Rapid growth in electric vehicle adoption is placing charging infrastructure at the center of energy, transportation, and urban planning decisions. The scale of the electric vehicle market has moved quickly from an early niche to a mass-market system. According to public security data cited in a comprehensive technical review, electric vehicle ownership in China rose from 2.11 million in 2018 to 18.13 million by June 2024. Policy guidance has also set a clear direction: by the end of the 14th Five-Year Plan period, the charging infrastructure system is expected to be moderately advanced, balanced in layout, intelligent, and efficient, with the capacity to meet the charging needs of more than 20 million electric vehicles.
That expansion makes the electric vehicle charging problem far broader than the installation of charging piles. It requires forecasting when and where electric vehicle charging demand will appear, planning electric vehicle charging facilities so that they serve users without wasting resources, and scheduling electric vehicle charging loads so that grid stability, operator economics, and user convenience can be balanced. A review of key technologies in electric vehicle charging facility planning and operation links these three stages and shows that they are not isolated technical tasks. Electric vehicle demand forecasting provides the decision basis for planning; planning determines the physical envelope within which operation can occur; and scheduling determines whether the planned electric vehicle charging system performs well in daily operation.
The integrated view is important because electric vehicle charging demand is shaped by vehicle ownership, travel behavior, traffic conditions, distribution network conditions, charging prices, and user psychology. A forecast that ignores spatial travel patterns may mislead electric vehicle charging station siting. A planning model that ignores grid constraints may create new electric vehicle charging load pockets that exceed local distribution capacity. A scheduling strategy that ignores user willingness may produce theoretical flexibility that cannot be used in practice. The review therefore treats electric vehicle charging as a chain of forecasting, planning, and scheduling, with each stage feeding the next.
1. The Three Pillars of Electric Vehicle Charging Analysis
The technical literature on electric vehicle charging can be organized into three major pillars. The first is electric vehicle charging demand forecasting, which estimates future electric vehicle charging load from electric vehicle population, travel characteristics, traffic conditions, and distribution network operation. The second is electric vehicle charging facility planning, including site selection and capacity planning for electric vehicle charging stations and charging piles. The third is electric vehicle charging scheduling, which controls charging time and charging power to improve outcomes for electric vehicle users, charging facility operators, and power grid companies.
These pillars interact. Electric vehicle charging demand forecasting supports both long-term electric vehicle charging infrastructure planning and short-term electric vehicle charging scheduling. Electric vehicle charging facility planning determines the number, location, and capacity of electric vehicle chargers, which in turn affects waiting times, travel distances, and the flexibility available for electric vehicle charging scheduling. Electric vehicle charging scheduling can reduce peak load, absorb renewable generation, provide frequency and voltage support, and lower electric vehicle charging costs, but its effectiveness depends on the physical charging infrastructure and the accuracy of electric vehicle demand forecasts.
| Stage | Core question | Main outputs | Decision link |
|---|---|---|---|
| Electric vehicle charging demand forecasting | When and where will electric vehicle charging demand occur? | Short-term electric vehicle load, medium- and long-term electric vehicle demand, spatial electric vehicle charging demand | Provides inputs for electric vehicle charging facility planning and electric vehicle charging scheduling |
| Electric vehicle charging facility planning | Where should electric vehicle charging facilities be built and how large should they be? | Electric vehicle charging station sites, charger numbers, charging power capacity | Sets the physical capacity for electric vehicle charging operation |
| Electric vehicle charging scheduling | How should electric vehicle charging be timed and controlled? | Charging start times, charging power profiles, V2G dispatch, price signals | Determines daily performance of electric vehicle charging systems |
2. Forecasting Electric Vehicle Charging Demand
Electric vehicle charging demand forecasting is the foundation for electric vehicle charging facility planning and electric vehicle charging scheduling. It can be divided into time-distribution forecasting and spatial-distribution forecasting. Time-distribution forecasting estimates electric vehicle charging load over time, while spatial-distribution forecasting estimates where electric vehicle charging demand will be concentrated. The two perspectives serve different decisions: time-distribution forecasts support electric vehicle charging scheduling and medium- and long-term electric vehicle charging infrastructure planning, while spatial-distribution forecasts support electric vehicle charging station siting, traffic coordination, and distribution network planning.
2.1 Short-Term Electric Vehicle Charging Demand Forecasting
Short-term electric vehicle charging demand forecasting usually targets the next few hours to one week and is often carried out for a residential community, a charging station, or another electric vehicle charging demand point. Short-term electric vehicle charging demand is affected by user charging habits, travel behavior, weather, holidays, and other random factors, which are difficult to model directly. Data-driven methods are therefore common.
Regression analysis establishes a relationship between electric vehicle charging demand and influencing factors. It is simple and fast, and it can use historical data to estimate parameters. One reviewed approach used multiple linear regression with temperature, humidity, holidays, electric vehicle charging demand at the same time on the previous day, and electric vehicle charging demand at the same time in the previous week. However, regression equations are often selected by human judgment, and prediction accuracy can be limited in practical electric vehicle charging applications.
