV charging patterns are adjusted through the proposed scheduling strategy to reduce the number of chargers required. Therefore, this study provides practical guidance and technical support for policy makers when deciding to build a new charging station. The main contributions of this study can be summarized as follows: (1) A generic framework is developed for the optimal sizing of EVCSs integrated renewable generators, ESS, and multi-types of chargers. The sizing problem is modeled as a mixed-integer linear program (MILP) with the consideration of seasonal variations of renewable energy generation. (2) A novel charging scheduling strategy is proposed to adjust the load curve, which can not only improve the utilization rate of chargers, but also promote the station economic operation. Moreover, the real-world EV charging data from public charging stations are used to determine the realistic characteristics of EV charging behavior for the estimation of EV charging demand. (3) In order to demonstrate the feasibility and effectiveness of the proposed model and strategy, different cases have been used for comparison. Meanwhile, the influence of the number of EVs and the capacity price of ESS on the economy of the charging station are discussed in depth, providing a valuable reference for the construction of the new charging station. The remainder of this paper is organized as follows. Section 2 presents the methodology, including the system framework, model formulation, and charging scheduling strategy. Section 3 reports the experimental results and discussion. Finally, Sect. 4 concludes the paper and outlines possible future research directions. Methodology The framework for the optimal sizing of the EVCS is illustrated in Fig. 1. First, to ensure the applicability of our study in practical scenarios, the models used in this study are validated by historical data of a practical EVCS. The output power of renewable energy resources and the characteristics of EV charging behavior can be derived based on the collected data. Second, a charging scheduling strategy is proposed to determine the optimal number of chargers to be installed for each charger type, and this strategy is embedded into the EVCS sizing problem. Hence, the charging time schedule of EVs strongly interacts with decisions on the number and type of chargers to be installed. Finally, an optimization model is proposed to determine the optimal sizing of the EVCS. To avoid excessive computation, a 4-day equivalent model is built and applied, which can represent the seasonal variability of the operation of the EVCS throughout the year. Thus, the seasonal characteristics of renewable energy output and the conventional load can be integrated into the optimal sizing problem of EVCSs. Framework for the optimal sizing of the EV charging station. Architecture model of the EVCS Figure 2 presents a schematic view of the studied EVCS. The station trades energy with the grid via a distribution transformer. The control center can not only determine whether to purchase electricity from or sell electricity to the main grid according to electricity price and load demand, but also execute the scheduling plan. Schematic of the EVCS. In such architecture, a collection of chargers is connected to the same bus as the integrated EVCS components, namely WT, PV and ESS. The efficiency of renewable energy (wind and solar) usage can be improved through the coordination of various components. The description and equations for these components are provided in the following subsection. WT power output model The power generation of a WT mainly depends on its rated power and the wind speed, and can be expressed as the following piecewise function. where is the WT rated power. , , and are the actual wind speed, rated wind speed, cut-in and cut-out wind speed of WT at timeslot t, respectively. PV power output model The output power from PV is strongly affected by cell temperature and sunlight intensity, and can be expressed as follows45. where denotes the rated power of PV under standard scenario, denotes the actual sunlight intensity received by PV, is the power reduction factor, is the temperature coefficient, is the cell temperature, is the rated temperature of PV cells under standard scenario. where is ambient temperature and NOCT is normal operation cell temperature. Charging and discharging model of ESS ESS can smooth output fluctuation and its operation status depends on whether the ESS is being charged or discharged. The stored energy of ESS at each timeslot is shown as follows. where denotes the energy level in ESS at timeslot t, σ denotes the self-discharge rate, which is assumed to be 0.0446. and denote the ESS charging and discharging power at timeslot t, respectively. and denote the charging and discharging efficiency of ESS, respectively. Δ T denotes the length of each timeslot. Modeling of EV