Simultaneous Electric Powertrain Hardware and Energy Management Optimization of a Hybrid Electric Vehicle using Deep Reinforcement Learning and Bayesian Optimization

In order to maximize the energy efficiency of hybrid vehicles, both the electric powertrain hardware and the energy management have to be optimized. Focusing on a certain deterministic driving cycle during design can lead to higher consumption in customer use, as real driving cycles differ to the deterministic cycles. This contribution uses a modern Deep Reinforcement Learning (DRL) energy management, which is able to optimize controls for real stochastic vehicle use. Additionally, a Bayesian Optimization sequentially operating with the DRL selects suitable hardware configurations. The best configuration achieves a fuel consumption reduction of 10% for a mild hybrid electric vehicle.