Distributed Energy Management of Electric Vehicles Under Prediction Based Asynchronous ADMM

Energy management for electric vehicles (EVs) requires controlling their power consumption in reference to a predetermined objective. However, optimality of energy management strategies can also depend upon extrinsic factors such as communication synchronization between EV agents and a coordinator. This paper proposes a distributed scheduling algorithm with computationally heterogeneous EV agents under learning-aided alternating direction method of multipliers (ADMM). The computational or communication delay between neighbouring EV agents is handled using asynchronous update of Lagrangian parameter. Moreover, an auto-regressive prediction model is developed for estimating the information lost due to communication disruption between EV agents. This ensures all agents are exempt from strict synchronization requirements between each other, thereby improving the time-complexity of the distributed algorithm. Simulations run for a typical EV charging station under a contractual power procurement limit, show that proposed algorithm reduces the iterations required for execution, while ensures improved optimality at convergence compared to uncorrected asynchronized ADMM.