What is it about?
Electric vehicles have flexibility in when, and sometimes at what rate, they can charge. Aggregating this flexibility across hundreds of vehicles can enable them to act as a virtual power plant (VPP) that can participate in the wholesale electricity market, like generators do. Electric vehicle owners could save money and even earn a profit by optimizing when they charge and by providing grid services, like frequency regulation. The first step to making this possible is modeling the aggregate energy and power capacity of the vehicles in an efficient way that can be easily communicated while preserving privacy. Many models have been developed, but they might perform better or worse when driver behavior is uncertain. This paper provides a way to compare different electric vehicle aggregation models and their performance when there is uncertainty in when drivers arrive and depart and in how much they need to charge.
Featured Image
Photo by Sevadis on Unsplash
Why is it important?
In addition to comparing four models that haven't been compared previously, this work provides a way to understand how models perform under uncertainty. Additionally, the results could provide a benchmark for more complex models that attempt to account for uncertainty.
Perspectives
We’re entering a new age in which electricity consumers can participate in the market as VPPs. We need a way to model VPP energy and power capacity efficiently and accurately. Every model has an ideal application; here we test simple models and how they perform under uncertainty. There’s an interesting trade-off in performance between capturing exact bounds based on assumed characteristics versus being a little more conservative, which can provide benefits when the system is subject to uncertainty. Without detailed uncertainty modeling, however, these methods are ad-hoc.
Jenna Knudtson
University of Michigan
Read the Original
This page is a summary of: Comparing the Performance of Electric Vehicle Aggregation Models Under Driver Behavioral Uncertainty, October 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3847238.3849048.
You can read the full text:
Contributors
The following have contributed to this page
