Data-driven, Internet-inspired, and Scalable EV Charging for Power Distribution Grid

Abstract: Electric vehicles (EVs) are finally making their way onto the roads. However, the challenges concerning their long charging times and their impact on congestion of the power distribution grid are still waiting to be resolved. With historical measurement data, EV chargers can take better-informed actions while staying mostly off-line. Proposed solutions that depend on heavy communication and rigorous computation for optimal operation are not scalable. The solutions that do not depend on power distribution topology information, such as Droop control, are more practical as they only use local measurements. However, they result in sub-optimal operation due to a lack of a feedback mechanism. This study develops a distributed and data-driven congestion detection methodology embedded in the Additive Increase Multiplicative Decrease (AIMD) algorithm to control mass EV charging in a distribution grid. The proposed distributed AIMD algorithm performs very closely to the ideal AIMD regarding fairness and congestion handling. Its communication need is almost as low as the Droop control. The results can provide crucial insights on how we can use data to reveal the inner dynamics and structure of the power grid and help develop more advanced data-driven algorithms for grid-integrated power electronics control.