Artificial intelligence methods have been widely applied to short-term electric vehicle charging demand forecasting. Support vector machines map original data into a high-dimensional space and perform linear regression there. They can handle nonlinear relationships and small data sets and can balance learning ability and model complexity. Neural networks can fit strongly nonlinear input-output relationships when large data sets are available. Back propagation neural networks are common but can suffer from overfitting and local optima. One reviewed approach used particle swarm optimization and Bayesian regularization to improve a back propagation neural network for electric vehicle charging demand forecasting. Because electric vehicle charging demand has temporal dependence, long short-term memory networks and gated recurrent units have been applied to capture long-sequence patterns. Attention mechanisms have also been introduced into long short-term memory and gated recurrent unit models to improve electric vehicle charging demand prediction.
Artificial intelligence methods generally achieve higher accuracy than traditional regression methods, but they depend on large data sets. At the current stage, public electric vehicle charging demand data sets are limited, and researchers often build their own data sets for specific problems. The quality of those data sets affects the reliability of electric vehicle charging demand forecasts.
| Method family | Representative methods | Advantages for electric vehicle charging demand | Limitations for electric vehicle charging demand | Typical use |
|---|---|---|---|---|
| Regression analysis | Multiple linear regression | Simple principle, fast prediction, easy parameter estimation | Model form depends on human selection; limited accuracy in complex electric vehicle charging scenarios | Short-term electric vehicle charging demand at a charging point |
| Support vector machines | Support vector regression | Handles nonlinear relationships and small samples; good generalization | Performance depends on kernel and parameter selection | Electric vehicle charging demand with limited data |
| Neural networks | Back propagation, particle swarm optimization, Bayesian regularization | Strong fitting ability for large data sets | Overfitting and local optima risks; data demand is high | Short-term electric vehicle charging load forecasting |
| Temporal neural networks | Long short-term memory, gated recurrent unit | Captures time dependence in electric vehicle charging demand | Requires careful training and sufficient sequence data | Electric vehicle charging demand with strong temporal patterns |
| Attention-enhanced models | Attention with long short-term memory or gated recurrent unit | Improves representation of important time steps in electric vehicle charging data | Model complexity increases | Short-term electric vehicle charging demand with complex patterns |
2.2 Medium- and Long-Term Electric Vehicle Charging Demand Forecasting
Medium- and long-term electric vehicle charging demand forecasting targets months to years and usually treats a city or region as a whole. Because regional electric vehicle charging demand is closely related to electric vehicle population, many studies forecast electric vehicle ownership and then derive electric vehicle charging demand. Data-based methods use historical electric vehicle ownership data and analyze the relationship between electric vehicle numbers and time. Time-series methods have been used to forecast motor vehicle ownership and electric vehicle penetration, and gray models combined with support vector machines have been used to forecast electric vehicle numbers. However, the electric vehicle industry has a short history, so data-based methods often have few sample points, and the reliability of medium- and long-term electric vehicle forecasts cannot be fully guaranteed.
The Bass model describes the market diffusion of a new product. Electric vehicle adoption in the early market stage can be represented by the Bass model, which divides product adoption into innovation and imitation effects. Innovation refers to external influence such as advertising, while imitation refers to influence from existing adopters. The model uses three parameters: maximum market potential, innovation coefficient, and imitation coefficient. One reviewed study used the Bass model to forecast electric vehicle numbers and then simulated single electric vehicle charging demand to obtain medium- and long-term electric vehicle charging demand. Another study used a modified Bass model that considered price changes and allowed the innovation and imitation coefficients to be adjusted dynamically according to electric vehicle development and consumer satisfaction.
Medium- and long-term electric vehicle charging demand forecasting still faces long-term uncertainty from policy, electric vehicle technology, and supporting infrastructure. Forecasts therefore have limitations, and decision-makers should explicitly consider the imprecision of medium- and long-term electric vehicle charging demand results when using them for electric vehicle charging infrastructure planning.
2.3 Spatial Distribution Forecasting for Electric Vehicle Charging Demand
Spatial distribution forecasting identifies electric vehicle charging demand across locations and is important for traffic and distribution network operation. It usually simulates electric vehicle travel and uses Monte Carlo methods to obtain electric vehicle charging demand at different points. Two main modeling approaches are used: origin-destination matrix methods and trip-chain methods.
An origin-destination matrix describes the transfer relationship between origins and destinations for a group of electric vehicles at a given time. For a traffic network with n nodes, the origin-destination matrix is an n by n matrix. The element in row i and column j represents electric vehicle travel demand from node i to node j. Origin-destination matrices can be obtained by survey statistics or by road flow inversion. Survey methods require large amounts of travel data and platform support and are costly. Traffic-flow-based methods use road section traffic volume surveys and invert the origin-destination matrix. The inversion result is not unique and depends on the model. Common inversion models include methods without congestion effects, methods based on user equilibrium, and methods based on stochastic user equilibrium. The TransCAD platform integrates multiple inversion algorithms and can be used to calculate origin-destination matrices.
