EV Charging Tech Advances Meet Grid Challenges in New Studies
The rapid rise of electric vehicles (EVs) is reshaping transportation, energy systems, and urban infrastructure. As EV adoption accelerates globally, two critical research studies published in June 2024 shed light on both the intelligent navigation strategies that guide drivers to charging stations and the unseen electrical impacts these stations impose on the power grid. These parallel investigations—one focused on optimizing driver decisions and the other on safeguarding grid stability—highlight the dual nature of the EV revolution: while software and artificial intelligence are making EVs smarter, the physical demands of fast charging are testing the limits of century-old electrical networks.
At Southeast University in Nanjing, a team led by Song Yuhang, Chen Yufan, Wei Yanling, and Gao Shan has introduced a novel reinforcement learning framework designed to improve how electric vehicles plan their routes when battery levels are low. Their work, published in the Journal of Automation of Electric Power Systems, presents a method that more accurately reflects real-world driving conditions by rethinking how charging stations are integrated into digital maps used for navigation.
Traditional EV routing algorithms often treat charging stations as simple nodes on a road network—points that can be reached by traveling from one intersection to another. This simplification, while computationally efficient, overlooks a crucial detail: charging stations are rarely located directly at street intersections. Instead, they sit off the main road, requiring drivers to slow down, turn, and sometimes navigate tight parking lots before plugging in. These maneuvers consume time and energy, factors that existing models frequently ignore.
The Southeast University team addresses this gap with what they call the “three-segment expression method.” Rather than viewing a trip to a charger as a single leg between two road points, their model breaks each journey into three distinct phases. The first segment involves moving from a random starting point—such as a home driveway or office parking lot—onto the nearest road segment. The second phase is standard city driving, where the vehicle travels along streets from one intersection to the next. The third and final segment is the approach to a destination, whether that’s a final stop or a charging station.
This seemingly subtle distinction has profound implications for route optimization. By separating the act of entering and exiting the road network from the core navigation process, the model can account for the energy lost during turns, the time spent decelerating into a charging bay, and even the potential delays caused by queuing at a busy station. More importantly, it allows the routing algorithm to treat the decision to charge not as a separate binary choice—“should I charge or not?”—but as a natural part of directional movement. In this framework, choosing to turn into a charging station is no different from choosing to turn left or right at an intersection, which simplifies the computational complexity of the decision-making process.
The researchers implemented this model using two reinforcement learning techniques: Q-learning and Deep Q-Networks (DQN). Both methods allow an AI agent to learn optimal behavior through trial and error, receiving rewards for successful outcomes—such as reaching a destination with minimal time, distance, and cost—and penalties for failures like running out of power. The environment in which the agent operates is built around the three-segment logic, with state variables that include the vehicle’s location, battery level, time of day, accumulated mileage, and total charging cost.
One of the key innovations is how the model handles the charging process itself. When the vehicle arrives at a charging station, the algorithm factors in not just the price per kilowatt-hour but also the expected wait time, which is modeled based on the number of available chargers and current demand. It also calculates the time required to charge to a specified target level, which depends on the charger’s power output and the vehicle’s battery capacity. This holistic view enables the system to compare the total cost—both in time and money—of charging at different stations, even if they are not directly on the most direct route.
Testing the system in a simulated environment representing a 34-square-kilometer area around Southeast University, the researchers found that their approach consistently produced efficient routes under various battery conditions. For example, when starting with only 2.8% battery charge and needing to reach a 3% minimum upon arrival, the algorithm successfully navigated to a charging station, replenished energy, and completed the journey. The DQN model, which uses neural networks to approximate decision values, showed slightly better performance in complex scenarios due to its ability to handle high-dimensional state spaces and learn from past experiences stored in a memory buffer.
Crucially, the study demonstrated that their method outperformed conventional approaches that treat charging stations as simple network vertices. In direct comparisons, the three-segment model resulted in shorter total travel times, lower energy consumption, and reduced charging costs. This improvement stems from the model’s ability to account for the real-world inefficiencies of accessing charging infrastructure, which traditional methods overlook. Moreover, the framework proved to be transferable—when tested in a different urban area, such as around People’s Park in Chengdu, the same algorithm adapted quickly and produced similarly optimized routes without requiring structural changes.
The implications of this research extend beyond academic interest. As automakers and tech companies race to develop autonomous driving systems and smart mobility platforms, accurate and realistic route planning will be essential. A self-driving EV cannot afford to miscalculate its energy needs or fail to account for the time it takes to enter a charging facility. The three-segment expression method provides a more granular and physically grounded way to model these interactions, potentially leading to more reliable and user-friendly navigation systems in future EVs.
However, while smarter software helps drivers find chargers, another study from Changsha University of Science and Technology reveals that the act of fast charging itself poses significant challenges to the electrical grid. Led by Wang Hongbiao, Su Shiping, Hu Yajie, and Ouyang Zhenyu, this research, published in Electrical Measurement & Instrumentation, investigates the transient power quality issues caused by large-scale EV fast charging stations.
