In this work, we conduct a data-driven simulation of ultra-fast charging station roll-out across Beijing, Shanghai, and Guangzhou, leveraging over 760,000 real-world public charging records. Our results show that large-scale deployment of these chargers increases the peak-to-valley differences of the total public charging load, substantially stressing grid flexibility and adequacy margins. We also find that while integrating on-site energy storage can smooth short-term load volatility, it can simultaneously exacerbate demand surges during electricity price changes. The analysis underscores that without new controls, deploying fast chargers with energy storage can significantly raise peak loads, highlighting the need for strategies to unlock their grid-stabilizing potential.
According to China’s New Energy Vehicle Industrial Development Plan for 2021 to 2035, by 2030 and 2060, the penetration rate of NEVs, primarily EVs, in vehicle ownership needs to reach 20% and 100%, respectively. To meet the rapidly growing demand for EV charging, the deployment of UFCS is essential. Our analysis focuses specifically on the charging demand of light-duty (LD) passenger vehicles, as they constitute the predominant user base for public urban charging infrastructure, including the UFCS central to our study.
The planning of UFCS in major Chinese cities reveals ambitious targets. For instance, Shenzhen aims to construct 300 UFCSs by 2025, increasing to 1,000 by 2030. Similarly, Beijing plans to establish 500 UFCSs by 2024 and reach 1,000 by 2025. Each station is typically equipped with 1 or 2 ultra-fast chargers with power outputs above 480 kW, along with 2 to 7 fast chargers with power outputs exceeding 250 kW. While current LDVs typically support DC fast charging rates between 60 and 250 kW, the deployment of 480 kW+ UFCS anticipates future advancements in vehicle battery technology, enabling charging times of 5-10 min. Under this trend, it is estimated that high-power chargers (240 kW+) could make up 44% of public chargers by 2030 and 80% by 2035.
Figure 1a illustrates a data-driven framework to simulate and model the dynamic fluctuations in EV charging load under UFCS deployment, assessing the impact of UFCS expansion on grid load (Fig. 1b shows the principle of how UFCS affects grid load). This study focuses on three major cities: Beijing, Shanghai, and Guangzhou. The basic characteristics of these cities and the TOU policy are summarized in Table 1. Although each city has implemented a TOU strategy, the actual charging prices at different stations may still vary; the distribution is provided in Supplementary Note 1. It is important to note that electricity tariffs for commercial and industrial users in China typically follow a two-part system, comprising an Energy Charge (often structured as TOU pricing, as detailed in Table 1) based on energy consumption, and a Capacity Charge based on peak power demand. However, under current supportive policies aimed at promoting EV infrastructure development, public centralized EV charging and swapping stations in cities like Beijing, Shanghai, and Guangzhou are temporarily exempted from the Capacity Charge.
The methodology first involves categorizing the charging stations and assuming that commercial and workplace stations will gradually be upgraded to UFCS. Three simulation scenarios were defined for the years 2030, 2035, and 2050. For each simulation scenario, a specified number of these stations are modeled as upgraded, and the upgraded UFCS are integrated with the non-upgraded stations to simulate the overall grid load. Additionally, we introduced an optimization model to simulate the deployment of on-site ESS, with unregulated market and load-control scenarios. The simulation results provide detailed spatiotemporal profiles of charging loads with minute-by-minute resolution, offering insights into the interplay between UFCS deployment and grid dynamics.