charging behaviors Considering EV charging characteristics, the feature information of EV charging behavior is extracted from real charging data, and maximum likelihood estimation is then used to determine the corresponding distribution curves. To this end, the probability of charging start time, charging duration, and charging demand are profoundly analyzed in this section. In this study, historical EV charging data are collected from the Internet of Vehicles Merchant Services Platform. The dataset records charging events during 2019–2020, including charging start time, charging duration, and the energy of each charging event. Unlike private charging stations with fixed charging modes, the charging behavior of EV users at the pubic EVCSs is random, and factors such as vehicle status, charging facility status, charging fees, and satisfaction with charging facilities have a significant impact on the charging behavior of EV users47. Based on the collected data, the EV charging start time is assumed to follow a normal distribution, which is expressed as follows. where , . Supplementary Figure S1 shows the distribution fitting chart on the EV charging start time. In Supplementary Figure S1, the red line is the curve fitting result of EV charging start time, and the probabilities are shown in pink histograms. It can be observed that EV users usually choose to charge at approximately 15.62 h in the public charging station. Here, the charging duration for each charging event is assumed to have an exponential distribution with its probability density function as follows. where . It should be noted that the distribution is calculated based on 10-minute bins48. The distribution fitting chart of charging duration is illustrated in Supplementary Figure S2. It can be seen from the above figures that more than 65% of all the EVs spent less than one hour to charge each time. In addition, EV charging energy demand is assumed to follow the Weibull distribution. The probability density function can be expressed as follows. where shape parameter k=1.409, scale parameter λ =18.315. Supplementary Figure S3 gives the distribution fitting chart of the charging energy demand of EVs. Charging scheduling strategy In the optimal scale problem of EVCSs, the EV charging demand varies greatly in different time periods, which makes the optimization model complex and difficult to solve49. To improve computational efficiency, a day is equivalently discretized into T time slots, and the length of each time slot is set as Δ T. Therefore, the charging start time and charging end time of each EV can be expressed as follows: where and represent the serial number of the time slot when EV i is connected to the charger and disconnected from the charger, respectively. represents the charging end time of the EV i. denotes the rounding to the previous integer, denotes the rounding to the next integer. Due to the operation potential and technical characteristics of EV charging demand and charging time, flexible adjustment in the time dimension can optimize the EVCSs cost structure and reduce costs51. Accordingly, we propose a charging scheduling strategy that satisfies charging demand with fewer chargers by adjusting the states of charging sessions, which is beneficial for the EVCS sizing problem. Supplementary Figure S4 illustrates a schematic comparison between the FCFS charging mode and the proposed charging scheduling strategy, highlighting how the latter improves charger utilization. The green bar represents the charging time period for each EV. The upward arrow indicates the start of charging, and the downward arrow indicates the end of charging. Charging must occur before the departure time of the EV (vertical dashed line). The following is a detailed explanation of each case described: Supplementary Figure S4 (a) shows the FCFS charging mode. It represents the traditional FCFS approach, where EVs start charging immediately upon arrival without any adjustment to the charging time. EVs are served in the order of their arrival. For example, EV1 arrives first and starts charging right away; EV2 arrives next and begins charging after EV1 is completed or when another charger is available. This mode typically results in unbalanced utilization of chargers. For instance, multiple EVs may arrive within a short period, leading to peak demand and requiring more HPCs to avoid delays. This increases the risk of charger congestion during peak hours, as charging tasks are not shifted to non-peak periods, which may raise infrastructure costs and grid pressure. This is consistent with the finding that FCFS requires 15 HPCs in Cases 1–3, whereas Case 4, which adopts the proposed strategy, only needs 10 HPCs. Supplementary Figure S4 (b) presents the proposed charging scheduling strategy. It demonstrates the new scheduling strategy, which dynamically adjusts the charging time slots of EVs to