When a hurricane is coming, can a city with lots of electric cars top-up all those batteries fast enough without frying the local power grid? Three Georgia Tech engineers—Alejandro Owen Aquino, Samuel Talkington and EPIcenter affiliate Daniel Molzahn—built a computer model to find out.
Using realistic-but-not-real data from Greensboro, North Carolina, the model shows that sticking strictly to every safety limit means charging takes about 18 hours. A long midday pause is needed because household demand peaks and cars must stop charging to keep voltages inside their safe band.
The authors tested giving grid operators a small “wiggle room” buffer—letting voltages drift modestly past their normal limits for a short time. With that flexibility, one of the tested scenarios with the highest tolerance for network voltage violations enabled all cars to finish charging in roughly 10 hours, a ~46 percent time-savings, and the midday pause disappears. Another scenario with a more modest increase in violation tolerance resulted in a ~9% decrease in charging times.
A key feature of the paper is how it approaches a computationally intractable problem, i.e. optimizing evacuation times when combined with the need to charge many electric cars. The optimization routine developed by Dan and his coauthors uses an “iterative constraint generation scheme” to find a workable schedule in only a handful of iterations, reducing computation time and making the problem manageable.
The takeaway for emergency managers is straightforward: a carefully chosen buffer can cut evacuation-charging time while keeping overload risks predictable. As the authors put it, “the formulation provides flexibility … to strike a balance between minimizing total EV charging time and safeguarding the power system.” In plain terms, cities can leave sooner—and safer—if they decide just how much grid slack they are willing to tolerate before the storm hits.
This summary was written with the assistance of Microsoft Copilot on July 1st, 2025. Its content was edited and verified by EPIcenter staff and affiliates.
Read the full paper: https://arxiv.org/pdf/2311.16975