In detail, charging order data from public charging stations serves as the primary input for the study. The dataset includes 1,702 charging stations (representing ~8.6% of the estimated 19,830 total public stations across the three cities, see Table 1) and a total of 769,225 charging orders recorded over a 31 day period (from January 18 to February 17, 2024) across the three cities. Figure 1c shows the load of different types of charging stations in the three cities and the total charging load at the city level. Analysis of this 2024 data reveals that ultra fast chargers meeting the 480 kW specification were virtually absent in our sample (only one such charger was present across the three cities, with 94.8% of chargers being 120 kW or less). Given this observation, we consider the current charging landscape represented by the 2024 dataset to be effectively free of the widespread adoption of next generation 480 kW UFCS, and it is therefore utilized as the baseline (pre-UFCS adoption in 2024) for our study. These data reflect real-world charging demand, exhibiting distinct temporal patterns influenced by factors such as TOU pricing schedules (detailed analysis in Supplementary Note 2). These samples are representative of high quality, as they were sourced from a single major charging station platform rather than from different platforms. Assuming that the dataset represents a uniform sampling of public charging stations, we scaled the data to align with the actual distribution of charging stations within each city to represent the citywide charging load of public charging stations. This scaling process preserves the aggregated charging patterns observed in our sample, which reflect user behaviors under the current TOU schedules and the existing station-specific price variations. Our simulation then focuses on the impact of upgrading charging speeds on the power load under these established demand patterns.
Considering the constraints of urban land use, deploying a large number of completely new UFCS sites is challenging. Therefore, our UFCS adoption assumption for future scenarios involves upgrading existing charging infrastructure rather than building entirely new stations. Given the varying demand characteristics by location and land use, the choice of upgraded points significantly impacts the resulting load profile. For example, daytime charging demand tends to concentrate in employment centers, whereas residential areas typically see higher evening demand. For the UFCS simulation, we first categorized the existing charging stations into “Residential,” “Commercial,” “Workplace,” and “Other” based on station address and name information (see Methods section for classification details; the load profile of different types of stations is provided in Supplementary Note 4). As fast charging stations are generally more often located in commercial districts and workplaces to maximize revenue, these categories were prioritized as candidate sites for upgrading to 480 kW UFCS in the initial scenarios.
Three scenarios were established based on city targets: deploying 1000 UFCSs by 2030, 2,000 by 2035, and replacing all public charging stations with UFCSs by 2050. In the 2030 and 2035 scenarios, commercial and workplace charging stations were prioritized for upgrade, whereas the 2050 target encompasses the conversion of all public charging stations to UFCS, with each UFCS configured with two 480 kW ultra-fast chargers. To reduce random errors, we performed multiple sampling tests for each scenario to reflect overall trends.
UFCS distortion of TOU-based regulated price signals
TOU tariffs have emerged as a particularly effective tool, clearly influencing when and how EV owners choose to charge their vehicles. The effectiveness of TOU in shifting load is substantiated by comparing observed charging patterns with a simulated baseline scenario without TOU influence (see Supplementary Note 3 for details on the Shanghai case study), based on the method proposed in our previous study. This comparison reveals that without TOU, charging demand would likely peak during post-commute hours, whereas the actual data shows a significant shift towards off-peak and valley price periods. Our analysis of EV charging patterns across Guangzhou, Beijing, and Shanghai illustrates the significant impact of TOU tariffs (as is shown in Fig. 2). These tariffs, which offer lower rates during off-peak hours, have successfully shifted the bulk of EV charging to nighttime, reducing the strain on the urban power supply during high-demand periods. This shift is evident across all three cities, regardless of local differences in the specific timing of these tariffs. For instance, in Guangzhou, the valley hours with low electricity price period extend from midnight to 8:00 AM, during which there is a noticeable increase in charging activity shortly after midnight and again just before the period ends. Although Beijing and Shanghai have earlier end times for their off-peak periods, the pattern of increased charging activity during these hours remains consistent. This demonstrates a widespread behavioral adaptation among EV owners, who are keen to take advantage of lower electricity prices, thereby underscoring the clear effectiveness of TOU tariffs in managing energy consumption patterns in urban China.