After origin-destination matrices for each period are obtained, electric vehicle travel demand can be simulated. The simulation process includes extracting an electric vehicle travel destination according to the origin-destination matrix, selecting a travel path according to traffic conditions and weather, judging whether electric vehicle charging is needed based on state of charge and travel distance, and advancing to the next simulation time until the simulation termination time is reached. Path selection can use shortest path, user equilibrium, or stochastic path equilibrium methods. Traffic conditions can be modeled as static or dynamic; dynamic modeling can reflect how electric vehicle travel affects traffic and can describe the relationship between electric vehicle user travel and traffic conditions more precisely.
Trip-chain methods describe the continuous travel behavior of electric vehicles across different areas in time order. They are generally used for electric vehicles with fixed travel patterns, such as private electric vehicles and electric buses. Trip-chain analysis usually divides trip purposes into home, work, and other affairs. Other affairs include shopping, dining, social activities, and leisure, corresponding to residential areas, work areas, and commercial areas. Four typical trip-chain structures are often considered, and the probability of each structure is obtained from statistical data or prior assumptions. Trip-chain forecasting requires distributions for the first departure time of the day, the travel distance of each trip, and the dwell time at the end of each trip. These distributions can be fitted using data such as the National Household Travel Survey in the United States. Common distributions include normal, Gaussian mixture, lognormal, and Weibull distributions. Using trip-chain probabilities and feature distributions, electric vehicle charging demand spatial distribution can be obtained through simulation.
| Approach | Suitable electric vehicle travel group | Input data | Traffic representation | Spatial output |
|---|---|---|---|---|
| Origin-destination matrix method | Electric vehicle fleets with relatively random travel demand, such as electric taxis | Origin-destination matrices by period, road flow data, path choice rules | Can represent dynamic traffic flow and path choice | Electric vehicle charging demand at individual traffic nodes |
| Trip-chain method | Private electric vehicles and electric buses with fixed or repetitive travel patterns | Trip-chain probabilities, departure time distributions, trip distance distributions, dwell time distributions | Traffic simulation is relatively coarse unless specific destinations are sampled | Electric vehicle charging demand in residential, work, and commercial areas |
3. Planning Electric Vehicle Charging Facilities
Electric vehicle charging facility planning must consider electric vehicle users, charging facility operators, the power grid, and the transportation network. The review organizes planning around objectives and constraints and then separates site selection from capacity planning. Electric vehicle charging facility planning is not only a location problem. It is a multi-actor problem in which electric vehicle charging station construction costs, electric vehicle user travel and waiting costs, and grid operation costs must be balanced.
3.1 Objectives and Constraints in Electric Vehicle Charging Facility Planning
The main planning objectives include electric vehicle charging facility construction and maintenance costs, electric vehicle user travel and waiting time costs, and power grid operation costs. Construction and maintenance costs depend on facility scale, charger type, charger power, and location. In general, larger electric vehicle charging facilities, higher-power chargers, and more central locations require greater investment. Electric vehicle user costs include travel to an electric vehicle charging facility and waiting when charging demand exceeds facility capacity. Grid operation costs arise because electric vehicle charging creates large electrical loads that can change load peak-valley characteristics and increase distribution network losses. Proper electric vehicle charging facility planning can reduce those costs.
Planning constraints include the charging capacity provided by electric vehicle charging facilities, user charging satisfaction, existing electric vehicle charging facilities, and power grid security. Charging capacity constraints ensure that the number of electric vehicle charging stations, the number of chargers in each station, and charger power can meet electric vehicle charging demand without excessive waste. User satisfaction constraints consider distance between electric vehicle charging stations and waiting time. Existing facility constraints require new electric vehicle charging facilities to maintain appropriate distances from existing facilities and to avoid resource waste. For multi-stage electric vehicle charging station planning, current-stage facilities must meet current electric vehicle charging demand and must not create severe barriers for later stages. Grid security constraints include voltage magnitude limits and line power limits because electric vehicle charging load can create significant voltage stress.
3.2 Site Selection Models for Electric Vehicle Charging Facilities
Electric vehicle charging facility site selection draws on classical facility location models. The main approaches are point-demand models and flow-demand models. Point-demand models include the p-median model, p-center model, set covering model, and maximum covering model. Flow-demand models focus on traffic flow along routes rather than demand at nodes.
The p-median model selects p electric vehicle charging facility sites to minimize the weighted distance between demand points and service facilities. It considers overall system cost and is intended to maximize social benefit. Many electric vehicle charging facility studies adapt the p-median model to include construction costs, user costs, traffic constraints, and distribution network constraints. One study minimized the total cost of electric vehicle charging facility operators and electric vehicle users. Another proposed a bi-level planning model that coordinated the transportation network and distribution network while considering both economics and security.
