Electric Car Models Need More Profiles for Accurate Grid Planning

As electric vehicles become more common, understanding their impact on the power grid is crucial for planning a reliable and cost-effective energy future. Researchers have found that the way these vehicles are represented in computer models can significantly skew , leading to inaccurate predictions about costs and infrastructure needs. This study focuses on a specific modeling approach that uses individual battery electric vehicle (BEV) profiles instead of aggregated data, highlighting a critical trade-off between detail and computational efficiency.

The key finding is that including too few BEV profiles in power sector models can lead to distorted outcomes, such as overestimating costs or misjudging the need for storage capacity. For example, when modeling a fleet of 15 million electric vehicles, using only 5 profiles instead of at least 20 can inflate estimated costs by 2.5 times for smart charging and underestimate cost savings by four times for bidirectional charging. The researchers discovered that beyond a certain point, adding more profiles does not significantly improve accuracy but drastically increases computation time, with runtimes jumping from under an hour to 10 hours or more when exceeding 80 profiles in some cases.

Ology involved generating synthetic BEV profiles using an open-source tool called emobpy, which creates detailed time series for mobility, driving electricity consumption, and grid availability at a 15-minute resolution. These profiles were then integrated into an open-source power sector model, DIETER, which optimizes costs over a year of hourly data. The study tested scenarios with varying numbers of profiles, from 5 to 120, and analyzed their effects on runtime and , including cost differences and optimal capacity mixes for technologies like stationary Li-ion batteries and natural gas turbines.

From the data show that runtime increases moderately with more profiles up to a threshold, but beyond 60-80 profiles, it spikes sharply, as illustrated in Figure 1. For cost accuracy, Figure 2 demonstrates that including at least 20 profiles stabilizes cost differences, with smart charging adding about 50 euros per BEV per year and bidirectional charging saving around 100 euros per BEV per year. Capacity , shown in Figures 3 and 4, are even more sensitive: optimal stationary Li-ion battery capacity can vary widely with too few profiles, leading to misleading conclusions about investment needs. The study also found that isolated long-trip events in profiles, when scaled up, create large charging spikes that distort capacity outcomes, as detailed in Figure 5.

Of this research are practical for energy planners and policymakers, as it provides a clear guideline: for fleets of 5 to 20 million electric vehicles, each BEV profile should represent between 200,000 and 250,000 vehicles to ensure accurate without excessive computation time. This rule of thumb helps balance the need for detailed modeling with the constraints of real-world computational resources, potentially improving grid planning and investment decisions. emphasize that while cost stabilize with fewer profiles, capacity and dispatch outcomes require more detail, especially for technologies like batteries and natural gas.

Limitations of the study include the use of a fixed pool of 200 BEV profiles, which may affect variability comparisons across different numbers of profiles. The researchers note that runtime values depend on hardware performance and should be interpreted cautiously, as better or worse equipment could shift the practical thresholds. Additionally, the study focuses on one modeling approach and does not compare it to aggregated s, suggesting future research should explore these comparisons and investigate how to design representative profile sets for nationwide applications.