While the TOU tariffs have been successful in redistributing load times, the integration of UFCS presents a challenge to this established pattern. UFCS allow for quicker battery replenishment, which, while convenient, leads to intense, short-term demand spikes. These load peaks are particularly acute during the shift from peak to off-peak tariffs, as many users wait to plug in their vehicles until the lower rates take effect. Such synchronization of charging behavior results in a dramatic increase in electricity demand exactly at the onset of the off-peak period. The magnitude of these spikes is further exacerbated by the rapid charging capabilities of UFCS, which, if adopted at a large scale, could see peak load demands doubling compared to non-UFCS scenarios.
The introduction of UFCS is anticipated to significantly amplify the variability of charging loads, leading to a pronounced increase in the short-term peak-to-valley load differences in the power grid. High peak-to-valley differences require the grid to adjust power output frequently to accommodate short-term fluctuations, which may induce instability in grid frequency and voltage, increasing the risk of grid faults and raising operational and maintenance costs. This instability ultimately threatens the reliability and stability of the power supply. To quantify these fluctuations, we use two key metrics: the daily peak-to-valley difference (the difference between the maximum and minimum load observed over a 24-hour period, representing the overall daily load swing) and the maximum hourly peak-to-valley difference (the maximum difference between the peak and valley load observed within any one-hour window during the day, quantifying short-term load volatility). Using simulation results, we calculated these differences in charging demand under various UFCS deployment scales and compared them with baseline EV public charging load levels (pre-UFCS adoption in 2024), as illustrated in Fig. 3.
As illustrated in Fig. 3a,b, expanding the UFCS network increases the daily peak-to-valley load gap almost linearly in all three cities. Up to roughly 1,500 stations, the trend is similar, but beyond that point, Guangzhou’s curve rises much faster than Shanghai’s, which in turn exceeds Beijing’s. This is attributable to the inherently higher baseline volatility of its charging load, driven by Guangzhou’s more frequently adjusted daytime TOU price periods.
From the daily peak-to-valley difference perspective (Fig. 3a), by 2030, deploying 1,000 UFCS is projected to increase the total public charging load peak-to-valley difference by up to 13.70% on a daily average basis and by 19.72% for short-term hourly differences compared to the non-UFCS baseline. By 2035, deploying 2,000 UFCS may elevate these differences by up to 31.61% (daily) and 46.15% (short-term). When the number of stations reaches 3,000, the increase in peak-to-valley difference is expected to reach 49.22% (daily) and 75.12% (short-term). Furthermore, due to the inherently higher volatility of charging load under the existing TOU policy in Guangzhou compared to other cities, the risks associated with UFCS deployment are significantly greater for Guangzhou than for the other two cities.
As illustrated in our simulation, once a valley pricing period begins, UFCS operators and EV users converge in their charging activities, triggering “secondary peaks” or even new “late-night peaks.” This pattern of demand reveals a critical mismatch between the fast-paced adoption of high-power ultra-fast charging technologies and the existing electrical grid and market designs, which still largely cater to traditional, steadier consumption patterns. The current market mechanisms, which prioritize short-term economic gains, fail to align with the long-term sustainability goals of environmental policies promoting EVs and renewable energy sources. This discord not only increases the risk of grid overload but also hinders investment in renewable energy, as the market fails to provide stable and favorable signals for such investments due to the intermittent nature of renewables juxtaposed against the peak demands of UFCS.
Grid risks amplified by UFCS under TOU pricing
The widespread deployment of UFCS under current TOU tariffs is expected to significantly increase peak charging power demand and short-term load volatility, introducing potential risks to power grid stability and adequacy. These risks primarily manifest as potentially exceeding the grid’s flexibility resources (Regulating Reserves) and its allocated power capacity (Capacity Reserves).