The p-center model minimizes the maximum distance from any demand point to its nearest electric vehicle charging facility. It is suitable when the worst-case electric vehicle travel distance must be limited, such as in emergency-oriented planning. However, electric vehicle charging planning usually emphasizes overall social benefit rather than only the worst case, so the p-center model has limited application. Because the number of facilities p is fixed in advance, the p-center model cannot guarantee that the worst-case electric vehicle charging travel distance meets a specific requirement. The set covering model addresses this by seeking the minimum number of electric vehicle charging facilities needed to satisfy all electric vehicle charging demand. One reviewed study used a set covering model to minimize electric vehicle charging station construction and operation costs while ensuring that the unreachable rate of electric vehicle charging demand met requirements, and it used a genetic algorithm to obtain charging station planning results for Shanghai. The maximum covering model considers a limited construction budget and seeks to satisfy as much electric vehicle charging demand as possible, directing limited resources to more electric vehicle users.
Flow-demand models study traffic flow on road sections rather than electric vehicle charging demand at traffic nodes. A travel route contains multiple road sections. When a relevant facility is built on one section, the traffic flow on that route is considered captured, meaning the electric vehicle charging demand of that flow is served. The flow-intercepting model maximizes captured flow subject to a fixed number of facilities. It assumes that one relevant facility on a route is enough to satisfy the route demand and does not consider electric vehicle driving range or facility capacity. Later studies proposed flow-refueling location models and capacity-constrained flow-refueling models for electric vehicle charging facility planning.
| Model | Core objective | Strength for electric vehicle charging planning | Limitation for electric vehicle charging planning | Typical electric vehicle context |
|---|---|---|---|---|
| p-median | Minimize weighted distance from electric vehicle demand points to facilities | Considers overall user travel cost and social benefit | May underemphasize distant electric vehicle demand points | Urban electric vehicle charging station siting |
| p-center | Minimize the maximum distance to the nearest facility | Limits worst-case electric vehicle travel distance | Fixed facility number; less focused on overall social benefit | Emergency or equity-oriented electric vehicle charging access |
| Set covering | Build the minimum number of facilities to cover all electric vehicle demand | Can enforce electric vehicle charging demand reachability | May assume unlimited facility construction in basic form | Electric vehicle charging station coverage planning |
| Maximum covering | Maximize covered electric vehicle demand under limited budget | Allocates limited resources to more electric vehicle users | Coverage definition affects results | Budget-constrained electric vehicle charging facility planning |
| Flow-intercepting and flow-refueling | Maximize captured electric vehicle traffic flow or refueled flow | Reflects route-based electric vehicle charging demand | Basic flow-intercepting model ignores driving range and capacity | Intercity highway electric vehicle charging facility planning |
Point-demand models are generally more suitable for urban electric vehicle charging facility siting because electric vehicles can choose a nearby electric vehicle charging station when state of charge falls below a certain level. Flow-demand models are more suitable for highway electric vehicle charging facility planning because facilities must be placed along routes to avoid electric vehicles running out of charge. Comparisons of the p-median, set covering, and maximum covering models show that the p-median model tends to place electric vehicle charging stations near communities with large electric vehicle charging demand, and planning results become more stable as the number of electric vehicle charging stations increases.
3.3 Capacity Planning for Electric Vehicle Charging Facilities
Electric vehicle charging facility capacity planning studies the number and power rating of chargers inside an electric vehicle charging station. Most capacity planning models use electric vehicle charging demand forecasts as input and apply queueing theory to estimate user waiting time. The optimization objective is often to minimize the sum of electric vehicle user waiting time cost and electric vehicle charging station construction cost. Some studies use the maximum electric vehicle charging demand over a period as the planned capacity, while others set capacity so that the queue length does not exceed a given value. With the growth of distributed renewable generation, some studies combine electric vehicle charging station capacity planning with renewable generation and optimize capacity configuration to support economical operation of the electric vehicle charging station.
4. Scheduling Electric Vehicle Charging
Uncoordinated electric vehicle charging can cause sharp fluctuations in power grid load, reduce power quality, shorten the life of electrical equipment, and increase grid losses. Electric vehicle charging load has transferable characteristics. A reasonable electric vehicle charging scheduling strategy can reduce the impact of electric vehicle charging on the power grid, lower electric vehicle charging costs for users and charging facility operators, and improve grid stability. With vehicle-to-grid technology, electric vehicles can not only draw power from chargers but also send power back to the grid, expanding the range of electric vehicle charging scheduling strategies.
4.1 Schedulable Potential of Electric Vehicle Charging
Electric vehicle charging scheduling requires forecasting the schedulable potential of electric vehicles, including charging and discharging time, power, and capacity. This forecasting must consider electric vehicle travel behavior, user willingness to participate, and electric vehicle state of charge. Because a single electric vehicle battery capacity is small and its charging and discharging power is relatively low, electric vehicle schedulable potential is usually forecast at the cluster level. Each electric vehicle cluster is controlled by an aggregator and participates in electric vehicle charging scheduling.