improve charger utilization while ensuring that all EVs complete charging before their departure. EVs do not necessarily start charging immediately upon arrival. Instead, the control center adjusts the charging time to less crowded periods based on factors such as charger availability, renewable energy output, and electricity prices. In this mode, the green bars are redistributed among time slots, and the overlapping charging tasks in FCFS are dispersed, reducing peak demand. For example, an EV that would have been charged during peak hours in FCFS is rescheduled to a later, less busy period, allowing the same charger to power more EVs in sequence. By maximizing the overlap of charging tasks across time slots without conflicts, the number of required chargers is reduced, especially HPCs. Grid pressure is lowered by smoothing the load curve (consistent with the result in Case 4 where the peak demand is 30% lower than that in Case 3). Economic efficiency is improved by matching charging with periods of high renewable energy production or low electricity prices. The comparison in Supplementary Figure S4 highlights that the proposed scheduling strategy avoids the inefficiency of FCFS by strategically delaying non-urgent charging tasks. This reduces the total number of required chargers; for example, in Case 4, the number of HPCs decreases from 15 to 10, thereby lowering operational costs while still ensuring that all EVs complete charging before their departure deadlines. In order to implement the charging scheduling strategy, EV users need to send charging information such as arrival time, departure time and expected charging energy to EVCSs in advance. EVs that cannot be served immediately upon arrival are rescheduled within their available parking duration. Based on the EV arrival time, departure time, expected charging energy, and charger availability, the control center assigns a suitable charger and charging time slot to ensure that charging is completed before departure. The vector X that represents the proposed charging scheduling strategy given by (13) defines the real charging state at each charging point. where is a binary variable that reflects whether EV i is charging at timeslot t. If , it indicates that EV i is charging at timeslot t; if , it means that the EV i is not charging at timeslot t. When an EV is to be charged, the number of timeslots required to complete EV charging task can be affected by the rated charging power of selected chargers. Although a charger with low rated charging power incurs the low installment cost, its charging duration is relatively long, which may result in the charging congestion phenomenon. Therefore, three typical kinds of chargers have been taken into consideration in this study. According to their rated charging power, these facilities are classified as low-power chargers (LPCs), normal-power chargers (NPCs), and high-power chargers (HPCs), respectively. Let j ∈ Jrepresent the index of charger types. Each EV is assigned to exactly one charger when charging begins. The classification criteria for EV charging options are shown as follows. where is a binary variable indicating whether EV i is assigned to charger j. is the rated charging power of charger j. In order to meet the expected charging energy demand of EV users and avoid overcharging, the following constraint must be satisfied. where is the expected charging energy of EV i, is the charge efficiency of EVs. Here, we have selected the FCFS manner as a benchmark for comparison. The FCFS manner differs from the proposed charging scheduling strategy in the sense that it sets priorities based on EV arrival time and then does not participate in the scheduling planning during the charging process. In this case, EVs are uncontrollable and cannot be scheduled. The following constraint shows this difference. Optimization model Objective function The aim of the sizing problem in this study is to minimize the equivalent annual total cost of the entire EVCS, including annual investment cost of various components, annual operation and maintenance cost, annual transaction cost between the EVCS and the power grid, and the revenue from providing charging services to EVs. The decision variables in the EVCS optimal sizing problem are the number of WTs, PV panels and each type of chargers, the energy capacity of ESS, EV charging state, the ESS charging and discharging power, and the exchanged power between EVCS and the grid at each timeslot. The objective function is expressed as follows. where the constants 4 and 96 represent the number of seasons in a year and time slots in a day, respectively. Therefore, 384 time slots are included in the objective function. Furthermore, the constant 91.25 represents the number of days in each season. (1) Annual investment cost of components in the EVCS. where is the cost of chargers; is the cost of WTs; is the cost