Optimal Energy Management for Connected Hybrid Electric Vehicles Based on Bayesian Optimization

Abstract

Abstract Due to the limitations in computing power and information, globally optimal control for hybrid electric vehicle (HEV) is difficult to achieve. The development of connectivity technology brings new opportunities for vehicle control to achieve quasi-global optimization by using future information. In this paper, an optimal energy management strategy for a parallel HEV is proposed and applied in the cloud computing side, based on Bayesian optimization (BO) algorithm. By seeking for the minimal accumulated equivalent fuel consumption (FC), the parameters energy management is optimized iteratively. Simulation results show that the proposed method can can converge within 5 steps and reduce the equivalent FC by more than 10% over manually tuned controller. In addition, the impacts of prediction horizon and accuracy on the FC optimization is quantitively investigated.