Electric vehicle cluster schedulable potential forecasting typically determines the minimum electricity demand for electric vehicle travel, then uses electric vehicle capacity constraints and charging and discharging power constraints to determine the charging and discharging power and capacity scheduling range of the electric vehicle cluster. To address uncertainty in electric vehicle travel behavior, one reviewed study used a bidirectional long short-term memory network to forecast single electric vehicle travel and applied Minkowski sum theory to obtain the schedulable potential boundary of the electric vehicle cluster. Another study considered state of charge and user price sensitivity and used a random forest method to forecast electric vehicle user willingness to participate in electric vehicle charging scheduling. For provincial and city-scale electric vehicle schedulable potential, one study proposed a multi-scenario schedulable potential assessment model based on clustering for peak shaving, frequency regulation, and voltage regulation, enabling electric vehicles to serve multiple ancillary services.
4.2 Peak Shaving and Valley Filling Through Electric Vehicle Charging Scheduling
Power grid load has peak and valley characteristics. Electric vehicle batteries can act as storage: electric vehicles can reduce charging during peak periods and increase charging during valley periods, reducing the peak-valley difference and lowering the cost of generation capacity needed for peak demand. Electric vehicle charging scheduling for peak shaving and valley filling is often studied for electric vehicle clusters in a residential community or electric vehicle charging station.
Under time-of-use pricing, electric vehicle users may concentrate charging during valley periods, which can create a new electric vehicle charging load peak. To address this, one reviewed study proposed a dual-sequence valley-period electric vehicle charging strategy to determine charging start times. The strategy allocates a proportion of electric vehicles so that their expected charging end time is the end of the valley price period, while the remaining electric vehicles start charging at the beginning of the valley price period. Monte Carlo simulation can determine the allocation proportion that minimizes the daily maximum power of the community load. Another study further considered the valley-period grid load acceptance capacity, calculated the probability distribution of charging start times for users with different charging durations, and selected charging start times according to that distribution. Compared with uncoordinated electric vehicle charging and dual-sequence valley-period charging, this approach reduced the peak-valley difference and variance of electric load. A further study divided an electric vehicle cluster into a controllable vehicle group and a conventional vehicle group according to battery state and charging time, and established an electric vehicle charging optimization model for the controllable group with the objective of minimizing overall grid load variance.
4.3 Frequency and Voltage Regulation Through Electric Vehicle Charging Scheduling
Compared with peak shaving and valley filling, frequency and voltage regulation rely on rapid changes in electric vehicle charging power. Because electric vehicles are connected to the grid through power electronic devices, charging power can change quickly. Electric vehicle charging can therefore respond faster than synchronous generators and can support frequency and voltage regulation. Frequency regulation requires electric vehicles to quickly change active charging power. One reviewed study used an electric vehicle cluster to provide grid frequency regulation and applied a sliding mode proportional-integral controller to maintain frequency stability while reducing electric vehicle cluster processing and system frequency regulation cost. Another study considered market mechanisms and used charging and discharging price incentives to encourage electric vehicles to participate in frequency regulation services.
Voltage regulation requires additional consideration of active and reactive power allocation in electric vehicle chargers. Reactive power can be adjusted to regulate voltage while meeting electric vehicle charging demand. One reviewed study proposed a voltage regulation method in which reactive power compensation devices and electric vehicle charging stations in a distribution network participate jointly. Another study proposed a day-ahead and real-time coordination model that considered reactive power compensation from electric vehicle charging stations, ensured that electric vehicles reached their target charging amount when leaving, and maintained system voltage stability.
4.4 Renewable Energy Accommodation Through Electric Vehicle Charging Scheduling
Renewable generation such as wind power and solar power fluctuates with natural conditions. As renewable generation increases, power system operation faces greater challenges. Electric vehicle charging load can respond quickly to renewable generation fluctuations, and electric vehicles can help absorb renewable energy and improve renewable utilization. One reviewed study considered electric vehicle accommodation of photovoltaic and wind generation, adjusted electric vehicle charging prices according to photovoltaic and wind output, and optimized power system load fluctuation and electric vehicle charging cost. Another used a bi-level optimization model for electric vehicle accommodation of wind power: the upper model minimized load fluctuation, and the lower model minimized actual electric vehicle charging cost and deviation between electric vehicle charging load and planned load from the perspectives of users and the grid.
These studies often assume that renewable generation can be predicted. In practice, renewable output uncertainty remains a challenge. One reviewed approach addressed photovoltaic and wind uncertainty by combining day-ahead forecasting with real-time dispatch and establishing an optimization model to minimize the cost of lost load caused by renewable generation fluctuations. Another study proposed a multi-time-scale scheduling model for a regional grid that considered electric vehicle flexibility and wind power accommodation.