of PV panels; is the cost of ESS. The investment cost of chargers can be described as follows. where r represents the discount rate. denotes the service lives of chargers, denotes the investment cost of each charger j, denotes the installation number of each charger j. The investment cost of WT can be described as follows. where denotes the service lives of WT, is the investment cost of each WT, is the number of WTs. The investment cost of PV generators can be described as follows. where denotes the service lives of PV, is the investment cost of each PV panel, is the number of PV panels. The investment cost of ESS can be described as follows. where denotes the service lives of ESS, is the cost of unit capacity energy storage, is the capacity of ESS. The formulas of , and in each season have the same symbolic meaning and are calculated in the same way. Thus, the subscript d can be omitted to make the description concise. Let , and represent daily operation and maintenance cost, daily transaction cost between the EVCS and the power grid, and the daily revenue from providing charging services, respectively. (2) Daily operation and maintenance cost of EVCS. where denotes the annual maintenance cost of each charger j., and denote the annual operation and maintenance coefficients of the WT, PV and ESS, respectively. (3) Daily transaction cost between EVCS and the power grid. where and represent the price of purchasing and selling electricity at timeslot t, respectively. and represent the quantity of purchasing and selling electricity at timeslot t, respectively. , , , , and represent the conventional electricity consumption, the EV charging demand, energy charged from ESS, energy supplied by WT, energy supplied by PV, energy discharged from ESS. If , , otherwise, . (4) Daily revenue from providing charging services. The revenue from providing charging services for EVs consists of two components, namely the electricity charge based on the TOU electricity price and the charging service fee50, which is described as follows. where N is the number of EVs. is the electricity price at the public charging station at time slot t, is the charging service fee at time slot t, and is the total charging energy of all EVs at time slot t. and denote the charging and discharging efficiency of ESS, respectively. Constraints The constraints in this subsection are mainly used to restrict the states of the overall EVCS and its components during the scheduling process. As with the expressions above, the subscript d will be removed for simplicity of description. (1) Supply-demand balance constraints. During each timeslot t, the EVCS must maintain the power balance between demand and supply. This equality constraint can be defined as follows. where denotes the conventional load without charging load at timeslot t. is the transmission power between EVCSs and the grid. When purchasing power from the grid, is positive. When selling power from the grid, is negative. and are the binary variables representing the charging and discharging state of ESS, respectively. (2) Requirements for each type of chargers. In the charging station, one charger can be connected to multiple EVs, but only one EV can be charged in each timeslot. Therefore, sufficient chargers should be installed to meet the corresponding charging demands at each timeslot. Constraints for each type of charger are established as follows. where represents the number of chargers of type j. (3) Constraints of ESS capacity. The SOC of ESS in each timeslot should be limited to the allowable range. where denotes the SOC of the ESS at each timeslot, which should remain within the allowable range. and are the upper and lower limits of the SOC of ESS, respectively. In this study, the lower limit of ESS charging state is set to 0.1, and the upper limit is set to 0.9 51. Moreover, the SOC of the ESS at the end of the day should be the same as that at the beginning of the day, which prevents the accumulation of energy in the ESS. This can be expressed as follows. (4) Regulations on the output power of components. The output power of all components at each timeslot must remain within a specified range. where denotes the rated power of the installed ESS. These constraints also mean that simultaneous charging and discharging of ESS at each timeslot is not permitted. (5) Transmission power constraints. To ensure the safety of the grid operation, the actual supplied and consumed power from the grid is shown as follows. where denotes the maximum transmission power. Experimental results and discussion Data and parameter setup To verify the effectiveness of the proposed scheduling strategy for EV charging, a practical EVCS in Anhui, China, has been selected as a test system. The proposed optimization model is built in the MATLAB R2018b environment and implemented on a computer with an Intel (R) Core (TM) i5-10500 CPU @ 3.10 