Scalable Optimization for Mobility-Aware Coordinated Electric Vehicle Charging in Distribution Power Networks

A team from UC Berkeley, led by Yi Ju, Lunlong Li, Jingchun Wang, and Scott Moura, has published a paper introducing MAC (Mobility-Aware Coordinated EV charging). This framework tackles a critical planning question for utilities and city planners: what is the absolute maximum benefit that can be extracted from flexible EV charging to avoid overloading and costly upgrades to the local power grid? The core innovation is a shift from per-session charging optimization to a full "mobility horizon" approach. Instead of forcing an EV to fully recharge immediately after a trip, MAC only requires the battery's state-of-charge (SOC) to be sufficient for the driver's upcoming travel needs. This creates a much larger pool of feasible, flexible charging schedules.

To make solving this massive optimization problem tractable at a regional scale—involving millions of variables with spatial and temporal coupling—the team developed a computationally scalable solution. They used an ADMM-based decomposition algorithm with custom subproblem solvers. This method also has a practical economic interpretation: the algorithm's dual variables can act as locational-temporal price signals, creating a decentralized market where individual EV charging decisions naturally align with the grid's social optimum. In a simulation for the San Francisco Bay Area with a future 30% EV adoption rate, using high-resolution mobility data, MAC showed it could significantly reduce the risk of distribution network overloads compared to unmanaged charging. The work provides a crucial, certifiable upper-bound benchmark for infrastructure planning, showing how much upgrade cost can potentially be deferred by smartly leveraging EV flexibility.