Power systems maintain stability and adequacy through various ancillary services, including Regulating Reserve (RR) and Capacity Reserve (CR). RR provides rapid-response capabilities in seconds to minutes to continuously balance supply and demand, managing fluctuations and maintaining system frequency; it addresses the grid’s need for flexibility against rapid load ramps. CR represents available capacity generation or demand-side resources held in reserve to ensure the system can meet peak load and handle contingencies; it addresses the grid’s need for adequacy to meet the magnitude of demand. The unique characteristics of UFCS-high power and potential for synchronized charging-can significantly increase the demand for both RR and CR compared to baseline conditions. It is important to recognize that while RR and CR are typically planned and dispatched by the grid operator across broader regional control areas (e.g., provincial level), the significant new stress and demand for these reserves originates from concentrated load centers. Large-scale UFCS deployment is primarily an urban phenomenon, and our simulations appropriately focus on quantifying the impact and the resultant demand for reserves generated at the city level.
A significant risk associated with UFCS deployment is the possibility of depleting the system’s RR – the fast-responding flexibility margin (AGC units, BESS, responsive load) that grid operators rely on to correct second-to-minute imbalances and keep frequency within tight bounds. Risk occurs when rapid or sudden variations in charging demand exceed the grid’s capacity to regulate and maintain stable operation via available RR, potentially leading to frequency instability or voltage fluctuations. Our simulations show that UFCS demand spikes-especially those clustered immediately after morning and evening TOU price transitions-can momentarily push the incremental balancing requirement above the available RR, creating a non-negligible probability of frequency or voltage excursions. Figure 4a-f plots, for Beijing, Guangzhou, and Shanghai in 2030 and 2035, the time-of-day profile of the probability of exceeding three illustrative RR thresholds, highlighting the intervals of greatest operational stress.
Furthermore, the high-power nature of UFCS places substantial short-term stress on local grid capacity, creating a risk of exceeding the CR. We define the baseline as the maximum simulated public-charging load in a non-UFCS scenario and impose three illustrative reserve buffers-15%, 20%, and 25%-in line with leading planning-reserve guidelines, which all recommend maintaining total reserve margins in the 15-20% range of system peak load. Applying these percentages specifically to the public-charging component allows us to isolate the incremental capacity stress introduced by UFCS. Figure 4g-i plots, for each city, the probability that the aggregate charging load surpasses the three CR thresholds as UFCS deployment scales up. The results reveal that 15% CR becomes insufficient once 1,000 UFCS are deployed in the 2030 case, as all three cities register a non-negligible chance of reserve exhaustion. 25% CR remains adequate only up to 1,500 UFCS, as beyond that level, approaching the 2035 target of 2,000 stations, the exceedance probability again rises sharply. Consequently, adherence to the current Chinese planning standard of 20% total reserve would leave the grid increasingly exposed between the 1,000 station and 2,000 station milestones, signaling that capacity-reinforcement or alternative flexibility measures will become imperative in the 2030-2035 window.
These results demonstrate that the large-scale integration of UFCS under the current TOU pricing model significantly increases grid risks, specifically the risk of exceeding RR due to heightened volatility and the risk of exceeding CR due to elevated peak public charging demand. This heightened risk arises from pronounced demand spikes during periods of lower tariffs, leading to potential challenges in managing rapid load variations and ensuring sufficient capacity. These issues necessitate urgent and comprehensive enhancements to grid management strategies, including the potential expansion of CR allocated for charging infrastructure and improvements in grid flexibility, potentially supported by enhanced RR or sophisticated demand response programs. Adapting these strategies is essential to effectively manage the substantial changes in energy consumption patterns and maintain the reliability and efficiency of the electrical grid as UFCS becomes more prevalent.
Risk of grid instability from improper ESS management in UFCS
The integration of ESS with UFCS primarily aims to enhance profitability through energy arbitrage-buying energy at lower prices during off-peak hours and selling during peak prices. However, while this approach can significantly enhance the economic returns of charging station operations, it also introduces new challenges in terms of grid stability. The use of ESS allows charging stations to disconnect from grid electricity prices temporarily, charging their batteries when it is cheapest and discharging during periods of highest electricity prices. This behavior can lead to increased load fluctuations as stations switch between drawing power from the grid and supplying power from their storage systems.
