EV Aggregator Strategy Balances Grid Needs and Driver Preferences
As electric vehicles (EVs) continue to gain traction across global markets, their role is evolving beyond personal transportation into dynamic participants within the energy ecosystem. A recent study published in the Transactions of China Electrotechnical Society presents a novel optimization strategy that enables electric vehicle aggregators (EVAs) to more effectively harness the collective power of EV fleets for grid support—while respecting individual user preferences.
The research, led by Fang Yuxuan, Hu Junjie, and Ma Wenshuai from the State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources at North China Electric Power University, introduces a Stackelberg game-theoretic model that integrates user behavioral psychology into the decision-making framework of EV aggregation. This approach marks a significant advancement in how energy systems can balance technical efficiency with human factors in the era of smart grids and vehicle-to-grid (V2G) integration.
The core challenge addressed in the paper lies in the inherent uncertainty of EV owner participation. While EVs represent a vast, distributed energy resource capable of providing critical grid services such as frequency regulation and peak shaving, their availability depends on human decisions. Unlike traditional power plants, EVs are privately owned assets whose primary function is mobility. Owners may be willing to contribute to grid stability, but only if it does not compromise their driving needs or damage their vehicle’s battery. This creates a complex negotiation space between the aggregator, which seeks to maximize revenue from energy markets, and the individual user, who prioritizes convenience, cost, and vehicle longevity.
Previous models often treated user behavior as a statistical given or assumed perfect rationality, leading to overly optimistic projections of available capacity. The new model, however, takes a more nuanced view by quantifying two key sources of user anxiety: battery degradation and loss of charging flexibility. These are not merely technical concerns but psychological barriers that influence willingness to participate in demand response programs.
Battery degradation anxiety stems from the fact that each charge and discharge cycle contributes to wear on the lithium-ion cells. Even under optimal conditions, repeated cycling reduces the battery’s capacity over time. For EV owners, this translates into a tangible concern about long-term vehicle value and performance. The researchers incorporate this by modeling the “extra” charging and discharging that occurs when an EV participates in grid services, distinguishing it from normal charging behavior. They further introduce a psychological amplification factor—what they call an “exclusion coefficient”—to reflect the fact that users often perceive battery wear as more damaging than it may be in reality. This captures the emotional weight of potential battery damage, which can deter participation even when financial incentives are attractive.
The second dimension, flexibility anxiety, relates to the disruption of the user’s charging routine. Most EV owners expect to plug in when they arrive home or at work and unplug when they are ready to leave. Participating in grid services may require them to delay charging, discharge during certain periods, or remain connected longer than desired. The model quantifies this inconvenience using a function that increases with the duration the vehicle is connected to the grid beyond its minimum charging needs. Crucially, the researchers recognize that users vary widely in their tolerance for such disruptions. Some may be indifferent to staying plugged in overnight, while others need the flexibility to leave at a moment’s notice. To reflect this diversity, the model includes decision factors that modulate the sensitivity of the anxiety function, allowing for a spectrum of user types—from highly flexible to highly inflexible.
These two anxiety components are combined into a composite cost metric, which is then weighed against the financial incentive offered by the aggregator. The result is a personalized assessment of whether a given user is likely to accept a particular grid service request. But the model goes a step further by acknowledging that human decisions are not perfectly rational or deterministic. Even among users with similar anxiety levels and incentives, some will opt in while others will opt out due to unpredictable personal factors. To account for this, the researchers employ a probabilistic framework based on binomial distribution, which introduces a degree of randomness into the decision process. This prevents the model from becoming too rigid and allows it to reflect the real-world variability of human behavior.
With this user-level model in place, the researchers construct a two-level game between the aggregator and the EV fleet. In game theory terms, this is a Stackelberg game, where one player—the leader—makes the first move, and the other players—the followers—respond optimally to that move. In this case, the EVA acts as the leader by setting the price it will pay for upward and downward reserve capacity (the ability to either reduce charging or inject power back into the grid). The EV users, acting as followers, then decide whether to participate based on the price and their individual anxiety costs.
The EVA’s goal is to maximize its own profit, which comes from selling reserve services to the grid at a higher price than it pays to the users. However, if it offers too little, few users will participate, and the available capacity will be insufficient. If it offers too much, participation will be high, but the profit margin will shrink. The optimal price lies somewhere in between, and finding it requires understanding how user participation responds to price changes—a response that is shaped by the anxiety model.
The solution is not found through a single calculation but through an iterative process. The EVA starts with an initial price offer. The user model then simulates how many EVs would accept that offer, based on their individual anxiety levels and the probabilistic decision rule. This gives the EVA an estimate of the available reserve capacity. The EVA then adjusts its price to improve its profit, and the process repeats. Over multiple iterations, the system converges to a stable equilibrium where neither the EVA nor the users have an incentive to change their behavior. This equilibrium represents the optimal balance between profitability and user satisfaction.
