Recurrent Neural Network-based Predictive Energy Management for Hybrid Energy Storage System of Electric Vehicles

Electrified vehicles (EVs) are one of the promising technologies for promoting the clean energy revolution. The hybrid energy storage system (HESS), which has multiple energy storage components, requires an energy management strategy (EMS) to reasonably allocate the overall power demand to sub-components. In this paper, a new predictive EMS is proposed to allocate the overall demanded current for the HESS of an EV. More specifically, an end-to-end prediction method is proposed using recurrent neural networks to forecast the bus current demand. Under power demand prediction, a rule-based EMS is developed to allocate the loads between the supercapacitor and Li-ion battery via multi-objective optimization. The proposed EMS is validated with respect to prediction accuracy and other metrics provided by the 2022 IEEE VTS motor challenge. And simulation results demonstrate the superior performance of the proposed algorithm, compared to other conventional methods.