GHz and 16 GB RAM. We assume that 100 EVs are served by the EVCS and simulate realistic EV charging situations based on the probability density function. Furthermore, the influences of seasonal characteristics of renewable energy output on the EVCS sizing optimization are considered in this work. The output curves of WT and PV generation are calculated using local meteorological data (wind speed and solar radiation) and the corresponding mathematical models. The duration of each time slot is set to 15 min, whereas the collected meteorological data and electricity prices are given on an hourly basis; therefore, the power and price are assumed to remain constant within each hour. The electricity transaction between the EVCS and the power grid is based on the TOU electricity price, as shown in Supplementary Figure S5. In this study, the sale price to the power grid is assumed to be 10% lower than the corresponding purchase price52. In addition, the charging price for EV users consists of two parts, namely the time-varying electricity price and a fixed charging service fee. The flat line at 0.8 yuan/kWh in Supplementary Figure S5 represents the charging service fee, whereas the TOU curve represents the electricity price. The output power and economic metrics for each type of charger are presented in Supplementary Table S1. According to relevant literature53, other parameters required in this study are shown in Supplementary Table S2. To compare and analyze the effects of the proposed strategy and model on the optimization result, four different cases are presented in this study, as listed in Supplementary Table S3. In Case 1, the total energy demand of the EVCS is fed from the grid. At this point, constraints concerning the transmission power between EVCS and the grid are not considered. Besides, first-come-first-serve (FCFS) manner is adopted for the charging of EVs. Compared with Case 1, the EVCS is isolated from the power grid in Case 2, where the entire energy demand is supplied by renewable energy. The application of ESS is also considered to achieve the balance of power supply and demand. The EV charging mode remains the same as Case 1. In Case 3, WT, PV and ESS are applied in the EVCS to provide cheap and safe energy. Additionally, the EVCS can be connected to the power grid, which makes the operation of the EVCS more flexible. Meanwhile, EV charging still adopts the FCFS manner. To highlight the role of the proposed method, the proposed charging scheduling strategy is used to charge EVs in Case 4. In addition, the energy demand is supplied by renewable energy and the power grid, and the excess energy is allowed to be sold to the grid. Optimal configuration results In this study, the optimal sizing problem of the EV fast charging station is modeled as a MILP. YALMIP platform with CPLEX commercial solver is used to solve the optimization model. The optimal solutions for the resource planning of an EVCS for each case are shown in Table 2. The annual cost of each component and the equivalent annual total cost of the whole EVCS under different cases are listed in Table 3. Table 2Optimal sizing results under different cases.Number of LPCsCase 1Case 2Case 3Case 41111Number of NPCs2222Number of HPCs15151510Number of WTs/403636Number of PVs/153939Capacity of ESS (kWh)/326.59331.34152.39Computational time (min)4.384.784.634.97 Optimal sizing results under different cases. Table 3Economic results under different cases.Charger cost (yuan)Case 1Case 2Case 3Case 4108,694108,694108,69474,727WT cost (yuan)–1,251,6271,126,4641,126,464PV cost (yuan)–332,463762,710762,710ESS cost (yuan)–56,94857,77526,572Operation and maintenance cost (yuan)80,000316,262355,985323,645Cost of exchanging energy with the power grid (yuan)4,144,308–−2,455,674−2,465,243Revenue from EV charging939,021939,455940,773924,627Annual total cost (yuan)3,393,9811,126,541−984,817−1,075,751 Economic results under different cases. As can be seen from Table 2, the choices for the number and types of installed chargers are the same in the first three cases because they adopt the same charging mode. Compared to the FCFS manner, the number of high-power chargers installed can be reduced significantly using the proposed charging scheduling strategy. In addition, for all cases, the number of low-power chargers and normal-power chargers installed is identical. The low-power and normal-power chargers are used to charge EVs with lower charging demand. This will contribute to the minimization of the unnecessary expenses for relatively expensive high-power chargers. Compared with Case 1, the annual total cost of Case 2 is reduced by nearly 67% (shown in Table 3). This result is consistent with our expectations because renewable generation, especially PV, has a lower life-cycle cost than purchasing electricity from the grid. However, the energy supply in Case 2 is unstable due to the isolation