DISTRIBUTED ALGORITHMS FOR ELECTRIC VEHICLE CHARGING
Abstract
Coordinated charging of plug-in electric vehicles (PEVs) can effectively mitigate
the negative effects imposed on the power distribution grid by uncoordinated charging.
Simultaneously, coordinated charging algorithms can accommodate the PEV user’s needs
in terms of desired state-of-charge and charging time. In this work, the problem of tracking
an arbitrary power profile, by coordinated charging of PEVs, is formulated as a discrete
scheduling process, while accounting for the heterogeneity in charging rates and restricting
the charging to only the maximum rated power. Then, a novel distributed algorithm is
proposed to coordinate the PEV charging and eliminate the need for a central aggregator.
It is guaranteed to track, and not exceed, the power profile imposed by the utility, while
maximizing the user convenience. A formal optimality analysis is provided to show that
the algorithm is asymptotically optimal in the case of homogeneous charging, while it has
a very small optimality gap for the heterogeneous case.
The work also discusses techniques for achieving aggregate load profiles with minimum
variance and peak in both centralized and decentralized settings. A theoretical analysis
that proves that peak minimization is inherently achieved as part of an variance minimization
process has also been presented.
The impact of interrupted and uninterrupted electric vehicle charging on the aggregated
load profile has been explored. The variance of the aggregate load profile is used
as the metric for measuring valley filling capability of the scheduling scenarios. It is
shown, that for low penetration levels (up to 30%), interrupted charging strategies result
in considerably lower variance values on the aggregated load profile as compared to the
uninterrupted case. It is also shown that the policy used for deciding the PEV priority for
scheduling has almost no impact on these variance values. All the proposed algorithms
and the related analysis are accompanied by numerical simulations under realistic charging
scenarios.