4.5 Lowering Electric Vehicle Charging Costs
Electric vehicle charging scheduling aimed at reducing charging costs can use direct control or price-guided control. Direct control is organized from the perspective of an electric vehicle cluster or electric vehicle charging operator. Electric vehicle users transfer charging scheduling authority to a control center in exchange for a subsidy, and the control center formulates the electric vehicle cluster charging schedule. One reviewed study optimized electric vehicle charging cost and grid load fluctuation under time-of-use pricing while considering the interests of the electric vehicle control center and the grid. Because the control center needs an incentive scheme to obtain electric vehicle charging scheduling authority, another study used cooperative game theory to study benefit allocation from orderly electric vehicle charging between electric vehicle users and charging facility operators, seeking a win-win outcome.
Direct control treats the electric vehicle group as the control object and may not fully consider individual electric vehicle user willingness. Price-guided indirect control gives charging choice back to electric vehicle users, who decide whether to participate in orderly electric vehicle charging according to charging prices. One reviewed study modeled the psychological effect of electric vehicle users changing charging time in response to charging price differences, used Monte Carlo simulation and particle swarm optimization to solve for optimal electric vehicle charging prices, and showed that electric vehicle users, charging operators, and the grid can all increase their benefits. Price-guided control better respects electric vehicle user charging willingness, but the randomness of electric vehicle user psychology and charging behavior makes modeling difficult and requires further study.
| Scheduling objective | Main control lever | Electric vehicle role | Key benefit | Reviewed implementation focus |
|---|---|---|---|---|
| Peak shaving and valley filling | Charging start time and charging power | Shiftable electric vehicle charging load | Reduce peak-valley difference and load variance | Dual-sequence valley charging, probability-based start times, group optimization |
| Frequency regulation | Fast active power change | Fast-responding electric vehicle cluster | Support frequency stability and reduce regulation cost | Sliding mode proportional-integral control, market incentives |
| Voltage regulation | Active and reactive power coordination | Distributed electric vehicle charging station resource | Maintain voltage while meeting electric vehicle charging demand | Reactive power compensation coordination, day-ahead and real-time model |
| Renewable energy accommodation | Price signals and charging dispatch | Flexible electric vehicle charging load | Absorb wind and solar generation and reduce curtailment | Price adjustment, bi-level optimization, day-ahead and real-time dispatch |
| Charging cost reduction | Direct control or price guidance | Controllable electric vehicle cluster or price-responsive electric vehicle user | Lower electric vehicle charging cost and share benefits | Cooperative game, user psychology modeling, particle swarm optimization |
5. Demonstration Projects and Operational Evidence for Electric Vehicle Charging
Electric vehicle charging scheduling research has moved from simulation to demonstration. Pilot projects in China and Europe have tested peak shaving, valley filling, frequency regulation, renewable energy accommodation, and electric vehicle charging cost reduction. These demonstrations provide operational evidence on how electric vehicle users, charging facility operators, and grid companies respond to incentives.
5.1 Shanghai Electric Vehicle Demand Response Pilots in 2019
Shanghai carried out three electric vehicle demand response pilots in 2019. The pilots set compensation prices for peak shaving and valley filling to encourage electric vehicle operators to participate in demand response. Private electric vehicle charging piles had a valley-filling demand response participation rate of 5.3 percent, which was limited but still represented large valley-filling capability because the number of private electric vehicle charging piles is large. Dedicated office charging piles reached a 75 percent response rate in peak-shaving demand response, creating significant economic benefit for dedicated charging pile operators. Battery swap stations had more controllable battery scheduling time and reached an 81.2 percent response rate when notified 30 minutes in advance, although the total market size was relatively small. The results show that different electric vehicle charging facility types have different demand response characteristics.
| Electric vehicle charging facility type | Peak shaving or valley filling | Response rate | Resource potential | Daily peak shaving feasibility |
|---|---|---|---|---|
| Private electric vehicle charging pile | Valley filling | Low | High | Feasible |
| Dedicated electric vehicle charging pile | Peak shaving and valley filling | Relatively high | High | Feasible |
| Electric vehicle battery swap station | Peak shaving and valley filling | High | High | Feasible |
5.2 Baoding Great Wall Industrial Park Vehicle-Grid Interaction Demonstration
The 2021 Baoding Great Wall Industrial Park vehicle-grid interaction demonstration project was described as the largest park-level V2G demonstration in China. The park deployed 50 V2G charging and discharging piles. Owners of electric vehicles with V2G capability could participate in energy and power market transactions through a mobile application. To protect owner interests and reduce electric vehicle charging and discharging costs, battery warranties were provided for all electric vehicles participating in the project.
During operation, electric vehicle owners could receive rewards for discharging. The operator set high, medium, and low discharge subsidy tiers according to each electric vehicle owner’s monthly discharge amount. Under this reward mechanism, each owner could receive up to RMB 2,000 per month. Future work was expected to study how prices affect owner discharge willingness and to develop battery charging and discharging degradation models.
