EV Clusters Pave the Way for Smarter Power-Transport Networks
The rapid rise of electric vehicles (EVs) is reshaping not just how we travel, but the very infrastructure that supports modern life. No longer seen merely as a replacement for internal combustion engines, EVs are emerging as dynamic, intelligent assets capable of bridging two critical systems: the transportation network and the power grid. This convergence, often termed the “power-transportation coupling network,” presents a unique opportunity to enhance the efficiency, sustainability, and resilience of both systems. However, harnessing this potential requires moving beyond simplistic models and embracing the complex, dynamic interplay between where EVs go and how they use electricity. A groundbreaking new study from Kunming University of Technology, published in the prestigious journal Automation of Electric Power Systems (AEPS), offers a sophisticated solution, proposing a novel strategy to unlock the full flexibility of EV clusters through dynamic, coordinated optimization.
The challenge is multifaceted. On one hand, the uncoordinated charging of a growing fleet of EVs can create significant stress on local power distribution networks, leading to voltage fluctuations, equipment overload, and increased operational costs for utilities. This is particularly acute during peak hours when electricity demand is already high. On the other hand, the decisions EV drivers make—where to go, which route to take, and which charging station to use—are deeply influenced by real-time traffic conditions, station availability, and pricing. This creates a feedback loop: traffic congestion can lead to queues at popular charging stations, which in turn discourages drivers from using them, potentially diverting traffic and charging demand to less optimal locations. This can result in an uneven distribution of load, where some parts of the grid are strained while others remain underutilized, and some roads are congested while others are idle. The key to breaking this cycle and achieving a harmonious, efficient system lies in treating the EV not as a passive consumer, but as an active, flexible participant in a larger, integrated network.
The research, led by Professor Liu Zhijian and his team, including graduate students Dai Jing and Yang Lingrui, directly addresses this challenge. Their work moves past the limitations of previous studies, which often examined the power and transportation systems in isolation or used static models that fail to capture the fluid, real-time nature of traffic and energy flows. For instance, earlier models might have determined an optimal charging price for a station based on a single snapshot of demand, ignoring how that price would influence driver behavior over the next hour as traffic patterns shifted. Other models attempted to guide drivers based solely on the shortest physical path, a strategy that is fundamentally selfish and leads to the “tragedy of the commons,” where everyone’s individual optimal choice results in a collectively suboptimal outcome—congestion and grid strain.
The core innovation of the Kunming team’s approach is its dynamic, two-layered optimization framework. This model doesn’t just look at the power grid or the road network; it creates a continuous dialogue between the two, using a powerful concept called the “instantaneous unit flow travel cost” as a universal signal. This cost is not a simple sum of fuel and tolls. Instead, it is a sophisticated, real-time metric that encapsulates the total economic burden of a journey at any given moment. It includes the direct cost of electricity for charging, the value of the driver’s time spent traveling (the “time cost”), and a crucial new element: a penalty cost for the delays caused by queuing at a charging station. By integrating these factors, the model creates a holistic view of the system’s health.
The process begins with a detailed simulation of the dynamic traffic network. The researchers use a method called Dynamic Network Loading (DNL) to model how traffic flows propagate through the city’s roads and intersections over time. A key feature of their model is the use of “virtual arcs” to represent both the starting points of journeys and the charging stations themselves. This elegant abstraction allows the complex process of queuing and charging at a station to be treated mathematically in the same way as driving down a road, with its own “travel time” that increases as more EVs join the queue. This travel time is calculated using an “arc impedance function,” a mathematical relationship that shows how the time to traverse a segment—whether a physical road or a virtual charging arc—increases non-linearly as the number of vehicles using it grows. This captures the reality that adding one more car to a lightly traveled road has little impact, but adding one more car to a heavily congested highway can dramatically slow down everyone.
With this dynamic traffic model in place, the focus shifts to the EVs themselves. The researchers recognize that not all EVs are the same. They come in different types—private cars, buses, ride-share vehicles—each with distinct driving patterns, battery sizes, and charging needs. More importantly, an individual EV’s flexibility is not static; it depends on its current state of charge, its destination, and how long it plans to stay at a charging station. The team develops a “flexible operation domain” for each EV, a mathematical representation of all the possible charging and discharging patterns it can follow while still meeting its driver’s needs. This domain defines the upper and lower limits of the power the EV can absorb from the grid or supply back to it at any given moment.
The true power of the model lies in its ability to aggregate thousands of individual EVs into a single, manageable “cluster.” Trying to manage each EV individually would be computationally impossible. To solve this, the researchers employ a sophisticated mathematical technique known as the Minkowski sum, combined with a linear approximation using a geometric shape called a “zonotope.” This allows them to efficiently combine the flexible operation domains of hundreds or thousands of EVs into a single, time-varying “flexible operation domain” for the entire cluster at a given charging station. This aggregated cluster can then be treated as a large, virtual battery, a “generalized energy storage” unit that the power grid can dispatch to help balance supply and demand.
This is where the two layers of the model come together in a powerful feedback loop. The top layer is the Transportation Network Optimization (TNO) model, which seeks to minimize the total travel cost for all EV users across the entire network. It uses the real-time “instantaneous unit flow travel cost” to guide drivers toward the most efficient routes and charging stations. This cost is derived from the current state of the network: high electricity prices at a station, heavy traffic on a route, or long queues at a charger will all increase the cost of using that path, discouraging drivers from choosing it.