Unlike the gradual charging process of a home outlet, fast chargers deliver power at rates exceeding 50 kilowatts, with some reaching over 350 kW. These high-power systems use complex power electronics to convert alternating current (AC) from the grid into direct current (DC) suitable for battery charging. The process involves rectifiers, high-frequency transformers, and DC-DC converters—all of which rely on rapid switching of semiconductor devices like IGBTs and diodes. While efficient, this switching generates electromagnetic transients—brief but intense bursts of electrical noise—that can propagate through the grid.
The research team developed a detailed simulation model of a distribution network incorporating a DC-bus-based fast charging station, a topology increasingly used to integrate renewable energy sources like solar panels. Their analysis focused on two types of disturbances: voltage sags and electromagnetic transients. Voltage sags are short-term reductions in voltage magnitude, typically lasting from a few milliseconds to a few seconds. They can occur when a large load, such as a fleet of EVs starting to charge simultaneously, draws a sudden surge of current from the grid.
In their simulations, the researchers observed that energizing a fast charging station caused an immediate voltage dip at 0.1 seconds into the process. The voltage remained below nominal levels throughout the charging period and only recovered when the station was disconnected. This sustained sag indicates that the grid’s voltage regulation systems may struggle to respond quickly enough to rapid load changes, especially in areas with limited generation or weak network infrastructure.
More insidious are the electromagnetic transients, which manifest as high-frequency oscillations or impulse spikes in the voltage and current waveforms. These disturbances arise from the switching actions within the charger’s power electronics. The study identified a pulse transient at 0.105 seconds and an oscillatory transient at 0.121 seconds after the station was activated. While short-lived, these events can interfere with sensitive equipment such as industrial controllers, medical devices, and communication systems that share the same power feeder.
The severity of these transients increases dramatically under fault conditions. During a two-phase short circuit, for example, the voltage on the affected phases dropped to 0.6 and 0.69 per unit, respectively, while the electromagnetic transients became more intense and occurred at different times—0.372 seconds for the pulse and 0.41 seconds for the oscillation. In a two-phase ground fault, the oscillatory transient appeared even earlier, at 0.302 seconds, and lasted for 5 milliseconds. Notably, the researchers found that transients did not occur during the clearing of certain faults, suggesting that the grounding configuration and fault type play a critical role in transient generation.
These findings have immediate practical implications. As cities install more fast charging stations—often clustered in commercial areas or along highways—the cumulative effect of multiple chargers coming online simultaneously could create a “perfect storm” of voltage sags and transients, particularly during peak charging hours in the morning or evening. Without proper mitigation, such as dynamic voltage restorers, active filters, or on-site energy storage, the reliability of the local grid could be compromised, leading to equipment malfunctions or even blackouts.
The study also underscores the importance of smart charging strategies. Uncontrolled, or “dumb,” charging—where vehicles begin charging as soon as they are plugged in—exacerbates these issues. In contrast, coordinated charging, where the timing and rate of charging are managed through communication with the grid, can smooth out load profiles and reduce the likelihood of severe disturbances. Integrating battery storage at charging stations can further mitigate inrush currents by providing instantaneous power during startup.
From a design perspective, the research suggests that manufacturers can improve power quality by optimizing control algorithms, incorporating soft-start mechanisms, and using higher-quality filtering components. The adoption of wide-bandgap semiconductors like silicon carbide (SiC) or gallium nitride (GaN) could also reduce switching losses and transient emissions, leading to cleaner power delivery.
Regulatory bodies may also need to update standards to address these challenges. Existing guidelines, such as IEEE 1159 and IEC 61000, provide general recommendations for power quality but may not fully capture the dynamic behavior of fast charging stations. The authors recommend that future revisions include specific provisions for EV charging infrastructure, particularly regarding transient immunity and emission limits.
Together, these two studies illustrate the multifaceted nature of the EV transition. On one hand, intelligent algorithms are making it easier for drivers to navigate the charging landscape. On the other, the physical reality of drawing massive amounts of power from the grid demands new engineering solutions to maintain stability and reliability. The work of Song Yuhang, Chen Yufan, Wei Yanling, and Gao Shan from Southeast University offers a more realistic and efficient way to plan EV journeys, while the research of Wang Hongbiao, Su Shiping, Hu Yajie, and Ouyang Zhenyu from Changsha University of Science and Technology highlights the hidden costs of fast charging on the electrical system.
As the world moves toward electrification, both aspects must be addressed in tandem. Smarter vehicles need smarter grids. The integration of EVs is not merely a matter of replacing internal combustion engines with batteries—it is a transformation of the entire energy ecosystem. Ensuring that this transformation is sustainable, efficient, and resilient will require continued innovation not only in vehicles and charging technology but also in the infrastructure and policies that support them.
Song Yuhang, Chen Yufan, Wei Yanling, Gao Shan, Southeast University, Journal of Automation of Electric Power Systems, DOI: 10.7500/AEPS20230621004; Wang Hongbiao, Su Shiping, Hu Yajie, Ouyang Zhenyu, Changsha University of Science and Technology, Electrical Measurement & Instrumentation, DOI: 10.19753/j.issn1001-1390.2024.06.021