5.3 Jiangsu Large-Scale Vehicle-Grid Interaction Demonstration
On August 23, 2023, a large vehicle-grid interaction demonstration center was completed in Wuxi. The center covers more than 14,500 square meters. Through intelligent upgrading of electric vehicle chargers, it can support 50 electric vehicles discharging back to the grid at the same time, with reverse discharge power of nearly 2,000 kW for 30 minutes. Since operation began, the center has used incentives to guide users into bidirectional vehicle-grid interaction and has completed more than 300 discharge events, with discharged electricity exceeding 1,800 kWh. The electricity has been connected to a city-level virtual power plant platform and can provide 2.2×10^6 kWh of electricity support during annual grid peak periods.
A second phase is expected to add super charging and mobile charging and battery swapping. It is expected to meet the charging demand of 144 electric vehicles at one time, the discharging demand of 50 electric vehicles, and the battery swapping demand of 400 electric vehicles. The center is expected to combine photovoltaic generation, energy storage, electric vehicle charging, and electric vehicle discharging, supporting grid supply-demand balance, improving overall power system efficiency, and supporting reliable grid connection of distributed renewable energy.
5.4 Denmark Parker Vehicle-to-Grid Project
The Parker project in Denmark was a demonstration project for commercial development of V2G frequency regulation. Its main goal was to test the ability of electric vehicles to interact with the grid and to support commercial development. The project ran from 2016 to 2018 and included 50 10 kW DC chargers capable of bidirectional interaction between electric vehicles and the grid.
The project focused on the ability of electric vehicles to provide power to the grid and on related technology. The electric vehicles and DC V2G chargers used in the project were able to provide all frequency regulation services needed in Denmark. The project also identified the CHAdeMO standard as the only V2G-supported standard in its context. When participating in power market regulation, charging economics were affected by multiple factors, and stable revenue had not yet been achieved, so further business model exploration was needed.
5.5 Switzerland SunnYparc Project
The SunnYparc project in Switzerland began in 2022 and is planned for completion in 2025. It studies how different power market pricing can guide orderly electric vehicle charging and support electric vehicle accommodation of photovoltaic generation. The project plans to build photovoltaic generation with annual output of 1 GWh to supply local residents and 250 electric vehicle charging piles. Through 50 V2G-capable chargers, electric vehicle batteries can store surplus photovoltaic generation during the day and discharge back to the grid during peak demand periods, providing electricity for residents. Because Switzerland allows private microgrid operation, the process can be guided by power market pricing and support market-based electric vehicle participation in grid regulation.
| Demonstration | Location and period | Electric vehicle charging or V2G scale | Reported operational focus | Electric vehicle grid interaction result |
|---|---|---|---|---|
| Shanghai demand response pilots | Shanghai, 2019 | Private piles, dedicated office piles, and battery swap stations | Peak shaving and valley filling compensation | Private pile valley response 5.3 percent; dedicated office peak response 75 percent; battery swap response 81.2 percent with 30-minute notice |
| Baoding Great Wall Industrial Park | Baoding, 2021 | 50 V2G charging and discharging piles | Park-level V2G and market participation | Electric vehicle owners could receive up to RMB 2,000 per month under discharge rewards |
| Jiangsu vehicle-grid interaction center | Wuxi, completed August 23, 2023 | 50 electric vehicles discharging simultaneously; 30-minute reverse power near 2,000 kW | Large-scale bidirectional electric vehicle charging and virtual power plant connection | More than 300 discharge events and more than 1,800 kWh discharged; 2.2×10^6 kWh annual peak support |
| Parker project | Denmark, 2016-2018 | 50 10 kW DC chargers | V2G frequency regulation and commercial development | Electric vehicles and DC V2G chargers provided all frequency regulation services needed in Denmark; stable revenue remained a challenge |
| SunnYparc project | Switzerland, 2022-2025 | 1 GWh annual photovoltaic output, 250 charging piles, 50 V2G chargers | Electric vehicle charging guided by market pricing and photovoltaic accommodation | Electric vehicle batteries store surplus photovoltaic generation and discharge during peak demand |
6. Research Gaps and Future Directions for Electric Vehicle Charging
The review identifies several limitations that remain important for electric vehicle charging research and practice. These limitations concern electric vehicle charging behavior modeling, long-term electric vehicle demand evolution, coordinated planning and operation of electric vehicle charging facilities and the grid, pricing mechanisms for orderly electric vehicle charging, and the forms through which electric vehicles participate in grid ancillary services.
6.1 Electric Vehicle Charging Behavior Modeling
Electric vehicle charging behavior modeling is essential for electric vehicle charging demand forecasting. Electric vehicle charging behavior is affected by vehicle type, owner charging preference, and travel plans, and it has significant randomness. Current origin-destination matrix and trip-chain methods often use simplified user charging behavior in electric vehicle charging simulation and do not fully consider traffic, holidays, and user travel behavior. They may therefore struggle in complex electric vehicle charging scenarios.
More detailed electric vehicle charging behavior modeling requires electric vehicle travel data, but data collection raises privacy concerns. Future work can explore user data collection schemes that use distributed technology to protect privacy while enabling data sharing for electric vehicle charging analysis.