from the grid, which results in less flexibility of power generation. The comparison of the annual total cost in Case 2 and Case 3 shows that the remaining energy can be sold to the power grid for profits when the output power of renewable energy meets the requirements of conventional loads and EV charging. In other words, the annual total cost of Case 2 is significantly higher than that of Case 3 because the remaining energy not sold to the power grid is wasted. It is worth noting that the annual total costs of Cases 3 and 4 are negative, which indicates that the annual total costs of the EVCS are lower than the revenues obtained from selling energy to the power grid and providing charging services to EVs. Compared with Case 3, adopting the proposed charging scheduling strategy can make the annual profit of Case 4 increase nearly 9%. From an economic perspective, Case 4 performs best, followed by Case 3, Case 2, and Case 1 successively. Implementation of the charging scheduling strategy not only reduces the number of high-power chargers and the ESS capacity, but also improves the economic performance of the EVCS by shifting charging demand to low-price periods. Therefore, the integrated optimization of the design of the EVCS, renewable energy and EV charging scheduling leads to a higher revenue and lower cost, indicating the superiority of the proposed method over others. As a key part of this study, we further analyze the operation of the EVCS and the corresponding load curves in Case 4, and compare them with those in Case 3. Figure 3 shows the total charging demand of the EVCS. In Case 3, the EV charging demand profile is almost consistent across seasons because FCFS is a fixed charging method. However, in Case 4, the use of the proposed charging scheduling strategy leads to differences in EV charging demand curves across seasons, because it can be affected by tariffs, the number of charging tasks, and renewable energy output. Daily EV charging demand of the EVCS. As can be seen from Fig. 3, one hundred uncontrollable EVs restricted by the FCFS charging mode are mainly charged from 12:00 to 20:00. The peak value of EV charging demand appears during 15:30–15:45 and exceeds 300 kW. However, the EV charging demand of Case 4 is relatively gentle during these periods. The peak load of the EV charging demand under Case 4 is approximately 30% less than that of Case 3. Therefore, deferring some of the EV charging load by implementing the proposed charging scheduling strategy can smooth the load curve and alleviate congestion to some extent. It should be noted that the total area under each curve, representing the charging demand, is the same for both cases shown in the figure. In other words, the application of different charging modes does not change the total charging demand for EVs, but only adjusts EV charging time and power. Figures 4 and 5 display the load curves of various types corresponding to the operation of the EVCS for Case 3 and Case 4, respectively, for which the electric load consists of EV charging demand and conventional load. Daily operation of the EVCS in Case 3. (a) Spring; (b) Summer; (c) Autumn; (d) Winter. Daily operation of the EVCS in Case 4. (a) Spring; (b) Summer; (c) Autumn; (d) Winter. As can be seen from Figs. 4 and 5, Cases 3 and 4 have the same PV and WT load curves of the four seasons because the number of PV panels and WTs is equal for the two cases. Meanwhile, the energy used in both cases is supplied from renewable sources. Installed WT and PV generation not only supplies the electric load demand, but also generates profit by selling surplus power to the power grid. Compared with the output curve of renewable sources, the fluctuations of ESS output in both cases are lower during the whole operation due to the limitations of its state of charge (SOC). As a result, the EVCS prefers to sell surplus power directly to the grid rather than store it in relatively expensive ESS. Compared with the electric load curves in Fig. 4, the proposed strategy results in smaller fluctuations in the optimized electric load curves shown in Fig. 5. It can also be seen that the application of the proposed charging scheduling strategy leads to an increase in the amount of electricity traded with the grid during the time periods of higher electricity prices. Therefore, the proposed charging scheduling strategy increases the benefits of selling energy to the grid by regulating the EVCS’s electric load curve. Influence of the number of EVs With the increasing adoption of EVs, EVCSs will face a growing number of charging tasks. Next, we analyze the effect that the different numbers of EVs have on the profitability of the EVCS. As shown in Fig. 6, the annual total profit in Cases 3 and 4 increases with the number of EVs, which may appear counterintuitive because the EVCS needs to install more chargers to satisfy higher charging