Decentralized Valley-fill Charging Control of Large-population Plug-in Electric Vehicles

Optimal charging control of large-population autonomous plug-in electric vehicles (PEVs) in power grid can be formulated as a class of constrained non-linear time-variant optimization problems. To overcome the computational complexity of this class of optimization problems, the author and his collaborators proposed a game-based decentralized control method such that individual agents update their best charging strategies simultaneously with respect to a common electricity price signal which is determined by the total demand in the grid. Due to the heterogeneity of individual PEVs, the game systems converge to a nearly valley-fill NE strategy with nontrivial deviation costs due to the heterogeneity property of individual PEV charging characteristics. In this paper the author proposed a novel algorithm to implement the optimal decentralized valley fill strategies for the charging problems of the PEV population which is composed of disjoint homogeneous subpopulations. The author introduces a cost which penalizes against the deviation of strategy of individual agent in a subpopulation from the average value of the subpopulation. It can be shown that in case that the update algorithm converges, the system reaches the optimal valley-fill equilibrium strategy where the introduced agent deviation cost vanishes. Simulation examples are used to illustrate the results developed in this paper.