6.2 Long-Term Evolution of Electric Vehicle Charging Demand
Electric vehicle charging demand and electric vehicle charging infrastructure influence each other. Reasonable construction and operation of electric vehicle charging infrastructure can better satisfy electric vehicle charging demand, and higher electric vehicle charging satisfaction can encourage more car owners to choose electric vehicles. Most current studies focus on how electric vehicle charging facilities can meet electric vehicle charging demand, while fewer studies examine how electric vehicle charging facilities affect long-term electric vehicle population evolution.
To describe the interaction between electric vehicle charging facilities and electric vehicle numbers more accurately, future work can start from the Bass model and further explore how electric vehicle charging facility construction affects the imitation coefficient. Such work could guide long-term electric vehicle charging infrastructure planning.
6.3 Coordinated Planning and Operation of Electric Vehicle Charging Facilities and the Grid
As electric vehicle charging demand grows, electric vehicle load pressure on the grid also rises. Current electric vehicle charging demand is mainly supported by grid capacity margin. As electric vehicle numbers increase further, existing distribution networks may need upgrading. Future distribution network planning and operation should fully consider electric vehicle charging load and achieve coordinated planning and operation of the distribution network and electric vehicle charging facilities.
Coordinated planning and operation must consider short-term and medium- and long-term uncertainty in electric vehicle charging demand. Multi-stage planning methods can address coordinated planning, and stochastic optimization methods can address coordinated operation scheduling.
6.4 Pricing Mechanisms for Orderly Electric Vehicle Charging
To use electric vehicle battery storage for power system regulation, electric vehicle users must be guided to participate in vehicle-grid interaction. Current electric vehicle charging research and demonstrations often assume that electric vehicle users sign charging agreements and that electric vehicle charging and discharging can be fully controlled. However, related price mechanism designs are still lacking.
Pricing design must consider both the benefits of the dispatching party and electric vehicle users and the psychology of user participation. Cooperative game methods can be used to design benefit allocation among participating parties. Data from existing electric vehicle charging demonstrations can be used to build a price sensitivity model of user participation in grid interaction and to achieve win-win outcomes for the dispatching party and electric vehicle users.
6.5 Forms of Electric Vehicle Participation in Grid Ancillary Services
Electric vehicles are distributed mobile energy storage resources. Most studies of electric vehicle participation in grid ancillary services use electric vehicle clusters, but current research lacks a unified architecture for utilizing electric vehicle resources, and existing studies remain relatively independent. To fully exploit the potential of electric vehicles in grid ancillary services, it is necessary to explore participation policies and contract mechanisms and to adopt effective control methods for electric vehicle dispatch so that power system stability can be maintained when large numbers of electric vehicles participate in ancillary services.
Electric vehicle participation in grid ancillary services can be classified by charging scenario. For private electric vehicle charging piles, blockchain technology can support peer-to-peer transactions. For public electric vehicle charging stations, a bi-level game model can be used, with charging price games between aggregators and electric vehicle users supporting electric vehicle participation in charging scheduling. For dedicated electric vehicle charging stations, relatively fixed charging times and power demand allow the station to be treated as a virtual power plant participating in grid ancillary services.
7. Conclusion
The electric vehicle charging problem spans demand forecasting, charging facility planning, and charging operation scheduling. Electric vehicle charging demand forecasting provides the basis for decisions. Electric vehicle charging facility planning determines where electric vehicle chargers should be located and how much charging capacity should be built. Electric vehicle charging scheduling determines how electric vehicle charging load is controlled in daily operation to reduce grid impact, lower electric vehicle charging costs, and support renewable energy accommodation.
The review shows that electric vehicle charging is a multi-actor and multi-constraint problem. Electric vehicle users care about travel distance, waiting time, and charging cost. Electric vehicle charging facility operators care about construction and operation economics. Power grid companies care about voltage, line loading, losses, and stability. Transportation networks determine electric vehicle travel and charging accessibility. Effective electric vehicle charging infrastructure planning and operation must consider all these factors together.
Demonstration projects provide practical evidence that electric vehicle charging can be scheduled and that electric vehicle users and operators respond to incentives. Private electric vehicle charging piles, dedicated electric vehicle charging piles, battery swap stations, public electric vehicle charging stations, and V2G-enabled electric vehicle chargers have different response characteristics and different potential roles in peak shaving, valley filling, frequency regulation, voltage regulation, renewable energy accommodation, and electric vehicle charging cost reduction. Future progress will depend on better electric vehicle charging behavior data, improved long-term electric vehicle demand evolution models, coordinated grid and electric vehicle charging facility planning, practical pricing mechanisms, and unified architectures for electric vehicle participation in grid ancillary services.
As electric vehicle ownership continues to rise, the planning and scheduling of electric vehicle charging facilities will become increasingly important for energy system efficiency and transportation service quality. The integrated development of electric vehicle demand forecasting, electric vehicle charging facility planning, and electric vehicle charging scheduling offers a practical pathway toward a charging infrastructure system that is adequate, balanced, intelligent, and efficient for electric vehicle users and the power grid.