demand. Nevertheless, the additional charging revenue obtained from EVs exceeds the additional investment cost of chargers, so this result is reasonable. Moreover, when the number of EVs reaches 500, the proposed charging scheduling strategy increases the annual total profit of the EVCS by more than 15%. In general, the difference increases with the growth of the number of EVs, indicating that it is indeed important to adopt the proposed charging scheduling strategy in the optimal sizing problem of EVCSs and the advantages of the proposed strategy will be more evident in the case of high penetration of EVs. Annual total profit for different numbers of EVs. Sensitivity analysis In the above case studies, the ESS is mainly used to maintain the balance between power supply and demand, while its full potential has not yet been exploited. With the development of ESS technology, the capacity price of ESS is expected to be reduced, which may have an impact on the capacities of ESS installation and the operation of ESS in the dispatching cycle. Accordingly, we analyze the impact of ESS capacity price on the results in Case 4. Figure 7 depicts the installed ESS capacity and the changes in EVCS profit under different ESS capacity prices. Impact of capacity price of ESS on profits. As can be seen from Fig. 7, the total profit of the EVCS decreases as the ESS capacity price increases. Note that when the capacity price of ESS is below 2000 yuan, the capacity of the installed ESS reaches 857 kWh. In this case, although the installment cost of ESS rises, more revenue can be obtained through electricity price arbitrage. However, due to the limitations of the maximum power of renewable energy and the power of ESS, the decrease in the capacity price will not lead to an increase in the installed capacity of ESS all the time. When the ESS capacity price exceeds 2700 yuan, the installed ESS capacity decreases. This result is quite reasonable. Discussion The four cases designed in this study exhibit significant differences in terms of charging modes, power supply methods, and economic performance, which can be analyzed in detail as follows: (1) Comparison of Equipment Configuration. Chargers: Cases 1–3 adopt the FCFS charging mode, so the number of LPCs (1), NPCs (2), and HPCs (15) is the same. In Case 4, with the proposed charging scheduling strategy, the number of HPCs is reduced to 10, a decrease of 33.3% compared to the other three cases. This indicates that the scheduling strategy can effectively improve the utilization efficiency of high-power chargers, reducing the demand for equipment investment. Renewable-energy equipment: Case 1 does not include WT or PV. Case 2 is equipped with 40 WTs and 15 PVs, whereas Cases 3 and 4 are equipped with 36 WTs and 39 PVs. The reduction in the number of WTs and the increase in the number of PVs in Cases 3 and 4 compared to Case 2 may be related to the local renewable energy resource endowment and the ability to trade with the power grid, making the configuration more economical. ESS capacity: Case 2 has an ESS capacity of 326.59 kWh, Case 3 is 331.34 kWh, and Case 4 is reduced to 152.39 kWh, a decrease of 54.3% compared to Case 3. This is because the scheduling strategy optimizes the charging time distribution, reducing the demand for energy storage to balance the load. (2) Economic Performance Comparison. Annual Total Cost: Case 1 has the highest annual total cost, at 3,393,981 yuan, mainly because it relies entirely on electricity purchased from the grid, resulting in high energy-exchange costs with the grid (4144308 yuan). Case 2, which relies solely on renewable energy and ESS, has an annual total cost of 1,126,541 yuan, 67% lower than Case 1, demonstrating the economic advantages of renewable energy. Case 3, which can trade with the grid, achieves a negative annual total cost of − 984,817 yuan, turning losses into profits compared with Case 2, because it can sell excess renewable energy to the grid and obtain revenue. The negative energy-exchange cost (− 2455674 yuan) indicates net revenue from electricity trading. With the scheduling strategy, Case 4 further reduces the annual total cost to − 1,075,751 yuan, corresponding to an increase of nearly 9% in profit compared with Case 3. Component Costs: The charger cost of Case 4 is 74,727 yuan, which is 31.3% lower than the 108,694 yuan in Cases 1–3. The ESS cost of Case 4 is 26,572 yuan, which is 54% lower than the 57,775 yuan in Case 3. These cost reductions are directly attributed to the scheduling strategy improving equipment utilization. Operation and Maintenance Costs: Case 1 has the lowest operation and maintenance cost of 80,000 yuan, while Case 3 has the highest at 355,985 yuan, which is related to the number of renewable energy equipment and the frequency of grid transactions. Case 4’s operation and maintenance cost of 323,645 yuan is lower than that