The bottom layer is the Distribution Network Optimization (DNO) model, which focuses on the economic and stable operation of the power grid. Its goal is to minimize costs, which include the expense of running traditional power generators, the penalties for having to curtail (waste) renewable energy from wind or solar farms when there’s too much supply, and the penalties for having to shed (cut) power to non-EV customers during periods of high demand. The DNO model uses the aggregated flexible operation domain of the EV clusters as a powerful new tool. When renewable generation is high and electricity prices are low, the model can signal the TNO layer to encourage more EVs to charge at stations connected to that part of the grid, effectively storing the excess clean energy. Conversely, when demand is high and prices are spiking, the model can incentivize EVs to discharge power back to the grid (vehicle-to-grid, or V2G), helping to meet peak demand and avoid costly penalties.
The brilliance of the system is its iterative nature. The TNO model calculates traffic flows based on current costs. These flows determine which EVs are at which charging stations and for how long, which in turn defines the flexible operation domain available to the DNO model. The DNO model then calculates a new set of optimal electricity prices and system costs. These updated prices are fed back into the TNO model, changing the “instantaneous unit flow travel cost” and prompting drivers to adjust their behavior. This cycle repeats until the system reaches an optimal, stable state where the decisions of the drivers and the needs of the power grid are perfectly aligned.
To validate their complex model, the researchers conducted a detailed case study based on a reconstructed version of the Nguyen transportation network, a standard benchmark in the field. They simulated a network with four charging stations, each connected to a different part of a simulated power grid. Two stations were located near wind farms, one near a solar farm, and one without any local renewable generation, creating a diverse and realistic scenario. The simulation ran for an 8-hour period, broken down into 5-minute intervals, with three different types of EVs making up the fleet.
The results were striking. When compared to a traditional “shortest path” strategy, the new dynamic coordination model demonstrated profound advantages. In the shortest path scenario, drivers naturally flocked to the two closest charging stations (c3 and c4). This led to severe congestion, with traffic volumes on the connecting roads exceeding their capacity and travel times soaring. Meanwhile, the roads leading to the other two stations (c1 and c2) remained completely unused, a massive waste of infrastructure. This congestion translated into a 47% higher total travel cost for users, primarily due to the immense time lost in traffic.
The power grid side of the story was equally dramatic. Because no EVs were charging at stations c1 and c2, those parts of the grid had zero flexibility. When the wind was blowing strongly at c2, the excess power had nowhere to go and was wasted, incurring high “curtailment penalties.” Similarly, when local demand spiked at c1, the grid had no way to meet it with flexible EV resources and had to resort to expensive and polluting backup generators. The total operational cost for the power grid was significantly higher.
In stark contrast, the dynamic coordination model achieved a balanced, efficient outcome. The “instantaneous unit flow travel cost” successfully guided drivers away from the congested paths and towards the underutilized stations. The traffic flow was distributed much more evenly across the network, with no single road experiencing crippling congestion. This alone led to a substantial reduction in travel time and cost for users.
The impact on the power grid was transformative. EV clusters at all four stations became active participants in grid stability. At station c2, the model successfully directed a large number of EVs to charge during periods of high wind generation, absorbing 67.9% of the otherwise-wasted wind energy. At station c3, it managed to utilize 61.7% of the excess solar power. This ability to “soak up” renewable energy not only reduced waste but also lowered the overall cost of electricity. Furthermore, the widespread availability of flexible EV clusters allowed the grid to meet peak demand more effectively. The model showed a significant increase in the “load supply capacity,” meaning the EVs were able to provide more power to the grid when needed, reducing the need for costly load shedding. Overall, the study found that the coordinated strategy led to a 33% increase in renewable energy consumption and a 25.8% increase in load supply response compared to the shortest path strategy.
The implications of this research are far-reaching. It provides a robust, mathematically sound framework for the future of smart cities. Rather than viewing EVs as a problem to be managed, this work shows how they can be a central part of the solution. By creating a seamless, real-time communication channel between the transportation and power systems, cities can achieve a level of efficiency and sustainability that was previously unattainable. Drivers benefit from shorter, more predictable travel times and potentially lower charging costs. Utilities benefit from a more stable, resilient, and cost-effective grid. Society benefits from reduced greenhouse gas emissions and a faster transition to a renewable energy future.
While the model is highly sophisticated, the underlying principle is elegant: information is power. By providing drivers with a single, intelligent signal—the true, real-time cost of their journey—that reflects the health of both the road and the grid, the system can self-optimize. This represents a paradigm shift from top-down, command-and-control regulation to a bottom-up, market-driven coordination. The success of this approach hinges on the development of advanced software platforms and communication protocols that can calculate these costs in real-time and deliver them to drivers through navigation apps and in-vehicle systems. As the number of EVs continues to grow exponentially, the insights from this research will be invaluable for building the intelligent, integrated infrastructure of tomorrow.
Liu Zhijian, Dai Jing, Yang Lingrui, Automation of Electric Power Systems, DOI: 10.7500/AEPS20230728004