One of the key innovations of the study is the method used to handle the complexity introduced by the probabilistic user decisions. Because the model includes discrete choices (participate or not) and continuous variables (prices, power levels), it could be computationally intractable. The researchers overcome this by transforming the probabilistic elements into deterministic equivalents at the fleet level. Instead of simulating every individual decision, they calculate the expected participation rate for each user type based on the distribution of anxiety costs. This allows the optimization to proceed efficiently while still capturing the aggregate effects of individual variability.
To validate the model, the researchers conducted a series of simulations using a fleet of 150 EVs, divided into two typical charging scenarios: residential charging (overnight) and workplace charging (during the day). They compared the performance of their game-theoretic approach against two benchmark strategies: a simple ordered charging method, where EVs charge as early as possible, and a multi-objective particle swarm optimization (MOPSO) algorithm, which seeks to balance multiple goals without explicit game dynamics.
The results were compelling. The game-theoretic model achieved a higher level of user participation—measured as a “response ratio” of 32%—compared to the MOPSO approach. More importantly, it delivered superior economic outcomes for both the aggregator and the users. The EV users saw an increase in their expected reserve revenue, from 421 yuan under MOPSO to 454 yuan under the game model, while their total costs (including charging and battery wear) decreased. The EVA, in turn, increased its reserve market revenue from 1,049 yuan to 1,240 yuan, demonstrating that a more user-centric approach can also be more profitable.
Beyond economics, the model also delivered significant grid benefits. By shifting charging activity away from peak hours and enabling discharging during periods of high demand, the game-theoretic strategy reduced the daily load peak-to-valley difference by a notable margin. This “peak shaving and valley filling” effect helps to stabilize the grid, reduce the need for expensive peaking power plants, and integrate more renewable energy. In a world where solar and wind generation are intermittent, the ability of EVs to store excess daytime solar power and release it in the evening is becoming increasingly valuable.
The implications of this research extend beyond the technical details of optimization algorithms. It represents a shift in how we think about demand-side resources. Instead of treating users as passive nodes in a network, the model treats them as active agents with their own goals and constraints. This human-centered approach is essential for the long-term success of V2G programs. If users feel that their needs are being ignored or that they are bearing an unfair share of the costs, they will opt out, undermining the entire system. By explicitly modeling user anxiety and incorporating it into the decision process, the researchers have created a framework that is not only more accurate but also more equitable.
The study also highlights the importance of scale. The researchers found that the stability of the user response ratio improves as the size of the EV fleet increases. With small fleets, random variations in individual decisions can lead to large swings in aggregate behavior. But with larger fleets—on the order of 100 vehicles or more—the law of large numbers smooths out these fluctuations, making the overall response more predictable. This suggests that EV aggregation is most effective when done at a community or city-wide level, rather than through isolated, small-scale programs.
Another important insight is the role of feedback. The iterative nature of the game allows the EVA to continuously learn and adapt to user behavior. This is in contrast to static pricing schemes, which are set in advance and cannot respond to changing conditions. In a dynamic energy market, where prices and grid needs can shift rapidly, the ability to adjust incentives in real time is a major advantage. The model’s feedback loop creates a responsive, adaptive system that can maintain balance even as external conditions change.
The researchers also emphasize the importance of transparency and trust. For users to participate in such programs, they need to understand how decisions are made and how they will be compensated. The anxiety model, by making the trade-offs explicit, provides a basis for clear communication. Users can see that their concerns about battery wear and charging flexibility are being taken into account, which can increase their willingness to engage.
Looking ahead, the framework could be extended in several directions. One possibility is to incorporate real-time data from user feedback, such as actual driving patterns and charging behavior, to refine the anxiety parameters. Another is to include other types of distributed energy resources, such as home batteries or smart thermostats, into the same game-theoretic framework. This would enable a more holistic approach to demand-side management, where multiple resources are coordinated to achieve common goals.
The model could also be adapted to different market structures. While the current study focuses on reserve markets, the same principles could apply to other services, such as voltage regulation or congestion management. In regions with high penetration of renewable energy, EVs could play a crucial role in balancing supply and demand on a minute-by-minute basis. The ability to model user behavior accurately will be key to unlocking this potential.
In conclusion, the research by Fang, Hu, and Ma offers a sophisticated yet practical solution to one of the most pressing challenges in the energy transition: how to integrate millions of privately owned EVs into a reliable, efficient, and equitable power system. By combining game theory, behavioral modeling, and optimization, they have created a framework that respects user autonomy while maximizing collective benefits. As the world moves toward a future of electrified transportation and decentralized energy, such approaches will be essential for building a grid that is not only smart but also human-centered.
Fang Yuxuan, Hu Junjie, Ma Wenshuai, State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University. Transactions of China Electrotechnical Society. DOI: 10.19595/j.cnki.1000-6753.tces.230923