of Case 3, indicating that the scheduling strategy also improves maintenance efficiency to a certain extent. Furthermore, the investment recovery performance of the EVCS is analyzed by calculating the simple payback period, which is defined as the ratio of the total initial investment to the annual net profit. According to the optimal sizing results in Table 2 and the unit investment costs listed in Supplementary Tables S1 and S2, the total initial investment of Case 3 and Case 4 is about 25.61 million yuan and 25.00 million yuan, respectively. Since the annual total costs of Case 3 and Case 4 are − 984,817 yuan and − 1,075,751 yuan, the corresponding annual net profits are 984,817 yuan and 1,075,751 yuan, respectively. Therefore, the simple payback periods of Case 3 and Case 4 are about 26.01 years and 23.24 years, respectively. This result further demonstrates that the proposed charging scheduling strategy can not only improve the annual profitability of the EVCS, but also shorten the investment recovery period. Conclusion Reasonable planning of EVCSs is beneficial and important for supporting the high penetration of EVs. In this work, PV, WT, ESS, and multiple types of chargers are considered in the EVCS sizing problem, and a system framework is proposed to incorporate these components into the optimal sizing of EVCSs. Meanwhile, a charging scheduling strategy is proposed and embedded into the optimization model to reduce unnecessary expenses by adjusting the charging time of EVs. The numerical results show that installing renewable generators and ESS, rather than simply purchasing electricity from the power grid, can significantly reduce the annual total cost of the EVCS. Cheap energy from renewable sources not only meets the electricity demand of the EVCS, but also allows excess energy to be sold to the power grid for profit. Case 3, which can trade with the grid, achieves a negative annual total cost of − 984,817 yuan, with revenue from selling energy to the grid reaching 2,455,674 yuan. After incorporating the proposed charging scheduling strategy, Case 4 further improves the economic performance, with an annual total cost of − 1,075,751 yuan, representing an improvement of nearly 9% compared with Case 3. In addition, the number of HPCs in Case 4 is reduced from 15 to 10, a decrease of 33.3%, and the ESS capacity is reduced from 331.34 kWh to 152.39 kWh, a decrease of 54.3%, significantly reducing equipment investment costs. Moreover, the advantage of the proposed strategy becomes more prominent as the number of EVs increases. When the number of EVs reaches 500, the annual total profit of Case 4 is more than 15% higher than that of Case 3. Finally, the sensitivity analysis shows that when the ESS capacity price is below 2000 yuan/kWh, the installed ESS capacity reaches 857 kWh, whereas the installed capacity decreases when the ESS capacity price exceeds 2700 yuan/kWh. It should be noted that the output of PV and WT is calculated using the collected weather data of a given day at a given location, and the uncertainty of renewable energy and EV charging behavior is not considered in this study. Consequently, a key direction for future research will be to develop a robust optimization method that can effectively capture these uncertainties in order to enhance the resilience and practical applicability of the planning solutions under real-world variability. Additionally, the planning problem for EVCSs needs to be further extended, which requires both determining geographic locations in the transportation network and optimizing the number of devices in the EVCS. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements The authors would like to thank the anonymous referees for their valuable comments and suggestions. Author contributions Wenbin Shao: Conceptualization, Methodology, Resources, Visualization, Writing-Original draft preparation. Dingding Hu: Conceptualization, Methodology, Data curation, Software, Visualization, Formal analysis, Writing-Original draft preparation, Writing-Reviewing and Editing, Project administration, Supervision. Funding This research was supported by the Key Project of Scientific Research Plan of Colleges and Universities in Anhui Province (No. 2025AHGXZK31132). Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Code availability The custom MATLAB code used to generate the results reported in this study is publicly available in the OSF repository at https://osf.io/u2v9d/overview. The repository contains the core scripts required to reproduce the main results reported in the manuscript. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. The custom MATLAB code used to generate the results reported in this study is publicly available in the OSF repository at https://osf.io/u2v9d/overview. The repository contains the core scripts required to reproduce the main results reported in the manuscript.