How to charge Electric Vehicles (EVs) efficiently when electricity price varies?

What is it about?

Consider a control system with time-varying and time coupling constraints and time-varying costs, such as a EV charging system. Assuming we have predictions about system costs and aggregate feedback of constraints, we provide regret analysis of a model-predictive control-based approach, termed "penalized predictive control". Our approach combines learning and control and we demonstrate that the PPC can be used to efficiently charge electric vehicles, even when the electricity prices vary.

Featured Image

Photo by Andrew Roberts on Unsplash

Photo by Andrew Roberts on Unsplash

Why is it important?

We provide theoretical analysis of control systems with time-varying and time coupling constraints, which are notoriously challenging to analyze. We illustrate that it is possible to do information aggregation for those constraints via a systematic approach. More interestingly, the control algorithm that we consider, termed "penalized predictive control" (PPC), is a combination of model predictive control and reinforcement learning. Our main result shows that it is possible to achieve a sub-linear regret, given enough predictions. As a real-world application, we demonstrate that the PPC can be used to efficiently charge electric vehicles and it achieves a lower overall cost compared with pure MPC.