To evaluate the operational dynamics of UFCS equipped with on-site ESS, we developed a model to simulate operation strategies of ESS-integrated UFCS. This simulation investigates how on-site ESS impacts the total charging load under various market conditions. We focus on three key scenarios:
The first is an Unregulated Market Scenario, where stations operate solely based on market dynamics to maximize cost savings. The second is a Capacity Charge Scenario, in which the standard TOU tariff is augmented with a capacity charge to incentivize stations to shave peak load. The third is a Demand Response Scenario, where grid operators implement direct control measures to ensure grid stability.
Figure 5 presents the simulation results across three cities for future scenarios under the following configurations: UFCS without ESS, UFCS with ESS under different scenarios. Generally, ESS integration significantly reduces daytime power demand and fluctuations compared to the UFCS-only case, leading to smoother daytime load curves (Fig. 5a-i) and enhanced grid stability during peak hours.
However, the Unregulated Market scenario reveals a critical risk: a substantial increase in nighttime demand and volatility. As shown in Fig. 5a-i, the transition from peak/shoulder to off-peak pricing under the TOU tariff (around hour 23 in Beijing, hour 22 in Shanghai, and hour 0 in Guangzhou) triggers a synchronized surge in charging activities as stations replenish their ESS during newly available low-price periods. Without regulatory constraints, ESS exacerbates this issue, concentrating energy demand. This potential grid risk is quantified in Fig. 5j,k. Deploying 1,000 UFCS stations with ESS under this scenario significantly increases peak demand: in 2030, the peak load rises by 71.4% in Beijing (0.239 GW to 0.410 GW), 84.6% in Guangzhou (0.308 GW to 0.569 GW), and 83.5% in Shanghai (0.306 GW to 0.562 GW) compared to the baseline without ESS. By 2050, under full deployment, these unregulated peaks could reach 4.31 GW, 3.17 GW, and 3.67 GW respectively, approximately 7.5, 3.9, and 4.6 times higher than without ESS. The average hourly peak-to-valley difference also increases compared to the UFCS-only case (Fig. 5k).
This nighttime surge presents a dual nature. On the negative side, the onset of the low-price period under the TOU tariff overlaps with the residual nighttime load peaks, thereby exacerbating the risk of grid overloading. On the positive side, the relatively extended duration of the low-price window provides operational flexibility, enabling the deferral of charging activities without compromising energy replenishment objectives. Given these conditions, the implementation of control strategies during this period is both necessary and practically achievable.
In contrast, both the Capacity Charge and Demand Response scenarios effectively mitigate these nighttime surges and improve overall load profiles (Fig. 5a-i). Comparative analysis using key metrics (Fig. 5j,k) highlights their distinct mechanisms and trade-offs. The Capacity Charge scenario, by introducing a financial penalty for high peak demand, incentivizes users to autonomously adjust ESS operations. This leads to constructive grid behavior, suppressing peak loads and mitigating fluctuations without direct intervention. It aligns individual economic optimization with grid stability objectives.
However, under the fully deployed ESS scenario in 2050, the Capacity Charge strategy faces limitations. To avoid creating new, penalized nighttime peaks, stations might limit their nighttime charging, potentially underutilizing ESS capacity and reducing its effectiveness in smoothing daytime load. Conversely, Demand Response scenarios, through direct control, can ensure ESS is fully charged during off-peak periods without creating uncontrolled peaks. This allows ESS to more effectively minimize reliance on grid energy during daytime operations and maximize its contribution to mitigating daytime grid stress.
These findings highlight that both pricing-based (Capacity Charge) and control-based (Demand Response) strategies possess distinct advantages and limitations. Careful and coordinated design of incentive and management mechanisms will be essential to fully harness the potential of ESS in stabilizing future power systems under deep UFCS deployment.
